Large Scale Heat Pump Installation Effects on the Interaction between Buildings and the Electric Grid

EIT InnoEne gy - MSc SELECT
En i onomical Pa hways o Sus ainable Ene gy Sys ems
Co-Simula ion Pla o m o he Assessmen o he
In e ac ion Be ween Hea Pumps and he Low
Vol age G id a he Feede Le el
Au ho : Jalomi Maayan Ta di
Di ec o : F ancisco Diaz-Gonzalez
Supe iso : Jaume Salom
Call: Sep embe 2018
Submi ed in pa ial ul ilmen o he equi emen s o he deg ee o
Mas e in Ene gy Enginee ing
Uni e si a Poli ècnica de Ca alunya – Ba celona Tech
Ca alonia Ins i u e o Ene gy Resea ch - IREC
I | P a g e
Abs ac
In he pas yea s Hea Pumps (HPs) ha e expe ienced signi ican g ow h a es, and his
elec i ica ion o esiden ial hea ing could con ibu e o deca bonizing he ene gy sec o in many
coun ies, as well as p o ide lexibili y po en ial o he g id. Howe e , he widesp ead use o HPs,
subs i u ing uel boile s, could lead o g id conges ion p oblems, pa icula ly a he low ol age
(LV) le el since he dis ibu ion equipmen may no be adequa ely dimensioned. This hesis will
p esen a co-simula ion en i onmen which uses he Func ional Mock-up In e ace (FMI) s anda d
wi hin a Py hon amewo k. Two case s udies a e p esen ed, one in Spain and one in Sweden,
ep esen ing he sou he n and no he n Eu opean clima es. Two newly de eloped a iable speed
ai - o-wa e and ai - o-ai HP models, which accoun o he deg ada ion o pe o mance due o
cycling condi ions, in e ac dynamically wi h esiden ial building models implemen ed in TRNSYS,
wi h s ochas ic h ee minu es ime-s ep demand p o iles. The building models a e each w apped
in o a Func ional-Mock-up Uni (FMU) and he co esponding elec ic loads a e e alua ed in he
CIGRE LV benchma k g id, wi h he Py hon based powe - low analysis ool PandaPowe .
The design and alida ion o a complex ne wo k, which in eg a es con ollable loads such as HPs,
Dis ibu ed Ene gy Sou ces (DERs), and ac i e powe elec onics, becomes challenging and
expe imen al da a is no easily ob ained o such la ge scales. The e o e, simula ion-based
app oaches a e possibly an app op ia e ool o suppo he de elopmen o inno a i e g ids,
howe e such sys ems equi e in eg a ed models o ep esen all key elemen , namely ene gy
ma ke s, g id equipmen , communica ion and con ol de ices, and each connec ed building. The
co-simula ion amewo k p esen ed in his hesis allows o he coo dina ion and synch oniza ion
o mul iple specialized ools o each elemen wi h high esolu ion domes ic load p o iles, which
cap u e sho e m powe spikes ha a e o he wise a enua ed. In addi ion o he gene al easibili y
o he p oposed amewo k o simula ion-based assessmen o complex g ids, he capaci y o an
u ban and a u al LV g id o accommoda e inc easing pene a ion a es o HPs is e alua ed. To
his end, a se o key pe o mance indica o s ha can be use ul o he Dis ibu ion Sys em Ope a o
(DSO) a e used: he minimum and maximum obse ed nodal ol ages, he ol age-d op ac oss he
eede , and he peak ans o me loading.
The esul s show ha he sensi i i y o he LV eede o a high le el o in eg a ion o domes ic
HPs is dependen on geog aphical loca ion, he HP echnology, and he ene gy densi y along he
eede . This wo k con i ms ha u al eede s a e mo e a isk o b eaching he minimum accep able
ol age le els, due o he long dis ances be ween deli e y poin s. In he Swedish case he
ans o me is o e loaded, while in Spain i emains wi hin accep able le els. Finally, HPs can
inc ease o dec ease he ans o me load, depending on which con en ional hea ing echnology
was p e iously commonly used in he egion unde in es iga ion.
P a g e | II
Acknowledgemen s
This hesis is he conclusion o my mas e deg ee in he EIT-InnoEne gy MSc. “SELECT” p og am
– En i onomical Pa hways o Sus ainable Ene gy Sys ems, which was unde aken a he “Royal
Ins i u e o Technology o S ockholm” (KTH) and “Uni e si a Poli ècnica de Ca alunya” (UPC). In
e ospec , I wan o hank his p og am and my ellow s uden s o o e ing me a unique as e o he
ene gy enginee ing ield om a ious coun ies pe spec i es, and o he oppo uni y o unde aking
challenging p ojec s ha ha e aken me all o e he globe.
Allo o people ha e help me along my pa h in he ields o Building and Ene gy Enginee ing. I would
like o hank my in e nship ad iso D . Jaume Salom o he “Ins i u de Rece ca en Ene gia de
Ca alunya” (IREC). He ga e me ample la i ude o ollow he s eams o my cu iosi y and explo e
a ious hemes, always suppo ed by his knowledge and expe ience. My g a i ude goes also o my
academic supe iso D . F ancisco Diaz-Gonzalez om he “Uni e si a Poli ècnica de Ca alunya”
(UPC) o his emendous pa ience acing my lack o p io knowledge in he ield o elec ical
enginee ing, and his use ul ad ice ha allowed his hesis o be cohe en om he elec ical g id
pe spec i e. I would like o exp ess my g ea app ecia ion o he in aluable guidance o Ma in Kegel
om “Canme Ene gy” in he o mula ion o he hea pump model p esen ed in his hesis, as well as
o sha ing wi h me along he pas 5 yea s an endless amoun o in o ma ion, da a, and ips abou
building simula ion models and no ms in ou Canadian clima e. Finally, al hough no in ol ed in his
hesis, I would like o hank D . Hua Ge om “Conco dia Uni e si y” in Mon éal, ha has o e ed
me immense suppo and oppo uni ies in my debu s in enginee ing du ing my bachelo deg ee o
Building Enginee ing, which has indi ec ly led me o his inal wo k.
On his pe sonal no e, I would like o s a / inish wi h a quo e aken om one o he many
disse a ions I ha e come ac oss du ing his hesis p ojec :
So I si be o e lowe s,
hoping hey will ain me in he a o opening up.
I s and on moun ain ops belie ing ha a alanches will each me o le go.
I know no hing,
bu I am he e o lea n.
– Shane Koyczan
The S uden
III | P a g e
Table o Con en
ABSTRACT _________________________________________________________ I
ACKNOWLEDGEMENTS___________________________________________ II
TABLE OF CONTENT _____________________________________________ III
LIST OF FIGURES __________________________________________________ V
LIST OF TABLES ________________________________________________ VIII
GLOSSARY ________________________________________________________ IX
1 INTRODUCTION ______________________________________________ 1
Mo i a ion ...................................................................................................................................... 1
Objec i es ....................................................................................................................................... 2
Scope & S uc u e o he Thesis ................................................................................................. 3
2 LITERATURE REVIEW _________________________________________ 4
Hea Pumps .................................................................................................................................... 4
Technical Cha ac e is ics ..................................................................................................... 4
Key Pe o mance Indica o s ............................................................................................... 7
Modelling .............................................................................................................................. 10
Elec ici y G id ............................................................................................................................ 11
Technical Cha ac e is ics ................................................................................................... 11
Key Pe o mance Indica o s ............................................................................................. 12
Modelling .............................................................................................................................. 13
In e ac ion Be ween Hea Pumps and he Elec ici y G id ................................................. 15
Economic Implica ions .............................................................................................................. 17
Na ional / In e na ional Le el .......................................................................................... 18
Co po a ion/O ganiza ion Le el ..................................................................................... 19
P i a e/Household Le el .................................................................................................. 22
3 METHODOLOGY & CASE STUDIES ____________________________ 23
Hea Pump Model ....................................................................................................................... 23
Va iable Speed Ai - o-Wa e Hea Pump ....................................................................... 23
Va iable Speed Ai - o-Ai Hea Pump ............................................................................ 24
Building Model ............................................................................................................................ 28
Building En elope .............................................................................................................. 28

P a g e | IV
HVAC ................................................................................................................................... 28
Occupancy and Elec ical Appliances ............................................................................. 29
Low Vol age G id Model ........................................................................................................... 30
PyPowe Valida ion ............................................................................................................ 35
Co-simula ion ............................................................................................................................... 38
4 RESULTS & DISCUSSION ______________________________________ 40
Inc easing Hea Pump Pene a ion Ra es ............................................................................... 40
Spain: 100% Hea Pumps .......................................................................................................... 46
Sweden: 100% Hea Pumps ...................................................................................................... 53
5 CONCLUSIONS _______________________________________________ 58
Main Poin s o In e es ............................................................................................................... 58
Limi a ions o his S udy ............................................................................................................ 62
6 BIBLIOGRAPHY ______________________________________________ 64
APPENDIX I: CIGRE EUROPEAN RESIDENTIAL LV GRID ___________ 70
CIGRE and PyPowe LV Residen ial G id ........................................................................................ 70
CIGRE Powe Flow Resul s ................................................................................................................. 72
APPENDIX II: COINCIDENCE FACTOR _____________________________ 74
Coincidence Fac o ................................................................................................................................. 74
Velande ’s Fo mula ................................................................................................................................. 75
Simul anei y Fac o .................................................................................................................................. 75
CIGRE Load Coincidence Fac o ........................................................................................................ 76
APPENDIX III: CO-SIMULATION TOOLS ___________________________ 77
Func ional Mock-up In e ace .............................................................................................................. 77
Mosaik ....................................................................................................................................................... 78
FUMOLA & P olemy II ........................................................................................................................ 79
APPENDIX IV: SIMULATION RESULTS _____________________________ 80
Modelling o he Va iable Speed Ai - o-Ai Hea Pump .................................................................. 80
Co-simula ion Ou pu G aphs .............................................................................................................. 81
Daily T ans o me Loading ................................................................................................................... 84
V | P a g e
Lis o Figu es
FIGURE 1:TECHNICAL AND ECONOMIC SUBSYSTEMS WITH WHICH THE ELECTRICAL SYSTEM OF A SINGLE RESIDENCE
INTERACTS [12] 3
FIGURE 2: BASIC HEAT PUMP SYSTEM CYCLE 4
FIGURE 3: TEMPERATURE AND COMPRESSOR SPEED PROFILES FOR A HP WITH AND WITHOUT AN INVERTER
CONTROL (COOLING MODE) [17] 5
FIGURE 4: UNITS SOLD BY TYPE [22] 6
FIGURE 5: UNITS SOLD BY COUNTRY [22] 6
FIGURE 6: AGGREGATION OF TYPICAL VALUES OF THE NUMBER OF CUSTOMERS PER LV FEEDER. THE MEDIAN IS
DENOTED “X” IN THE BOX. THE BOX IS DEFINED BY THE 25TH AND 75TH, THE WHISKERS ARE THE MAXIMUM
AND MINIMUM VALUES 12
FIGURE 7: PERFORMANCE CURVE FOR PART-LOAD CONDITIONS; INTERPOLATED CURVES BY POLYNOMIAL
REGRESSION (SOLID LINES) VS. MEASURED PERFORMANCE CURVES (DASHED LINED). 25
FIGURE 8: TYPICAL DEFROST CYCLE [32] 25
FIGURE 9: MODEL VALIDATION; HEAT OUTPUT MEASURED ON THE CONDENSER AIR SIDE (Q_AIR) AND
REFRIGERANT SIDE (Q_REF) VS. MODELLED HEAT OUTPUT (Q_INTERP) (TOP), SUPPLY AIR TEMPERATURE
MEASURED (TSUP_MEAS) VS. MODELLED (TSUP_INTERP) AT MEASURED OUTDOOR TEMPERATURE (TOUT)
AND INDOOR TEMPERATURE (TIN) (BOTTOM) 27
FIGURE 10: POWER PROFILE IN DEFROST-CYCLING OPERATION CONDITIONS 27
FIGURE 11: REPRESENTATIVE 12 STOCHASTIC PROFILES OF ONE DAY. THE FIRST NUMBER OF THE MODEL NUMBER
REPRESENTS THE NUMBER OF OCCUPANTS, THE SECOND IS THE APPLIANCE STOCK LEVEL AND TWO LAST
NUMBERS CORRESPOND TO A UNIQUE IDENTIFIER (I.E 2411 HAS 2 OCCUPANTS AND AN APPLIANCES LEVEL OF
4) 30
FIGURE 12: TOPOLOGY OF CIGRE EUROPEAN LV DISTRIBUTION NETWORK BENCHMARK [36] 31
FIGURE 13: PEAK APPLIANCE LOAD OF HOUSEHOLDS PER OCCUPANCY LEVEL AND LOCATION (E.G. SWD_4 AND
SPN_2 SIGNIFY THE SWEDISH MODELS WITH 4 OCCUPANTS AND THE SPANISH MODEL WITH 2 OCCUPANTS,
RESPECTIVELY) 33
FIGURE 14: VOLTAGE VALIDATION 35
FIGURE 15: CURRENT VALIDATION 35
FIGURE 16: THREE PHASE VOLTAGE AND CURRENT DIAGRAM 36
FIGURE 17: LV NETWORK CO-SIMULATION FRAMEWORK 39
FIGURE 18: IMPACT OF HPS ON THE VOLTAGE ACROSS THE FEEDER IN THE SPANISH RURAL BASE CASE SCENARIO 41
FIGURE 19: IMPACT OF HPS ON THE VOLTAGE ACROSS THE FEEDER IN THE SWEDISH RURAL BASE CASE SCENARIO
41
FIGURE 20: IMPACT OF HPS ON THE TRANSFORMER IN THE SPANISH RURAL BASE CASE SCENARIO 42
FIGURE 21: IMPACT OF HPS ON THE TRANSFORMER LOAD IN THE SWEDISH RURAL BASE CASE SCENARIO 43
FIGURE 22: DAILY TRANSFORMER LOADING IN THE SPANISH R_BC CASE STUDY 43
FIGURE 23: DAILY TRANSFORMER LOADING IN THE SWEDISH R_BC CASE STUDY 44
FIGURE 24: TWO HOUSEHOLDS IN THE SPANISH HP_100 SCENARIO - INDOOR TEMPERATURE (TIN4211, TIN3411
[°C]), APPLIANCE LOADS (APP4211, APP3411[W]), AND HP POWER DRAW (PHP4211, PHP3411[W]). 44
FIGURE 25: TRANSFORMER MAXIMUM, MINIMUM, AVERAGE AND MEDIAN LOADING % WITH INCREASING HP
LOADS. 46
P a g e | VI
FIGURE 26: AGGREGATED LOAD OF EACH BUILDING CLUSTER FOR THE SPANISH U_BC AND U_WC SCENARIOS IN
WINTER 47
FIGURE 27: AGGREGATED LOAD OF EACH BUILDING CLUSTER FOR THE SPANISH U_BC AND U_WC SCENARIOS IN
SUMMER 48
FIGURE 28: VOLTAGE AT THE FIRST NODE (V2) AND THE LAST NODE (V18) OF THE SPANISH FEEDER IN WINTER 49
FIGURE 29: VOLTAGE AT THE FIRST NODE (V2) AND THE LAST NODE (V18) OF THE SPANISH FEEDER IN SUMMER 49
FIGURE 30: CORRELATION BETWEEN VOLTAGE (V11) AND LOAD (S11) IN THE SPANISH R_BC AND R_WC GRID 50
FIGURE 31: CORRELATION BETWEEN VOLTAGE (V18) AND LOAD (S18) IN THE SPANISH R_BC AND R_WC GRID 50
FIGURE 32: SPANISH TRANSFORMER LOAD SENSITIVITY TO THE ENERGY DENSITY IN WINTER 51
FIGURE 33: DAILY TRANSFORMER LOADING IN SUMMER 52
FIGURE 34: SPANISH URBAN GRID - TRANSFORMER LOADING PEAKS AND VALLEYS OVER THE THREE-DAY PERIOD52
FIGURE 35: AGGREGATED LOAD OF EACH BUILDING CLUSTER FOR THE SWEDISH URBAN BASE-CASE (BC) AND
WORST-CASE (WC) SCENARIOS 53
FIGURE 36: VOLTAGE AT THE FIRST NODE (V2) AND THE LAST NODE (V18) OF THE SWEDISH FEEDER 54
FIGURE 37: CORRELATION BETWEEN VOLTAGE (V11) AND LOAD (S11) IN THE SWEDISH R_BC AND R_WC GRID 55
FIGURE 38: CORRELATION BETWEEN VOLTAGE (V18) AND LOAD (S18) IN THE SWEDISH R_BC AND R_WC GRID 55
FIGURE 39: SWEDISH TRANSFORMER LOAD SENSITIVITY TO THE ENERGY DENSITY 56
FIGURE 40: SWEDISH URBAN GRID - TRANSFORMER LOADING PEAKS AND VALLEYS 57
FIGURE 41: PEAK LOAD PER HOUSEHOLD FOR INCREASING NUMBER OF HOUSEHOLDS CONSIDERED 59
FIGURE 42: CLUSTER OF AGGREAGATED HOUSEHOLD LOADS - SMAX [KVA] 60
FIGURE 43: TRANSFORMER LOADING WHEN USING A SINGLE HP OR USING THREE HPS MODELS WITH DIFFERENT
CYCLING PERIODS 62
FIGURE 44: COORDINATING HPS FOR PEAK SHAVING (LEFT) AND LOAD SHIFTING (RIGHT) FOR REDUCED
TRANSFORMER STRESS [33] 63
FIGURE 45: TRANSFORMER - CIGRE DATA AND PYPOWER MODEL 70
FIGURE 46: LOADS - CIGRE DATA AND PYPOWER MODEL 70
FIGURE 47: GEOMETRY OF UNDERGROUND LINES - CIGRE DATA AND PYPOWER MODEL 71
FIGURE 48: LINES - CIGRE DATA AND PYPOWER MODEL 71
FIGURE 49: CIGRE POWER FLOW RESULTS I 72
FIGURE 50: CIGRE POWER FLOW RESULTS II 73
FIGURE 51: PEAK LOAD CONTRIBUTION PER CUSTOMER AS THE AGGREGATION LEVEL INCREASES FOR DIFFERENT
COINCIDENCE FACTORS [115] 74
FIGURE 52: MAPPING THE FUNCTIONS AND INTERACTION BETWEEN THE SIMAPI AND FMI[112] 78
FIGURE 53: VISUAL RENDITION OF CO-SIMULATION IN PTOLEMY II GRAPHICAL INTERFACE [110] 79
FIGURE 54: SCHEMATIC OF SOFTWARE LAYERS AND THEIR PURPOSE WITHIN THE SIMULATION ENVIRONMENT [110]
79
FIGURE 55: HP MEASURED DATA AND MODEL COMPARISON - JANUARY 27TH 80
FIGURE 56: HP DATA AND MODEL COMPARISON - FEBRUARY 24TH 80
FIGURE 57: HP DATA AND MODEL COMPARISON - FEBRUARY 27TH 80
FIGURE 58: : HP MEASURED DATA AND MODEL COMPARISON - APRIL 3-4 80
FIGURE 59: TWO HOUSEHOLDS IN THE SWEDISH HP_0 SCENARIO 81
FIGURE 60: TWO HOUSEHOLDS IN THE SWEDISH HP_30 SCENARIO 82
VII | P a g e
FIGURE 61: TWO HOUSEHOLDS IN THE SWEDISH HP_100 SCENARIO 83
FIGURE 62: TWO HOUSEHOLDS IN THE SPANISH R_BC COOLING SCENARIO 84
P a g e | 4
2 Li e a u e Re iew
The HP ma ke expe ienced signi ican g ow h a es and has ob ained a no ewo hy sha e o hea
supply sys ems [6]. I has been sugges ed ha hey can play a majo ole in educing he ene gy
consump ion and de-ca bonizing he hea ing sec o , by linking he elec ic and he he mal ene gy
domains [3], [13], [14]. Mo eo e , HP ope a ion could po en ially p o ide ancilla y se ices o he
g id such as ol age con ol, conges ion managemen and p o ision o spinning/non-spinning
ese es, o allow a s able and cos -e icien ope a ion o he elec ic g id [15]. Howe e , a high
pene a ion a e o esiden ial HPs migh inc ease he peak elec ic demand and he eby cause
s ess on many le els o he g id. The ollowing will p o ide an o e iew o HPs and LV g ids, as
well as he in e ac ion be ween hem in e ms o a single building and a building clus e .
Hea Pumps
The ollowing sec ion will p o ide basic in o ma ion abou HP ope a ing p inciples, elec ical
cha ac e is ics, and con ol s a egies, as well as me hods o modelling a HP.
Technical Cha ac e is ics
A basic apou comp ession HP cycle is composed o wo hea exchange s (one ac ing as an
e apo a o and ano he as a condense ), a comp esso , and an expansion al e (Figu e 2).
Figu e 2: Basic hea pump sys em cycle
A luid will abso b hea om a low empe a u e enewable hea sou ce (ambien ai , g ound, wa e ),
and ans e i o a hea sink (ai , wa e ) o be used o space hea ing and/o domes ic ho wa e . In
such HPs he hea sou ce is used o e apo a e a s eam o liquid e ige an a low p essu e. The
e ige an apou is hen comp essed, leading o a high p essu e and high empe a u e s a e. The

5 | P a g e
accumula ed hea is ans e ed o he hea sink, and he now liquid e ige an is expanded back o a
low-p essu e le el and he cycle is epea ed. In esiden ial applica ions, depending on he sou ce and
sink empe a u es, ela i ely li le addi ional ene gy is needed o he elec ical comp esso .
In esiden ial buildings, he ype o hea sink is linked o he ype o hea ing dis ibu ion and s o age
sys em. Wa e is used o adia o o loo hea ing sys ems and he p epa a ion o DHW, whe eas ai
is mos ly used in en ila ion and hea eco e y applica ions. The empe a u e o he luid ha needs
o be p o ided by he HP di e s depending on he equi emen s o he hea sink. Depending on he
building physics, common highes empe a u es a e needed o he p epa a ion o DHW (up o 65 °C)
ollowed by adia o hea ing (up o 55 °C), en ila ion (up o 40 °C) and loo hea ing (up o 35 °C)
[13].
The annual pe o mance o a HP sys em is in luenced by how he HP is designed o supply he
a iable hea demand. Fixed speed HP uni s a e ope a ed in on/o cycles, whe eas a iable speed
HPs wi h an in e e con ol use a con inuous egula ion o he comp esso speed o e a wide
ope a ional ange. I can be obse ed ha in e e powe con ol is capable o eaching he
empe a u e se -poin mo e apidly han an on/o HP, and has a mo e accu a e empe a u e con ol,
he e o e ob aining a mo e s able com o le el (Figu e 3). The appliances equipped wi h his
echnology a e mo e e icien and ha e lowe ene gy consump ion [16].
Figu e 3: Tempe a u e and comp esso speed p o iles o a HP wi h and wi hou an in e e con ol (cooling
mode) [17]
The e iciency o he HP can be signi ican ly in luenced by cycling due o pa ial loads o de os
ope a ion, and by s and-by powe consump ion, which can lead o a pe o mance educ ion o 5-30%
P a g e | 6
[3], [18]–[21]. De os cycling occu s as a p ecau ion in o de o sa egua d he hea pump om os
accumula ion on he e apo a o hea exchange . Depending on he ou doo ai humidi y, he os
can al eady occu as he empe a u e go below 7°C [20].
The e e sible ai - o-ai HP echnology ype domina es he Eu opean ma ke , ollowed by ai - o-
wa e hea ing, e e sible uni s and sani a y ho wa e uni s (Figu e 4). This is he case in Sweden and
Spain which a e he hi d and ou h la ges HP ma ke s in Eu ope, espec i ely (Figu e 5)[22].
Figu e 4: Uni s sold by ype [22]
Figu e 5: Uni s sold by coun y [22]
The e olu ion om ai - o-ai o ai - o-wa e sys ems, which p esen s mo e oppo uni ies in e ms o
7 | P a g e
he lexibili y o e ed by he connec ed he mal wa e ank ha e been in es iga ed [23] and migh be
inc easingly ins alled in newly cons uc ed buildings, o as a e o i op ion in buildings ha p e iously
we e designed o hyd onic boile sys ems.
Howe e , his end o subs i u ion o gas boile s o space hea ing and DHW needs o be
suppo ed by an adequa e con ol s a egy, as he s a ing cu en o hea pumps a e swi ch-o
o blackou s can be highe han in s eady ope a ion and p oblema ic o he elec ici y g id[3], [17],
[24].
Key Pe o mance Indica o s
The pe o mance o HPs in buildings is a ec ed by a la ge numbe o ac o s, such as[17]:
• Clima e - annual hea ing and cooling demand and maximum peak loads.
• Tempe a u es o he hea sou ce and empe a u e se poin s o he hea dis ibu ion sys em.
• Auxilia y ene gy consump ion - pumps, ans, supplemen a y hea o hyb id sys em e c.
• Technical s anda d o he HPs and he e ige an s.
• Sizing o he hea pump – designed o supply he ull hea demand o jus a po ion o i , and
he ope a ing cha ac e is ics o he HP.
• Con ol sys em.
The mos common indica o s o he pe o mance o a HP (in hea ing mode) a e he hea ou pu o
he condense (𝑄), he elec ic inpu (𝑃), and he Coe icien o Pe o mance (𝐶𝑂𝑃). The 𝐶𝑂𝑃 which
now anges be ween 3.2 and 5.2 depending on he echnology [13], is he a io:
𝐶𝑂𝑃=𝑄
𝑃(2.1)
To e alua e he ull load pe o mance o he HP based on he a ed pe o mance, usually p o ided
by he manu ac u e , [20], [25] sugges s co ela ions simila o he ollowing equa ion:
1
𝐶𝑂𝑃= 𝑃𝑟𝑎𝑡
𝑄𝑟𝑎𝑡∗(1+𝐵1∗∆𝑇 + 𝐵2∗∆𝑇2) (2.2)
The alue o ∆𝑇 is in luenced by he empe a u e a he inle o he e apo a o (𝑇𝑖𝑛,𝑒𝑣) and
condense (𝑇𝑖𝑛,𝑐𝑜𝑛):
∆𝑇= 𝑇𝑖𝑛,𝑒𝑣+273.15
𝑇𝑖𝑛,𝑐𝑜𝑛+273.15 − 𝑇𝑖𝑛,𝑒𝑣,𝑟𝑎𝑡+273.15
𝑇𝑖𝑛,𝑐𝑜𝑛,𝑟𝑎𝑡+273.15 (2.3)
P a g e | 8
The pa ame e s B1 and B2 a e de e mined by he leas squa e me hod using manu ac u e da a.
Two se s o pa ame e s a e calcula ed o below and abo e he os limi empe a u e, since
os ing on he e apo a o a ec s he pe o mance and he eby he slopes o he pe o mance
cu e.
As men ioned p e iously, he e iciency o he hea pump depends on he ope a ing condi ions
and comp esso echnology, ha dic a e he pe o mance in pa -load condi ions. The e iciency
o he hea pump a pa -load can be desc ibed by plo ing he Pa Load Ra io (PLR) e sus he
Pa Load Fac o (PLF). The PLR is calcula ed as he a io o he ac ual building hea ing/cooling
load o he ull-load a ed capaci y o he HP.
𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑦 𝑅𝑎𝑡𝑖𝑜=𝑃𝐿𝑅= 𝑄𝑐𝑦𝑐
𝑄𝑟𝑎𝑡 (2.4)
Whe e 𝑄𝑐𝑦𝑐 is he hea ing o cooling demand a a ce ain ime, and 𝑄𝑟𝑎𝑡 is he a ed capaci y a
ull-load. The PLF is co ela ed o he PLR as such:
𝑃𝐿𝐹= 𝐶𝑂𝑃𝑐𝑦𝑐
𝐶𝑂𝑃𝑟𝑎𝑡= 𝑄𝑐𝑦𝑐 𝑃𝑐𝑦𝑐
⁄
𝑄𝑟𝑎𝑡 𝑃𝑟𝑎𝑡
⁄= 𝑃𝐿𝑅
𝑃𝑐𝑦𝑐 𝑃𝑟𝑎𝑡
⁄(2.5)
In he case o a ixed speed HP [26] u he de elops his equa ion assuming a linea ela ion
be ween he pa -load Powe Ra io and PLR:
𝑃𝑜𝑤𝑒𝑟 𝑅𝑎𝑡𝑖𝑜=𝑃𝑐𝑦𝑐
𝑃𝑟𝑎𝑡=𝑎∗𝑃𝐿𝑅+𝑏 (2.6)
This app oach sugges s ha a single pa -load es is su icien o calcula e he linea co ela ion
be ween he pa -load powe a io and PLR and, consequen ly, he PLF. A simila linea co ela ion
is de ined by [20], which simpli ies and elimina es he need o expe imen ing, whe e 𝑎 is equal o
1 and 𝑏 is he s andby powe ac ion de ined as:
𝑏= 𝑃𝑠𝑏
𝑃𝑟𝑎𝑡 (2.7)
Whe e 𝑃𝑠𝑏 is he s andby powe inpu , and 𝑃𝑟𝑎𝑡 is he o al powe inpu o he HP, including
comp esso and ans consump ion.
In he case o a a iable speed comp esso , he pe o mance will o en be de ined by a minimum
PLR abo e which i modula es he COP, and below which i will ope a e as a ixed speed On/O
HP [20], [25]. A simple linea co ela ion be ween he poin a nominal equency and he poin a
9 | P a g e
minimum equency (minimum PLR) is sugges ed by [20][26], allowing o cha ac e ize he
co ela ion be ween PLF and PLR by u ilizing only h ee es poin s: one ull load es a he
maximum equency, one ull load es a he minimum equency, one on/o es a he minimum
equency. Addi ionally, [20] sugges s ha wo COP cu es a e o be de eloped a each side o he
os empe a u e limi o he HP, bo h o he ull-load and pa ial load pe o mance e alua ion.
In he case o pa -load he deg ada ion due o cycling can be accoun ed o h ough a deg ada ion
coe icien (𝐶𝑑). Howe e [18] compa ed se e al o he gene alized app oaches o de ine he
co ela ion be ween he PLR and he PLF wi h de aul deg ada ion coe icien s (UNI 10963, EN-
14825 and EN-14511) and ound hem o be insu icien o cha ac e izing he dynamic
pe o mance o a hea -pump. The e o e [27] p esen s h ee common me hods o cha ac e ize he
s a -up beha iou o HPs:
𝐸𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑖𝑎𝑙: 𝑄󰇗𝑐𝑦𝑐𝑙𝑒
𝑄󰇗𝑠𝑡 =1−exp[−𝑡
𝜏] (2.8)
Whe e 𝑄󰇗𝑐𝑦𝑐𝑙𝑒 and 𝑄󰇗𝑠𝑡 a e he dynamic hea ing he mal powe and he s eady s a e he mal powe ,
espec i ely, is ime and 𝜏 is a ime cons an ep esen ing he ime a e s a -up ha i akes he
HP o each 63.2% o s eady s a e capaci y.
𝑆𝑖𝑔𝑚𝑜𝑖𝑑: 𝑄󰇗𝑐𝑦𝑐𝑙𝑒
𝑄󰇗𝑠𝑡 =1
1+exp[𝑡50−𝑡
𝑠](2.9)
Whe e 𝑠 is a e m ela ed o he slope o he sigmoid and 𝑡50 ep esen s he ime a e s a -up ha
i akes he HP o each 50% o s eady s a e capaci y.
𝑃𝑜𝑙𝑦𝑛𝑜𝑚𝑖𝑎𝑙: 𝑄󰇗𝑐𝑦𝑐𝑙𝑒
𝑄󰇗𝑠𝑡 = 𝑏6𝑡6+ 𝑏5𝑡5+𝑏4𝑡4+ 𝑏3𝑡3+ 𝑏2𝑡2+ 𝑏1𝑡 (2.10)
Whe e 𝑏6 o 𝑏1 a e polynomial i ed coe icien s. These me hods a e hen ex ended o he
deg ada ion coe icien :
𝐶𝑑, 𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑖𝑎𝑙= 𝑡𝑠𝑡−∫ 1−exp[−𝑡
𝜏]𝑑𝑡
𝑡𝑠𝑡
𝑡0(2.11)
𝐶𝑑, 𝑠𝑖𝑔𝑚𝑜𝑖𝑑= 𝑡𝑠𝑡−∫ 1
1+exp[𝑡50−𝑡
𝑠]𝑑𝑡
𝑡𝑠𝑡
𝑡0(2.12)
𝐶𝑑, 𝑝𝑜𝑙𝑦𝑛𝑜𝑚𝑖𝑎𝑙= 𝑡𝑠𝑡−∫ (𝑏6𝑡6+ 𝑏5𝑡5+𝑏4𝑡4+ 𝑏3𝑡3+ 𝑏2𝑡2+ 𝑏1𝑡)𝑑𝑡
𝑡𝑠𝑡
𝑡0(2.13)

P a g e | 10
Modelling
Hea pump models a e o en ca ego ized depending on he physical app oach [28]:
• The modynamic/physical app oach - based on he geome y o he hea pump and gene al
physics laws (hea and mass ans e )
• Black box app oach - in which empi ical co ela ions de e mine hea pump pe o mances as
a unc ion o he condi ions o use (ou doo and indoo empe a u es, powe demands, e c.)
• G ey box app oach - which is a he in e sec ion o he wo abo e-men ioned app oaches.
Ano he ype o classi ica ion is acco ding o he pe o mance indica o s:
• Ra ed pe o mances - calcula e he pe o mances o he sys em o poin s based on
empe a u es and o ull-load and s eady-s a e ope a ion. Some models a e based on he
ypically a ailable manu ac u e published da a wi h an equa ion ob ained using he leas
squa e i me hod [16], [20].
• Quasi-s a ic - his app oach conside s ime as a sequence o s eady s a es. Du ing a sequence,
he dynamics o he sys em esponse mus be as e han he dynamics o he dis u bances.
Fo ins ance, hou ly ime s ep simula ion conside s ha he ope a ing condi ions (like he
ou side empe a u e) does no change a lo compa ed o he hea pump pe o mances du ing
he hou ly cycle. The hea demands a e a e aged on one hou and he pe o mances a e
adjus ed in unc ion o sink empe a u e and PLF. The beha iou o he hea pump is
assimila ed o s eady-s a e snapsho s. This app oach is used o model ixed and dual speed
comp esso models in DOE-2, using deg ada ion coe icien s o accoun o educ ions in
e iciency due o cycling [29]. Ene gyPlus app oxima es a VSHP using a mul ispeed model
ha allows up o 4 disc e e comp esso speeds. [30]
• Dynamic - hese models calcula e hea a e and elec ic powe a any ins an o he simula ion,
so ha he ansien phases and he s eady s a es a e bo h simula ed. Some o hese models
a e physical, based on di e en ial physic laws and [31] desc ibe ex ensi ely a whi e box
physical sys em o equa ions o model he condense , e apo a o and comp esso . O he s a e
empi ical, based on equa ions wi h one, o se e al, ime cons an s o ep esen he ansien
phases. Models o duc less mini-spli and mul i-spli sys ems a e e y a e, and mos models
ound in li e a u e a e based on expe imen al empi ical da a [29], [32].
11 | P a g e
Elec ici y G id
The elec ical g id could be di ided in o he subsys ems o elec ic powe gene a ion, ansmission,
dis ibu ion and u iliza ion. Each subsys em ope a es a di e en ol age le els wi h ans o me s
linking be ween hem and adjus ing he ol age acco ding o se equi emen . The ene gy is
gene a ed in la ge powe plan s connec ed o a high ol age (HV) ansmission g id which is
ollowed by he dis ibu ion g id. The elec ical dis ibu ion g ids can be di ided in o Medium
Vol age (MV) and Low Vol age (LV) le els. Typical ol age le els in Eu ope a e 10kV o 20kV o
he MV g id and 0.4kV o he LV g id [33]. Essen ially, buildings a e he las elemen o he
elec ici y dis ibu ion sys em and he e o e hei impac is mos signi ican and isible on he le el
o LV ne wo ks and MV g ids[34]. The e o e, he ollowing sec ion will ocus on he LV g id
sys em and on he LV eede le el, as i is he subjec o conce n when discussing high pene a ion
a es o HPs in he esiden ial sec o .
Technical Cha ac e is ics
The majo i y o LV g ids ha e a adial opology, wi h a numbe o eede s s a ing om an LV
busba s connec ed o a MV/LV subs a ion (also called “seconda y subs a ion”) [33], [35].
T adi ionally, he LV dis ibu ion ne wo k was planned and buil o a unidi ec ional powe - low
om he seconda y subs a ion o he cus ome s. As such, he LV g id was conside ed passi e and
li le ol age con ol o measu emen s we e pe o med in his subsys em. The g id would be
dimensioned o deal wi h a peak load, while main aining a su icien ol age le el a he cus ome
deli e y poin s.
U ban and u al LV g ids ha e di e en cha ac e is ics and he e o e expe ience di e en isks [8]:
• Ru al g ids a e cha ac e ized by small (ye o en o e sized) ans o me s, and o e head lines
wi h limi ed c oss sec ions which co e la ge a eas. They usually ha e long eede s, which
expe ience a highe ol age d op ac oss he lines compa ed o u ban eede s. In a esiden ial
u al g id he e is a isk o lack o load di e si y, which in u n can inc ease he isk o
unde ol age o simul aneous domes ic loads, o o e ol age a deli e y poin s o DERs.
• U ban ne wo ks a e cha ac e ized by la ge ans o me s, and unde g ound cables wi h la ge
c oss sec ion and sho dis ances. The impac ha loads and DERs ha e on he ol age p o ile
h oughou he eede is less se e e han in u al dis ibu ion ne wo ks, howe e he e is a isk
o unde ol age i he agg ega ed loads a e oo high, as well as a isk o o e loading he
ans o me .
The e is a g ea di e si y o dis ibu ion g ids and DSOs and he e is a g ea need o a consolida ed
ep esen a i e es sys em ha would acili a e he analysis and alida ion o inno a i e g id
de elopmen me hods and echniques [36]–[39]. In li e a u e can be ound some en a i e
P a g e | 12
benchma ks, which a e especially di icul o ob ain o he LV g id as he le el o con ol and
moni o ing is lowe han in he highe ol age le els [39]. In Figu e 6 is p esen ed an a emp o
consolida e he in o ma ion om many LV g id s udies [6], [8], [46], [35], [38], [40]–[45] and iden i y
he ypical a ibu es o he LV g id a he eede le el. I can be no ed ha he a e age numbe o
esidences pe LV eede is app oxima ely 24. I should be men ioned ha hese many s udies do no
always speci y:
• G id ype (u ban o u al)
• Exac numbe o cus ome s pe eede s
• Leng h o he eede o a e age dis ance be ween deli e y poin s
• Building ypology
The e o e, he conca ena ion o he da a om hese a ious s udies can only p o ide a e y ough
es ima e.
Figu e 6: Agg ega ion o ypical alues o he numbe o cus ome s pe LV eede . The median is deno ed “x” in
he box. The box is de ined by he 25 h and 75 h, he whiske s a e he maximum and minimum alues
Key Pe o mance Indica o s
In e ms o he LV g id s uc u e and eliabili y, DSOs indica o s can be [47]:
• LV consume s pe a ea.
• LV ci cui leng h pe LV consume .
13 | P a g e
• LV ci cui leng h pe a ea o dis ibu ion.
• LV unde g ound cable a io.
• Numbe o LV consume s ( esiden ial and comme cial) pe MV/LV subs a ion.
• A ea pe MV/LV subs a ion.
• Capaci y o MV/LV subs a ions pe consume .
• A ea co e ed pe capaci y o MV/LV subs a ion.
As o design indica o s, DSOs migh de ine he g id in e ms o :
• Typical MV/LV subs a ion ans o me capaci y in an u ban/ u al a ea
• A e age numbe o MV/LV subs a ions pe eede in u ban/ u al a eas
• Vol age le els
• Au oma ion equipmen and deg ee o au oma ion
• T ans o me low bandwid h (di e ence be ween maximum and minimum loading)
Signi ican indica o s o he g id s abili y a e:
• Vol age di e ence ac oss he eede (be ween he i s and las connec ed consume ).
• Vol age de ia ion om he nominal ol age a he eede deli e y poin s.
• The mal o e loading o he lines.
• Load o e load o he ans o me .
• Vol age le el sensi i i y o powe d aw o o e 1kW a a node.
The main wo cons ain s ha usually limi he dis ibu ion eede s a e he maximum/minimum
admissible ol age and cu en s [33], [39].
Modelling
The modelling o a LV ene gy sys em can be cha ac e ized by h ee di e en app oaches:
1. Building model ocused - in his app oach, he ene gy demand is gene ally gene a ed om a
de ailed building simula ion and desc ibed in hou ly load p o iles o e a yea . Following his,
he ene gy sys em is simula ed sepa a ely (pos -p ocessing) using he building simula ion
ou pu as a g id inpu , hinde ing a possible eedback om he sys em o he buildings.
P a g e | 20
• Rein o cemen o o e head lines (OH) due o iola ion o he mal/cu en limi (o e load)
iola ion o ol age limi . The cos o hese cables can ange be ween 32.8 – 65.0 €k/km.
• Rein o cemen o g ound moun ed ans o me (GMT) due o iola ion o he mal limi
(o e load). The cos o his equipmen can ange be ween 14.0 – 33.8 €k.
• Rein o cemen o pole moun ed ans o me due (PMT) o iola ion o he mal limi
(o e load). The cos o his equipmen is a ound 3.4 €k.
An inc ease s ess can ei he cause he equipmen o ail comple ely o o sho en hei li e ime and
has en hei eplacemen .[75]. In a scena io om he s udy abou he elec ic ehicles i was ound
ha he in eg a ion o 24 EVs called i s o a ol age ein o cemen o unde g ound cables a he 3 d
yea a e ins alla ion. As he numbe o EVs was inc eased, i also became necessa y o ein o ce he
ans o me . Towa ds he end o he 25-yea pe iod unde e alua ion, cu en ein o cemen o
unde g ound cables was also equi ed as hei he mal limi s we e exceeded. Al hough HPs p esen a
di e en load magni ude han EVs, hei in eg a ion would p obably equi e a simila imeline in e ms
o which equipmen would need o be ein o ced i s .
An example o e alua ion he cos o g id ein o cemen as a consequence o HP ins alla ion in 10
million households in he UK is p o ided in [24] whe e is assumed ha HPs could inc ease he LV
load by oughly 50%. The o al cos o ein o cing he LV g id was di ided in he ollowing
componen s:
• Cable eplacemen : he ein o ced g id would equi e a ein o cemen o ideally 30% (bu
mo e p obably 50%) o he eede cable leading om he ans o me o he consume s. I
was no ed he new cable po ion o he eede would be connec ed nea he ans o me
po ion, as o no dis u b o ha e o in e ac wi h each indi idual household.
• LV subs a ion ein o cemen : i was es ima ed ha app oxima ely 40% o he ans o me s
a e ope a ing well below hei maximum capaci y. Hence i was es ima ed ha only
app oxima ely 50% o he ans o me s, mos ly in u ban spaces, would need o be upg aded.
• T a ic managemen and s ee wo k cos : i was es ima ed ha ha he cos ela ed o he
ield wo k (such as digging enches o unde g ound cables) would be app oxima ely an
addi ional 10% o he ins alla ion cos .
• Asse eplacemen : i is epo ed ha mos LV cables ha e been unde only li le he mal
s ess. The e o e, e en i he calcula ed g id ein o cemen due o high HP pene a ion cos

21 | P a g e
would be lowe i he cables we e scheduled o eplaced anyway, i canno be assumed ha
his is he case.
Including ein o cemen s o he highe ol age le els, s a and o e head, he g id ein o cemen was
es ima ed o cos app oxima ely ₤2000 (€2247) pe con e ed house.
These cos s migh be mi iga ed by he oppo uni y o lexibili y o e ed o DSOs by HPs o manage
conges ion in a line o a ans o me . The quan i ica ion o he payback esul ing om he coo dina ed
modula ion o he HP loads allows he ac i a ion o he lexibili y wi hou c ea ing conges ions u he
in ime. A DSO could also ely on his se ice o balancing pu poses [25]. The lexibili y oppo uni y
o e ed by high HP pene a ion a es was es ima ed o he U.S. by he Rocky Moun ain Ins i u e [76]
which ex apola ed o he en i e U.S indings om an analysis o he ma ke (in collabo a ion wi h
local u ili ies) in ou s a es and ound ha he po en ial g id-le el cos sa ings om he lexibili y
o e ed by a high deploymen and con ol o elec ical HVAC and DHW is signi ican and ha abou
8% o U.S. peak demand could be educed while main aining com o and se ice quali y. I was
s a ed ha in he esiden ial sec o alone a widesp ead implemen a ion o demand lexibili y could
sa e 10–15% o po en ial g id cos s o DSOs in he U.S. O hese sa ings, mos come om a oided
in es men in gene a ion, ansmission, and dis ibu ion equipmen (app oxima ely 69%). Con olling
he iming o he ene gy demands o op imize hou ly ene gy p ices can p o ide addi ional sa ings o
abou 23%, while p o iding ancilla y se ices o he g id could ep esen a e enue which can moun
up o 8% o he o al ne sa ings o e ed by demand lexibili y.
2. Elec ici y e aile
The elec ici y e aile could use he lexibili y o e ed by HPs ei he o balance i s po olio as a
balance- esponsible pa y o o adjus i s consump ion acco ding o day-ahead spo ma ke p ices.
In he elec ici y ma ke , de ia ions om he ag eemen s wi h he DSO, incu economic penal ies
based on an imbalance a i [25] which could be mi iga ed wi h HP con ol s a egies.
3. Elec ici y Equipmen Manu ac u e s and Vendo s
The new se ice o HP lexibili y would suppo he ma ke o powe elec onics and “Sma G id
Ready” echnologies. The capaci y o ex ac inancial bene i om he lexibili y se ice; ei he
om he DSO which would sa e money on g id ein o cemen , o he indi idual consume which
would make money by selling lexibili y, could p obably encou age a p o i able expansion o such
ools.
P a g e | 22
4. The “Agg ega o ”
The concep e ol es a ound he idea o agg ega ing mul iple willing building owne s/occupan s
o con ac hei lexibili y po en ial wi h u ili ies. This ole is o special in e es in he con ex o
HPs as he indi idual esidence canno o e much lexibili y, and only becomes ele an when
agg ega ed on a highe le el. The agg ega o would con ol and coo dina e esiden ial hea pumps
o o e a di ec con ol lexibili y se ices o a u ili y/DSO company. The se ice could consis s
o a powe modula ion, upwa d o downwa d, ha is ac i a ed o a chosen ime o e a ixed
numbe o pe iods. The se ice modula ion can be ela i e o an op imized baseline ha minimizes
he ene gy cos s and a e i is ac i a ed i should be ollowed by s a egy o cons ain he ebound
e ec . Resul s om p e ious in es iga ions o his concep show ha wi h a se o one hund ed
hea pumps, a load agg ega o could o e modula ion ampli udes o up o 138 kW upwa d and up
o 51 kW o downwa d modula ion. These alues s ongly depend on he p oposed lexibili y
se ice and con ol s a egy, as hey could dec ease down o 2.6 kW and 0.4 kW, espec i ely, i
no ebound e ec was allowed [25].
P i a e/Household Le el
Many s udies ha e ound ha domes ic HPs can ep esen ene gy and consequen ly cos sa ings
o he p i a e consume . Fo example a s udy by [77] in cold clima es demons a ed ha e en in
sub-ze o empe a u es, some o he HPs could ope a e wi h a COP g ea e han 1, he eby
equi ing less ene gy han mos hea ing echnology. Mo eo e , he s udy s a ed ha he billing
analysis showed ha e en in colde clima es such as he No hwes egion o he USA he HPs
could p o ide sa ings o mo e han 5,000 kWh/y .
In e ms o lexibili y po en ial, ei he he building owne o he building occupan who pays o
he elec ici y, could bene i by di ec ly selling lexibili y, esponding o a signal sen by he DSO
o he agg ega o .
23 | P a g e
3 Me hodology & Case S udies
Two case s udies a e p esen ed in his hesis in o de o e alua e he impac o high pene a ion o
VSHPs in di e en loca ions wi h inhe en ly di e en clima es. The i s case ep esen s a Sou h
Eu opean clima e and is based in he a ea o Te assa (Ca alunya, Spain), and he second case
ep esen s a No h Eu opean cold clima e and is based in S ockholm (Sweden). The case s udies
a e simula ed o i e days wi h a ime s ep o h ee minu es. The i s wo days se e he pu pose
o ini ializing he s a e o a iables and a e disca ded om he analysis which ocuses on he h ee
subsequen days. Acco ding o he Join Resea ch Cen e (JRC) o he Eu opean Commission, he
main load in an a e age Eu opean esidence is hea ing. This is de ini ely ue in he Swedish case,
while in he Medi e anean clima e o Spain he hea ing and cooling loads a e ela i ely balanced,
howe e hea ing is s ill dominan [78], [79]. The e o e, his s udy shall ocus on hea ing, ye a
cooling scena io will be o e ed o he Spanish case. The chosen simula ion days ep esen a
coldes /ho es sequence, and he e o e a wo s -case scena io in e ms o hea ing/cooling loads.
Hea Pump Model
As p e iously men ioned in sec ion 2.1.1 he e a e nume ous HP echnologies, and o he scope o
his hesis he VSHP was selec ed as hey a e a ma u e echnology and highe pe o ming han ixed
speed comp esso s, and a e mo e likely o domina e u u e HP ins alla ions. Since dynamic models
o VSHP a e sca ce and o en neglec o cap u e pa -load e iciency deg ada ion and cycling s a -up
losses, his hesis uses wo newly de eloped models; a Va iable Speed Ai - o-Wa e Hea Pump (VS
AWHP) and a Va iable Speed Ai - o-Ai Hea Pump (VS AAHP). In all cases he VSHP is hea
d i en, meaning ha i s p io i y is main aining he com o o he occupan s, ega dless o ime-o -
use elec ici y a i s o o he mo e complex con ol s a egies.
I should be no ed ha he di e ence in sink luid can ha e a no able in luence on he pe o mance
cha ac e is ics o he VSHP. As he VS AWHP o en u ilizes an in e media y wa e ank i has a highe
ine ia capaci y han he VS AAHP wi h hea supplied di ec ly o he oom ai . Al hough bo h
echnologies su e om a lowe hea ou pu a e s a -up, and consequen ly a lowe COP, he VS
AAHP can expe ience a high powe d aw a s a -up [3], [32] which can be less signi ican in he case
o he VS AWHP [21], [80]. The e o e, al hough he wo black-box VSHP models used in his hesis
we e de eloped di e en ly, he signi ican cha ac e is ics o each echnology seem o be add essed
adequa ely, as will be de ailed in he ollowing sec ions.
Va iable Speed Ai - o-Wa e Hea Pump
The choice o an ai - o-wa e sys em in he Spanish case s udy is explained in [81] as he only ealis ic
choice o a hyd onic sys em ha is o supply DHW as well. Wi h an ai - o-ai HP he e is no op ion
P a g e | 24
o supply DHW. A g ound-sou ce HP was also disca ded, as he ac ual building is loca ed in an u ban
en i onmen and comp ises o 16 indi idual esidences wi h independen hea ing and cooling sys em.
The eason is ha he possibili y o d illing 1 o 16 sepa a e e ical bo eholes o se ice his building
was assumed o be un ealis ic, while combining all he esidences in o single load wi h a cen al hea ing
sys em would no ep esen accu a ely he building s ock in Spain. I was also conside ed a good
op ion as i could possibly eplace hyd onic boile wi h adia o sys ems which a e he dominan
echnology ound in Spain [82]–[84] and can se e bo h space hea ing and DHW.
The VS AWHP was de eloped by [85] and is based on a ailable manu ac u e da a (ou doo uni :
HITACHI Yu aki-S-Combi RAS-4WHVNPE, indoo uni : RWD-4.0NWE). In o de o accoun o
pa -load pe o mance and cycling deg ada ion he PLF equa ion de eloped by [27] is used and he
PLF-PLR cu e is ob ained wi h he a ailable pa -load da a using a quad a ic i me hod. This cu e
is used when condi ions a e abo e he minimum PLR limi o he hea pump, below which he VS
AWHP eso s o an on/o ope a ion.
Va iable Speed Ai - o-Ai Hea Pump
The VS ASHP is used in he Swedish case whe e i is he p e alen ins alled echnology [72], [86]
and whe e he domina ing con en ional hea ing sys em is elec ical. The VS AAHP model was
de eloped based on expe imen al esul s [32]. An app oach simila o [87] was used, whe e each
obse a ion (a one-minu e da a in e al) was agg ega ed in o bins acco ding o ou doo
empe a u e (2.5 °C each) and equency (5 hz each). Keeping only he da a wi h a deg ee o
con idence o 95%, he mean alue o powe inpu and hea ou pu wi hin each bin se ed as he
basis o i ed pe o mance cu es by means o 2nd o de polynomial eg ession. In Figu e 7 a e
compa ed he esul ing pe o mance cu es o selec ed equencies (i.e. H30 is he pe o mance
a 30 Hz) wi h he measu ed da a a ha equency (e.g. H30_m).
These cu es a e used o in e pola ion o he s eady-s a e and pa -load pe o mance o he black
box VS AAHP model w i en in py hon, which was alida ed by compa ing he esul ing hea ou pu ,
equency, supply ai empe a u e, and powe inpu wi h he measu ed da a. A pe o mance ile is
c ea ed wi h he pa ame e s:
• F equency [Hz]: 0, 30, 50, 52.5, 70, 90, 110
• Tin [°C]: 18, 21, 24
• Tou [°C]: -25, -20, -15, -10, -5, 0, 5, 10, 15
25 | P a g e
Figu e 7: Pe o mance cu e o pa -load condi ions; in e pola ed cu es by polynomial eg ession (solid lines)
s. measu ed pe o mance cu es (dashed lined).
This ile is used o in e pola e o he pa -load Capaci y Ra io and Powe Ra io o he condi ions a
each ime-s ep. Then an addi ional algo i hm is used in o de o mimic he dynamic beha iou o he
VS AAHP in cycling condi ions. Fi s he ypical beha iou wi hin a o al de os cycle is de ined in
e ms o he du a ion o he cycle ( _cycle), he du a ion o de os ( _du a ion), and he ime o
eco e y a e s a -up ( _ eco e y) (Figu e 8).
Figu e 8: Typical de os cycle [32]
0
0.25
0.5
0.75
1
1.25
1.5
1.75
2
2.25
2.5
-30 -25 -20 -15 -10 -5 0 5 10 15 20
Powe inpu [kW]
Ou doo empe a u e [C]
HZ30 HZ50 HZ60 HZ70
HZ90 HZ110 HZ30_m HZ50_m

P a g e | 26
Fo a dynamic s a -up model as is sough in his case, he e iciency deg ada ion unde cyclic
condi ions can be exp essed in se e al ways (sec ion 2.1.2) and in his model a penal y ac o wi h an
exponen ial exp ession was ound o be su icien o mimicking he educed e iciency and lowe
hea ou pu a he condense a e s a -up [32].
𝑃𝑒𝑛𝑎𝑙𝑡𝑦 𝐹𝑎𝑐𝑡𝑜𝑟=1 − exp ((−(𝑡+𝜏)
𝜏(3.1)
𝑄󰇗𝑐𝑦𝑐=𝑅𝑎𝑡𝑒𝑑 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑦 𝑥 𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑦 𝑅𝑎𝑡𝑖𝑜 𝑥 𝑃𝑒𝑛𝑎𝑙𝑡𝑦 𝐹𝑎𝑐𝑡𝑜𝑟 (3.2)
The penal y ac o ge s close o 1 as “ ” (cu en ime s ep) inc eases a e de os . The penal y
ac o egula es he hea ou pu o all ime s eps be ween 0 and “ _ eco e y” minu es a e de os .
In Figu e 9 can be seen an ex ac o he collec ed da a (2 days) ha was used o he alida ion o
he model, whe e he VS AAHP ope a ed in de os cycling mode due o he low ou doo
empe a u e (Tou ) which emained below 0 °C, and ied o main ain he indoo empe a u e
(Tin) a he se poin o 21 °C. The esul ing hea ou pu o he model (q_in e p) ollows he end
o he hea ou pu on he ai side o he condense (q_ai ) which was ound o ep esen mo e
accu a ely he hea ou pu a e cycling [32]. The co esponding ai supply empe a u e
(Tsup_in e p) co esponds well he measu ed da a (Tsup_meas). I should be men ioned ha he
speci ic VS AAHP model es ed con ains a sel -lea ning algo i hm o i s own which is no publicly
a ailable and makes i di icul o mimic exac ly he ime gap be ween each de os cycle. The e o e,
keeping in mind he scope and objec i e o his hesis, he ime gap be ween each de os cycle
has been simpli ied and is aken as he a e age de os cycling ime. An analysis o he da a also
clea ly deno ed ha he VS AAHP did no ope a e in equencies below 25 Hz and as such, he
modelled HP will shu -o i he load is oo low and will s a -up again once he hea ing load is
su icien ly high.
The powe d aw by he comp esso du ing he eco e y pe iod is highe han wha i would be o
he same empe a u e condi ions in s eady s a e ope a ion. In Figu e 10 can be seen he measu ed
powe inpu du ing one es cycle (p_meaus ed) and he modelled powe inpu (p_in e p). The exac
pe iod be ween each cycle, as well as he exac eco e y pe iod a e s a -up could no be exac ly
ep oduced, as he es ed VS AAHP has an in e nal sel -lea ning algo i hm and o he con ol s a egies
ha a e no publicly a ailable, and ha e no all been iden i ied. The e o e, he a e age cycle pe iod
and eco e y pe iod ha e been assumed o his model, and he gene al end and p o ile
cha ac e is ics ha e been mimicked o a le el ha sa is ies he pu pose o his hesis, which aims
mos ly o cap u e he peak powe d aw a e s a -up.
27 | P a g e
Figu e 9: Model alida ion; hea ou pu measu ed on he condense ai side (q_ai ) and e ige an side (q_ e )
s. modelled hea ou pu (q_in e p) ( op), supply ai empe a u e measu ed (Tsup_meas) s. modelled
(Tsup_in e p) a measu ed ou doo empe a u e (Tou ) and indoo empe a u e (Tin) (bo om)
Figu e 10: Powe p o ile in de os -cycling ope a ion condi ions
Since i would be inaccu a e o model all esidences wi h he same VS AAHP, h ee sligh ly di e en
models ha e been c ea ed, al e ing he ime he hea pump is in de os ( _du a ion) and he o al
de os cycle ( _cycle) ha a e depic ed in Figu e 8. The di e ence in he h ee VS AAHP is gi en in
Table 1, whe e “HP_6” ep esen s he a e age alues ound du ing he expe imen s. The models o
HP_3 and HP_9 emain wi hin he bounda ies o he expe imen esul s, as he manu ac u e limi s
“ _du a ion” o a maximum o 10 minu es.
Table 1: VS AAHP models
HP_3
HP_6
HP_9
_du a ion (minu es)
3
6
9
_cycle (minu es)
90
150
210
The ime i akes o eco e ( _ eco e y) is cons an o all h ee models and co esponds o he
a e age alue o 30 minu es.
P a g e | 28
Building Model
The building simula ion so wa e TRNSYS is used o model he building sys em. The ollowing
will desc ibe he cha ac e is ics o he wo esiden ial models used in his hesis.
Building En elope
Two alida ed esiden ial models a e used, which ep esen s he common building s ock ha is
abo e 20 yea s old and would p obably need o be e- u bished i hey we e o mee he new and
mo e s ingen ene gy s anda ds in he EU. The esidence in Te assa was buil o mee he
s anda ds om he 1980s (NRE – AT- 87[88]). The ypical Swedish esidence ep esen s he
building s ock o he 1980s and is an adap a ion o a p e iously alida ed model o a single- amily
house in Canada. Adop ing he Canadian model o Sweden was possible due o he g ea
esemblance in loo a ea and insula ion le els o he ypical 1980s single house buil in Mon eal
acco ding o he Canadian Single-De ached and Double/Row Housing Da abase (CSDDRD)[89],
and he single- amily esidence building s ock in he 1980s in Sweden desc ibed in EU Building
Da abase[90]. The main building p ope ies can be ound in Table 2, and mo e de ails o he
Te assa and Mon eal esidences can be ound in [91] and [92], [93], espec i ely.
Table 2: Key Building En elope P ope ies
Pa ame e
Uni
Cold Clima e
Medi e anean
Clima e
Loca ion
-
S ockholm (Sweden)
Te assa (Spain)
Building da e
-
1980-1990
1991 – 2007
Floo a ea
m2
210
108.5
Roo U- alue
W/m2K
0.208
0.546
Wall U- alue
W/m2K
0.358
0.6
Windows U- alue
W/m2K
2.53
2.5 o 5.7
In il a ion n50
1/h
4.4
3.0
HVAC
In o de o assess he impac o an inc easing pene a ion a e o HPs, wo hea ing sys ems a e
modelled o each clima e. A base-case model ep esen s he con en ional hea ing sys em used in
Te assa and in S ockholm in he 1980s-1990s, and a second model has an in eg a ed VSHP. While
he base case in Te assa has a con en ional hyd onic boile sys em, in Sweden an di ec elec ical
hea ing sys em is modelled. This is because his echnology domina ed he space hea ing ma ke
o single- amily esidence and in 2001 app oxima ely 70% o he Swedish single amily esidences
had elec ic space hea ing sys ems, o which, a ound 34% we e hea ed by di ec -ac ing elec ici y
29 | P a g e
and wa e -based elec ic hea ing [71]. Be ween 2010-2018 he pe cen o newly ins alled hea
pumps eplacing o complemen ing di ec elec ic hea ing is be ween 21% o 25% [94].
Table 3: HVAC sys em o he case s udies
Cold Clima e
Medi e anean Clima e
Base-case
Elec ic esis ance hea e
Na u al gas boile
Hea Pump
Va iable speed
ai - o-ai hea pump
+
Back-up elec ic esis ance hea e s
Va iable speed
ai - o-wa e hea pump
+
Fan coils
Se poin s
Hea ing: 20.5-21°C
Cooling: -
Hea ing: 20°C + nigh se back o 18°C
Cooling: 25°C
Occupancy and Elec ical Appliances
Residen ial ene gy consump ion p o iles a e di icul o p edic o se e al easons: occupan
beha iou can a y widely, p i acy issues limi he access o indi idual household ene gy da a, and
usually he de ailed me e ing o end-uses consump ion a e cos ly. The impo ance o he demand
side modelling is pa icula ly high when e alua ing he peak consump ion o an agg ega ed building
clus e , upon which elies he design he LV g id. Common me hods o size dis ibu ion g ids a e
he Velande o mula and me hods based on simul anei y and coincidence ac o s. These me hods
a e conside ed eliable when he indi idual loads a e ela i ely homogeneous, (e.g. all esiden ial),
and he numbe o esidences is la ge enough (abou 200 connec ions). Conce ns exis abou he
eliabili y o his me hod o eede s se ing smalle clus e , such as 20-60 esidences[95].
In e ms o modelling load p o iles, he high impo ance o load di e si y is highligh ed in [7]
whe e he load p o iles used had a high coincidence ac o and consequen ly he base case u al
Aus ian g id modelled su e ed om a ol age de ia ion iola ion (below 0.9 p.u.), e en be o e
adding HPs.
As his hesis ocuses on he LV eede , i was chosen o use 25 s ochas ic p o iles de eloped by
[81], [96] which ep esen homes o 1, 2, 3 o 4 esiden s, and ha e 3 di e en appliance s ock
le els. The p opo ions o esidences wi h 1, 2, 3 o 4 occupan s on a LV eede is weigh ed
acco ding o he s a is ical da a p o ided in Table 4.
Table 4: Pe cen sha e o esidences in building s ock occupied by 1,2,3, and 4 people [90]
Coun y/Occupan s
1
2
3
4
Spain (%)
24.6
30.6
21.1
17.8
Sweden (%)
39.9
32.8
10.2
12.1
A sample o 12 s ochas ic p o iles is p esen ed in Figu e 11, which highligh s he impo ance o unique
P a g e | 36
balance o he sys em.
Figu e 16: Th ee phase ol age and cu en diag am
In a h ee-phase balanced sys em he ins an aneous alues o ol age and cu en a e [103]:
𝑢𝐴(𝑡)= √2𝑉sin(𝜔𝑡)
𝑢𝐵(𝑡)= √2𝑉sin(𝜔𝑡−120°)
𝑢𝐶(𝑡)= √2𝑉sin(𝜔𝑡+120°)
𝑢𝐴(𝑡)+ 𝑢𝐵(𝑡)+ 𝑢𝐶(𝑡)=0 (3.3)
𝑖𝐴(𝑡)= √2 𝐼 𝑠𝑖𝑛(𝜔𝑡−𝜙)
𝑖𝐵(𝑡)= √2 𝐼𝑠𝑖𝑛(𝜔𝑡−120°−𝜙)
𝑖𝐶(𝑡)= √2𝐼𝑠𝑖𝑛(𝜔𝑡+120°−𝜙)
𝑖𝐴(𝑡)+ 𝑖𝐵(𝑡)+ 𝑖𝐶(𝑡)=0 (3.4)
𝑝3𝜙(𝑡)=𝑝(𝑡)=𝑢𝐴𝑖𝐴+ 𝑢𝐵𝑖𝐵 +𝑢𝐶𝑖𝐶 = 𝑝𝐴+ 𝑝𝐵+ 𝑝𝐶
𝑝𝐴(𝑡)= 𝑉 𝐼 𝑐𝑜𝑠𝜙{1−𝑐𝑜𝑠2𝜔𝑡}− 𝑉 𝐼 𝑠𝑖𝑛𝜙𝑠𝑖𝑛2𝜔𝑡
𝑝𝐵(𝑡)= 𝑉 𝐼 𝑐𝑜𝑠𝜙{1 − 𝑐𝑜𝑠 (2𝜔𝑡−120°}− 𝑉 𝐼 𝑠𝑖𝑛𝜙𝑠𝑖𝑛 (2𝜔𝑡−120°)
𝑝𝐶(𝑡)= 𝑉 𝐼 𝑐𝑜𝑠𝜙{1− 𝑐𝑜𝑠 (2𝜔𝑡+120°}− 𝑉 𝐼𝑠𝑖𝑛𝜙𝑠𝑖𝑛(2𝜔𝑡+120°)(3.5)
In he case o a balanced sys em, many e ms in equa ions (3.3) cancel each o he and i can be
simpli ied in o: 𝑝3𝜙(𝑡)=𝑝(𝑡)= 3𝑉𝐼𝑐𝑜𝑠𝜙 (3.6)
Equa ion (3.4) indica es ha o balanced h ee-phase sys em, he o al ins an aneous powe is equal
o he eal powe o a e age ac i e powe (P), which is cons an . The o al eac i e powe can be
simila ly de ined as: 𝑄=𝑄𝐴+ 𝑄𝐵+𝑄𝐶= 3𝑉𝐼𝑠𝑖𝑛𝜙 (3.7)

37 | P a g e
Whe e 𝑉 and I a e he ol age and cu en magni udes, espec i ely, and a e he same o all h ee
phases. The line- o-line ol age is: 𝑉𝐿𝐿=𝑉𝐴𝐵= 𝑉𝐵𝐶= 𝑉𝐶𝐴
𝑉𝑁𝑒𝑢𝑡𝑟𝑎𝑙= 0 (3.8)
The line cu en is: 𝐼=𝐼𝐴= 𝐼𝐵= 𝐼𝐶
𝐼𝑁𝑒𝑢𝑡𝑟𝑎𝑙=0 (3.9)
Fo a h ee-phase, h ee-wi e balanced sys em, he e ec i e appa en powe is ound a e
calcula ing e ec i e ol age and cu en as:
𝑉𝑒=√𝑉𝐴𝐵
2+𝑉𝐵𝐶
2+𝑉𝐶𝐴
2
9=𝑉𝐿𝐿
√3(3.10)
𝐼𝑒=√𝐼𝐴2+𝐼𝐵
2+𝐼𝐶2
3=𝐼 (3.11)
And he e o e, he appa en powe can hen be calcula ed as:
𝑆𝑒=𝑆 =3𝑉𝑒𝐼𝑒=√3𝑉𝐿𝐿𝐼 (3.12)
In he unbalanced sys em he e is a cu en “leak” and 𝐼𝑁𝑒𝑢𝑡𝑟𝑎𝑙≠0, and he e o e he appa en
powe canno be calcula ed as desc ibed abo e as √3𝑉𝐿𝐿𝐼 , and he ull equa ion ha accoun s o
each phase and he neu al needs o be used, whe e:
𝑉𝑒=√𝑉𝐴2+𝑉𝐵2+𝑉𝐶2
3(3.13)
𝐼𝑒=√𝐼𝐴2+𝐼𝐵
2+𝐼𝐶2+𝐼𝑁
2
3=𝐼 (3.14)
While bo h CIGRE and PyPowe use he appa en powe inpu load o calcula e he supplied
ol age and cu en , PyPowe assumes a ully balanced sys em [102], while CIGRE accoun s o
he sligh unbalance. This can explain he disc epancy in he cu en calcula ion, whe e in he
balanced sys em he e is no cu en “leak” o he neu al and he e o e he phase cu en alues
a e highe han in he unbalanced sys em. Assuming a balanced h ee phase sys em has been done
in o he s udies o he LV g id and LV eede s, such as he “G id Impac s udies o elec ic ehicles
– Rein o cemen cos s in LV g ids” pe o med by he Danish Ene gy Associa ion and leading
P a g e | 38
ene gy companies Ibe d ola, EDF, ENEL, and RWE [74]. The e o e, al hough he line-cu en
esul s o he powe low a e di e en han he CIGRE benchma k, hey a e s ill alid o his
analysis.
Co-simula ion
The ini ial goals o he co-simula ion pla o m in his wo k we e:
1. Link o he so wa e TRNSYS, which is a us ed ool by esea che s in he building simula ion
ield wi h mul iple alida ed models o building a che ypes in di e en clima es.
2. P io i ize open-sou ce pla o ms/ eewa e which inc eases he possibili y o mul iple use s
o ins all and use i .
3. Explo e he FMI s anda d which migh allow o u u e exchange o models be ween pee s
wi hou comp omising in ellec ual p ope y and con iden iali y ag eemen s.
The e a e a ew co-simula ion p ojec s wi h TRNSYS, mos ly using an ad-hoc c ea ed link wi h a
speci ic single ool. A common en i onmen linked o TRNSYS in a co-simula ion pla o m is
MATLAB, which comp ises nume ous oolboxes ha enables es ing o ad anced op imiza ion
s a egies, as well as a link o Simulink. A desc ip ion o he in e changeabili y o models be ween
TRNSYS and MATLAB/Simulink is gi en in [104], and examples o es ing model p edic i e con ol
s a egies can be ound in [105] o o gene ic algo i hm op imiza ion [106]. The use o FMI o couple
TRNSYS wi h o he ools was made easie wi h he FMI complian w appe de eloped by [107],
which enables he w apping o a TRNSYS model in o a FMU. This TRNSYS FMU w appe is used
wi hin he dedica ed co-simula ion amewo k FUMOLA (de eloped on op o he P olemy II
simula ion en i onmen and he FMI++ Lib a y) which se es as he mas e and p o ides he
synch oniza ion algo i hm. Wi hin FUMOLA he TRNSYS FMU is capable o exchange wi h
MATLAB/Simulink [108], [109], Ene gyPlus, Dymola-Modelica & Powe Fac o y-DIGSILENT
[110], [111]. FUMOLA could ha e been a good op ion, howe e i called o he ins alla ion o
addi ional ools and pla o ms and was mo e complica ed o debug as he code is no as easily
accessible as open-sou ce Py hon packages, and he e o e was disca ded. I should be no ed ha an
addi ional limi a ion was imposed by he TRNSYS-FMU w appe which was designed a he ime o
TRNSYS17 which has a x32 a chi ec u e and he e o e is incompa ible wi h some co-simula ion
pla o ms ha a e based on a x64 a chi ec u e.
The choice o he ool o modelling he LV g id and pe o ming he powe low analysis was based
39 | P a g e
on goal #2, as he us ed and well-es ablished ool DIGSILENT was disca ded since i is no ee.
The g id modelling Mosaik pla o m seemed p omising as i is ee and based on py hon sc ip (which
is a high-le el language and he e o e conside ed mo e accessible and easy o unde s and o a wide
public). Mosaik has p e iously been coupled wi h o he ools such as OMNeT++ (communica ions
ool) ia di ec co-simula ion [51] o wi h Powe Fac o y, Ma lab and OpenModelica ia FMI [58],
[112] . As Mosaik is Py hon based i uses PyPowe unc ions o pe o m powe - low analysis and can
use FMI complian packages such as PyFMI [113] and FMpy[114] o co-simula ion wi h an FMU.
Howe e , al hough his op ion o e ed many ad an ages, such as a g aphic in e ace and g id analysis
ools, i also had a small addi ional s ep in e m o synch oniza ion; mosaik manages simula ion ime
as in ege s while FMI sees ime as loa s - so a loa "s ep ac o " needed o be in oduced ha
ansla es a mosaik- ime in ege in o a FMI- ime loa . The e o e, i was decided o use Py hon sc ip
o di ec ly w i e a mas e algo i hm which would coo dina e he exchange be ween he Py hon based
VS AAHP model, he TRNSYS FMUs, and he PyPowe based g id model p e iously desc ibe in
sec ions 3.1, 0 and 3.3, espec i ely. The capaci y o combine di ec ly Pypowe and PyFMI o simula e
an u ban ene gy sys em wi h FMU w apped building models has p e iously been demons a ed in
[11] and he speci ic amewo k de eloped o his wo k po ayed in Figu e 17.
Figu e 17: LV ne wo k co-simula ion amewo k
P a g e | 40
4 Resul s & Discussion
The main esul s o he loads, ol age le els and ans o me loading o he all scena ios (Table
8) will be p o ided in he ollowing sec ion
Table 8: Simula ed Scena ios
Sensi i i y analysis
Scena io
Nomencla u e
HP pene a ion a es
Ru al Base Case - no HPs
HP_0
Ru al Base Case - 30% HPs
HP_30
Ru al Base Case - 60% HPs
HP_60
Ru al Base Case - 100% HPs
HP_100
Ene gy densi y
Ru al Base Case - 100% HPs
R_BC
Ru al Wo s Case - 100% HPs
R_WC
U ban Base Case - 100% HPs
U_BC
U ban Wo s Case - 100% HPs
U_WC
Fi s will be assessed he impac o inc easing HP ins alla ions on he eede , and hen a sensi i i y
analysis o eede ene gy densi y is p o ided o he scena io whe e hea pumps domina e he en i e
esiden ial sys em, in o de o e alua e which pa s would be mos a isk and which a iables ha e
mos impac in such a case. In he ollowing analysis all compa isons will be e alua ed in e ms o
pe cen di e ence, calcula ed as:
% 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒= ‖𝑁2−𝑁1‖
𝑁2+𝑁1
2𝑥 100 (4.1)
Inc easing Hea Pump Pene a ion Ra es
The use o HPs in 0%, 30%, 60%, and 100% o he connec ed household is e alua ed in e ms o
he impac on ol age le els and ans o me loading.
1. Vol age
The ol age de ia ion o he i s node R2 (“V2”) and las node R18 (“V18”) on he eede (as
depic ed in he CIGRE model in Figu e 12), a e shown o he Spanish case in Figu e 18 and o
he Swedish case in Figu e 19. The u al base case (R_BC) is shown as i is mo e sensi i e o ol age
de ia ion han he u ban case. I can be seen ha in he Spanish case he addi ional HPs cause he
elec ical load o inc ease and he e o e he ol age le els in bo h V2 and V18 d ops mo e
signi ican ly. I can also be no ed ha he ol age le els a e less cons an (less s able) as he HP
le el inc eases, howe e hey emain in all scena ios abo e he lowe limi o 0.93 p.u..
41 | P a g e
Figu e 18: Impac o HPs on he ol age ac oss he eede in he Spanish u al base case scena io
Figu e 19: Impac o HPs on he ol age ac oss he eede in he Swedish u al base case scena io
In he Swedish case an opposi e end can be seen, and as he HP le el inc eases, so does he s abili y
o he ol age, and he ol age d op bo h a V2 and V18 dec eases. Al hough V2 s ays wi hin allowable
limi s in all cases, he ol age le el in V18 is only accep able in he case o 100% HP hea ing (HP_100).
In all cases, since he g id is domina ed by loads, and has no in eg a ed DERs, hen he e is a no able
unde ol age endency. The di e en end be ween he ol age d op inc ease o dec ease wi h
inc easing HP le els in he wo case s udies can be due o he di e ence in he hea ing load, in he
hea ing echnology being eplaced and in he HP echnology being ins alled.
2. T ans o me
The maximum ans o me loading is conside ed in his hesis as 80% o he a ed capaci y. The
ans o me loading o e he h ee analysed days wi h di e en le el o in eg a ion o HPs can be
seen in Figu e 20 o he Spanish case, and Figu e 21 o he Swedish case. The u al base case is
shown in o de o be consis en wi h he p e ious ol age analysis and because i has he same
load p o ile (same numbe o esidences pe clus e ) as he u ban base case. I can be seen in he

P a g e | 42
Spanish case ha in a pu ely esiden ial sys em wi h only con en ional na u al gas boile s (HP_0)
he only load on he LV eede is he elec ic appliances, which eaches a maximum o 12% o he
ans o me capaci y. As he HP pene a ion a e inc eases om, he peak load inc ease om 12%
o 14%, 19% and 25% o he HP_30, HP_60, and HP_100 scena ios, espec i ely (pe cen
di e ence is 17%, 45% and 70%).
Figu e 20: Impac o HPs on he ans o me in he Spanish u al base case scena io
Howe e , in all cases he ans o me is a om being o e loaded, which conco ds wi h he base
assump ion ha in mos u al cases he ans o me is o e sized and is no a isk, and ha ol age
d ops a e he main challenge.
In he Swedish case can be seen he opposi e si ua ion, whe e he con en ional hea ing sys em is di ec
elec ic hea ing hen he scena io wi h no HPs (HP_0) has he highes ans o me loading and is in
ac almos cons an ly nea he maximum allowed capaci y. I should be men ioned howe e ha a
se ies o wo s -case coldes days was chosen and he e o e, hese esul s ep esen he highes
ans o me loading ha would occu du ing he yea . In such a case he highe COP o he HP helps
educe he peak load o 110% o he ans o me capaci y in he HP_0 case o 98%, 90%, and 66%
o HP_30, HP_60 and HP_100, espec i ely (di e ence o 11%, 20% and 50%). This end was
an icipa ed by he Swedish go e nmen and is he main eason o i s ac i e suppo o new and
e o i ins alla ions o HPs.
43 | P a g e
Figu e 21: Impac o HPs on he ans o me load in he Swedish u al base case scena io
The di e ence be ween he p o iles o he Spanish and Swedish cases could be due he hea ing load
di e ence. As can be seen om he daily ans o me loading p o iles o Spain (Figu e 22) and Sweden
(Figu e 23), he i s case has an ou doo empe a u es (Tou ) oscilla ing be ween 0-10°C and he load
spikes do no all co espond wi h he empe a u e p o ile. In Figu e 24 can be seen a snapsho o
some o he moni o ed pa ame e s o he Spanish HP_100 scena io, whe e he HP o en u ns o
due o low loads, and he low cons an powe d aw du ing he day is due o elec ic appliances
(addi ional snapsho s o he esul s can be ound in Appendix IV: Simula ion Resul s).
Figu e 22: Daily ans o me loading in he Spanish R_BC case s udy
In he Swedish case he ans o me loading can be said o be in e sely co ela ed o he ou doo
wea he , as he load inc eases when he empe a u e dec eases and ice e sa. The e is a cons an
hea ing load and he small sudden cyclic spikes in he ans o me loading, ha a e mo e p ominen
as he HP pene a ion le el inc eases, a e no om low-load cycling bu a he om he HP cycling
0
5
10
15
0
5
10
15
20
25
30
24 27 30 33 36 39 42 45
Tempe a u e [°C]
T ans o me Load [%]
Time [h s]
Spain - Day 3
HP_100 HP_60 HP_30 HP_0 Tou
P a g e | 44
ope a ion in de os mode. These spikes migh be mo e mode a e in eali y, whe e he pool o hea -
d i en hea pumps include o a wide a ie y o HP echnologies (di e ing in cycling beha iou , hea
sink, cycling du a ion, and powe d aw magni ude).
Figu e 23: Daily ans o me loading in he Swedish R_BC case s udy
Figu e 24: Two households in he Spanish HP_100 scena io - indoo empe a u e (Tin4211, Tin3411 [°C]),
appliance loads (app4211, app3411[W]), and HP powe d aw (php4211, php3411[W]).
In bo h case s udies, he maximum load on he ans o me changes signi ican ly wi h inc easing HP
loads. The addi ional HPs cause he daily peak loading in he Spanish case o go up om an a e age
-25
-20
-15
-10
-5
0
0
20
40
60
80
100
120
48 51 54 57 60 63 66 69
Tempe a u e [°C]
T ans o me Load [%]
Time [h s]
Sweden - Day 3
HP_100 HP_60 HP_30 HP_0 Tou
45 | P a g e
o 11.5% o he ans o me capaci y, o 13.6%, 16.9% and 22.3% (HP_30, HP_60 and HP100,
espec i ely), while he minimum loading emains cons an in all cases, a 1% o he ans o me
capaci y (Table 9). On he o he hand, he subs i u ion o di ec elec ic hea ing by HPs in he Swedish
case educes he peak loading, as he a e age daily peak load passes om 93.4% o he ans o me
capaci y (HP_0), o 43.2% loading (HP_100)( Table 10).
Table 9: De ining alues o he ans o me loading in he Spanish case s udy
Spanish case s udy
Tou
[°C]
HP_100
[%Load]
HP_60
[%Load]
HP_30
[%Load]
HP_0
[%Load]
Daily a e age
Max
9.8
22.3
16.9
13.6
11.5
Min
0.9
1.0
1.0
1.0
1.0
A e age
5.0
7.8
5.4
4.0
2.7
Median
4.8
6.9
5.0
3.6
2.1
Table 10: De ining alues o he ans o me loading in he Swedish case s udy
Swedish case s udy
Tou [°C]
HP_100
[%Load]
HP_60
[%Load]
HP_30
[%Load]
HP_0
[%Load]
Daily a e age
Max
-3.7
43.2
64.1
77.9
93.4
Min
-13.6
5.7
24.1
36.7
48.0
A e age
-9.6
18.3
40.5
58.5
74.6
Median
-9.6
17.6
39.7
58.2
74.6
Di e en ly han in he Spanish case, in he Swedish case he minimum ans o me loading is no
cons an in all scena ios and passes om 48.0% (HP_0) o 5.7% (HP_100). This change could be
because in he Swedish case simula ed some hea ing is always equi ed.
The aging o he ans o me is accele a ed when i is epea edly o e loaded and he daily O e load
Ra io (OR) is calcula ed in his hesis as he ime he load exceeds 80% o he ans o me capaci y
o e he o al analysed ime. I is no ewo hy ha he Spanish case does no su e om o e loading
while in he Swedish case he highes occu ence o o e loading occu s in he HP_0 scena io, implying
ha i would accele a e he aging o he ans o me mo e han he scena ios wi h HPs (Table 11).
Table 11: O e loading o he ans o me o e he h ee day pe iod
HP_100
HP_60
HP_30
HP_0
O e load
a io
Spain
0.000
0.000
0.000
0.000
Sweden
0.000
0.003
0.066
0.401
In Figu e 25 can be ound a g aphical ep esen a ion o he de ining cha ac e is ics o he Spanish and
Swedish scena ios, o e he en i e h ee days.
P a g e | 52
end occu s in he summe scena io, whe e e en in he U_WC scena io he maximum loading o he
ans o me eaches 40.27% o he ans o me capaci y (Figu e 33).
Figu e 33: Daily ans o me loading in summe
The gap be ween he maximum and minimum loading o he ans o me a e shown in Figu e 34.
Only he u ban scena ios a e depic ed as bo h u al ans o me s ha e he same load as U_BC.
Al hough he magni ude is di e en be ween U_BC and U_WC, he pe cen di e ence be ween he
maximum and minimum alues is 185% in bo h cases.
Figu e 34: Spanish u ban g id - ans o me loading peaks and alleys o e he h ee-day pe iod
The o e load a io o he ans o me o e he h ee-day pe iod in U_BC and U_WC is 0, signi ying
ha he load in bo h cases does no cause much h ea o he aging o he ans o me .

53 | P a g e
Sweden: 100% Hea Pumps
1. Loads
The agg ega ed loads o each clus e o building is depic ed in Figu e 35, whe e he highes
appa en powe load o 65.50 kVA (S15_BC) and 139.59 kVA (S15_WC) occu s o he la ges
clus e o buildings (R15), while he lowes occu s a he single esidence connec ion poin (R11),
wi h 8.75 kVA (S11_BC) and 15.35 kVA (S11_WC).
Figu e 35: Agg ega ed load o each building clus e o he Swedish u ban base-case (BC) and wo s -case (WC)
scena ios
As in he Spanish case, he coincidence ac o plays an impo an ole in he ac ual peak powe , as he
single esidence (R11) has he highes peak pe cus ome (8.75 kVA) and he clus e o 11 esidences
(R15) has he lowes peak pe cus ome (5.95 kVA).
Table 14: Sweden - Peak pe cus ome o each building clus e
U_BC
U_WC
Name o clus e
R11
R15
R16
R17
R18
R11
R15
R16
R17
R18
Numbe o esidences
1
11
4
3
6
2
22
8
6
12
Peak pe esidence
8.75
5.95
7.93
6.79
6.64
7.67
6.35
6.31
6.53
6.07
In his case he pe cen change in load be ween scena ios U_BC and U_WC a he single esidence
(R11) does no change as ex emely as in he Spanish case. The g ea es di e ence be ween he
scena ios occu s in clus e R17 whe e he minimum alues changes om 0.59 kVA o 2.07 kVA
(Table 15).
P a g e | 54
Table 15: Swedish dwelling clus e load [kVA] - di e ence be ween he U_BC and U_WC cases
R11
R15
R16
R17
R18
Maximum Load U_BC
8.7
65.5
31.7
20.4
39.9
Maximum Load U_WC
15.35
139.59
50.45
39.16
72.88
Pe cen Di e ence
55%
72%
46%
63%
59%
A e age Load U_BC
1.73
19.81
7.79
5.15
10.58
A e gae Load U_WC
3.63
39.51
15.03
10.69
21.27
Pe cen Di e ence
71%
66%
63%
70%
67%
Median Load U_BC
1.27
18.41
7.82
4.81
9.79
Median Load U_WC
3.41
36.76
14.81
9.67
19.82
Pe cen Di e ence
91%
67%
62%
67%
68%
Minimum Load U_BC
0.14
4.72
0.97
0.59
1.92
Minimum Load U_WC
0.29
9.51
3.41
2.07
5.18
Pe cen Di e ence
72%
67%
111%
112%
92%
2. Vol age
The ol age de ia ion o he i s and las eede nodes a e show in Figu e 36. As in he Spanish
case, he highes ol age de ia ion occu s a he las node on he eede (R18) and is mos
signi ican in he u al cases due o he longe cable dis ances.
Figu e 36: Vol age a he i s node (V2) and he las node (V18) o he Swedish eede
I can be seen ha in R_BC he minimum ol age does no go below he s anda d limi 0 ±7% o
he nominal ol age, howe e i app oaches and eaches a minimum o 0.945 p.u.. In he case o
R_WC, ex ending he dis ance be ween he deli e y poin s causes he ol age o go well below he
55 | P a g e
allowed limi , eaching a minimum o 0.907 p.u.. The ol age d op ac oss he eede (V2-V18) eaches
a peak o 0.033 p.u. (7.59 V) and 0.064 p.u. (14.72 V) in R_BC and R_WC, espec i ely.
The co ela ion be ween he ol age le el and he load a e depic ed o he i s load node (R11) in
Figu e 37 and he las load node (R18) in Figu e 38.
Figu e 37: Co ela ion be ween ol age (V11) and load (S11) in he Swedish R_BC and R_WC g id
a he i s load node (R11)
Figu e 38: Co ela ion be ween ol age (V18) and load (S18) in he Swedish R_BC and R_WC g id
a he las load node (R18)
Along he h ee days analysed he load p o ile o R_BC eached a peak amp a e o appa en powe
o 1.16 kW/min and 3.01 kW/min in R11 and R18, espec i ely. The maximum di e ence o ol age
0
2
4
6
8
10
12
14
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1.00
0
3
6
9
12
15
18
21
24
27
30
33
36
39
42
45
48
51
54
57
60
63
66
69
Appa en powe [kVA]
Vol age [p.u.]
Time [hou s]
V11 R_BC V11 R_WC S11
0
10
20
30
40
50
60
70
0.86
0.88
0.90
0.92
0.94
0.96
0.98
1.00
0
3
6
9
12
15
18
21
24
27
30
33
36
39
42
45
48
51
54
57
60
63
66
69
Appa en powe [kVA]
Vola ge [p.u.]
Time [hou s]
V18 R_BC V18 R_WC S18
P a g e | 56
o e one imes ep in R_BC is 2.08V (R11) and 3.79V (R18), and in R_WC is 3.21V (R_11) 6.65V
(R_18). As in he Spanish case, he maximum sensi i i y o a powe change o e a 3-minu e imes ep
occu s a he las load node R18 wi h a alue o -0.84 V/kW in R_BC and -1.43 V/kW in R_WC.
The alues a e nega i e, signi ying ha a ise in powe a he node will dec ease he ol age le el a
he eede . The ol age sensi i i y ange o -0.6·10-4 o -0.3·10-4 p.u./kW is simila o he ange ound
in he s udy o an LV g id by [33].
3. T ans o me
The ans o me loading p o ile is depic ed in Figu e 39 and he same indi e ence o cable leng h
as in he Spanish case is obse ed. In he U_BC scena io occu s a peak loading o 64.3% o he
ans o me capaci y, which is s ill below he maximum allowed capaci y, and his peak coincides
wi h he lowes ou doo empe a u e pe iod o he simula ion; whe e he VS AAHP is no capable
o mee ing he en i e hea ing load and he e o e he addi ional back-up esis ance hea ing also
ope a es and inc eases signi ican ly he powe d aw. The same phenomenon occu s in he U_WC
scena io, in which case he ans o me is o e loaded and eaches a peak loading o 131.5% o i s
capaci y.
Figu e 39: Swedish ans o me load sensi i i y o he ene gy densi y
I should be men ioned ha using a hea pump wi h a highe capaci y could ha e educed he use o
he auxilia y hea e s, howe e since hea pumps end o pe o m less well as he empe a u es dec ease
i can be expec ed ha he auxilia y hea e s could s ill be needed when he ou doo empe a u e a e
especially low (in his case when i goes below -10 °C).
The gap be ween he maximum and minimum loading o he ans o me a e shown in Figu e 40.
-30
-25
-20
-15
-10
-5
0
5
0
20
40
60
80
100
120
140
0
3
6
9
12
15
18
21
24
27
30
33
36
39
42
45
48
51
54
57
60
63
66
69
Tempe a u e [ºC]
T ans o me load [%]
Simula ion Time [hou s]
R_BC
R_WC
U_BC
U_WC
Tou
57 | P a g e
Figu e 40: Swedish u ban g id - ans o me loading peaks and alleys
Al hough he magni udes changes be ween U_BC and U_WC, he pe cen di e ence be ween he
maximum and minimum alues a e he same (175%) in bo h cases.
The o e load a io o he ans o me is 0 in he U_BC case, as he loading emain below 80% o he
ans o me capaci y, and goes up o 0.019 in he U_WC case, implying ha he inc ease numbe o
cus ome s in his scena io migh cause an accele a ion o he aging o he ans o me .

P a g e | 58
5 Conclusions
The ollowing will discuss he mos impo an indings o his wo k, i s limi a ions and in e es ing
u u e esea ch di ec ions.
Main Poin s o In e es
As single buildings, in gene al, will only ma ginally in luence he s a e o he elec ical g id, and
he e o e an analysis on he LV eede le el and he MV/LV ans o me seconda y subs a ion
pe o med in his wo k is o impo ance o many s akeholde s in he ene gy ield. This esea ch
di ec ion is especially ele an o he e alua ion o he po en ial o hea pumps o Demand Side
Managemen (DSM), o he s a egic in eg a ion o DERs, and o assessmen o demand and
supply lexibili y oppo uni ies.
1. Co-simula ion
• The p oposed co-simula ion pla o m ook ad an age o es ablished domain-speci ic simula o s;
TRNSYS o building simula ion, and Pypowe o g id modelling and powe - low simula ions.
• The co-simula ion en i onmen is es ed wi h wo case s udies and he esul s seem plausible and
cohe en wi h esea ch in his ield.
• Co-simula ion suppo s he use o high- esolu ion da a.
2. Household Loads
• The agg ega ed loads o indi idual households wi h high- esolu ion unique p o iles is shown in
his wo k.
• I is obse ed in Figu e 41 ha he peak load pe household wi h inc easing agg ega ion le els
(SPN and SWD ep esen he Spanish and Swedish case s udies, espec i ely) ollows a ypical
end simila o he Velande load agg ega ion me hod (VL_NoElec, VL_Elec1, VL_Elec2
ep esen calcula ions o a house wi hou elec ic hea ing, a co age wi h elec ic hea ing and a
la ge house wi h elec ic hea ing, espec i ely) and he Spanish egula ion Simul anei y Fac o (SF)
me hods (SF_NoElec and SF_Elec, ep esen a house wi hou and wi h elec ic HVAC,
espec i ely). Fo mo e de ail abou hese me hods see “Appendix II: Coincidence Fac o ”. I
should be no ed ha he cu es o all he case s udies o his hesis a e less smoo h o 3 and 4
agg ega ed households because a hese le els he in luence o he single esidence on he peak is
59 | P a g e
highe . Fo example, i a clus e o 3 households includes an “ou lie ” p o ile wi h ela i ely high
appliance loads, i will acco dingly ha e a highe load pe cus ome han wha would be expec ed
on a e age. Wi h highe le els o agg ega ion, he coincidence ac o ends o be lowe and he
balance be ween he household powe d aws end o smoo hen ou “ou lie ” p o iles.
Figu e 41: Peak load pe household o inc easing numbe o households conside ed
• The subjec o me hods o agg ega ing domes ic load p o iles (using indi idual s ochas ic p o iles
o using coincidence o simul anei y ac o equa ions) is indi ec ly app oached in his wo k. I can
be seen in Figu e 42 ha he esul ing maximum clus e loads (Smax) in his wo k, occu ing i all
households wi hin a clus e ha e a peak a he same ime (coincidence ac o =1), do no exac ly
ma ch he maximum loads desc ibed in CIGRE, o ea lie wo k by Pa hanassiou [35] on which
he CIGRE LV g id is based. I should be no ed ha al hough no commonly occu ing in eali y,
Smax ep esen s an ex eme case which could occu occasionally o special e en s, desc ibed in
e ms o he social andom ac o in [34] ha al e s he ypical daily load p o ile o a household
and inc ease/dec ease he coincidence ac o o ce ain appliances (e.g. FIFA Wo ld Cup
championship). The esul ing ac ual peak powe o he building clus e s (So) and he maximum
possible peak powe (Smax) o R_BC and U_WC scena ios o he Spanish and Swedish case
s udies a e p esen ed in Table 16. The ocus on hese wo scena ios is because he g id
con igu a ion o R_BC is closes o he o iginal benchma k, while U_WC has he highes loads.
F om he compa ison be ween he esul s o his hesis and he benchma ks desc ibed by CIGRE
and Pa hanassiou e al. he me hod employed o ep esen ing he load di e si y and coincidence
ac o could ha e a signi ican in luence on he simula ion. Howe e , he gap can also be due o
he ac ha CIGRE p esen s a s a ic model wi h p e-de ined esiden ial loads which a e no
de ailed in e m o building ypology, en elope, and HVAC sys ems, and mos p obably di e
om he models used in hesis. I is no ewo hy o men ion also ha ypical load p o iles di e
acco ding o socio-geog aphical condi ions and as CIGRE o e s a gene al Eu opean benchma k,
he su ounding en i onmen is unknown and migh be ano he sou ce o he di e ence in he
0
2
4
6
8
10
12
14
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Peak load pe household [kW]
Numbe o households
VL_NoElec
VL_Elec1
VL_Elec2
SF_NoElec
SF_Elec
SWD - HP_0
SWD - HP_100
SPN - HP_0
SPN - HP_100
P a g e | 60
loads
Figu e 42: Clus e o agg eaga ed household loads - Smax [kVA]
I should be men ioned ha in he ollowing able can also be seen he di e ence ben weeen he wo
benchma ks by CIGRE and Pa hanassiou e al., whe e R11, R16, and R18 a e iden ical bu he CIGRE
R15 is slighl y lowe and R17 is much highe han in he benchma k o Pa hanassiou e al. Addi ionally,
he numbe o esidences pe clus e is no desc ibed in he CIGRE benchma k, and nei he a e he
So alues, while in Pa hanassiou e al. R15 and R18 a e only desc ibed as apa men buildings, wi hou
men ioning he numbe o cus ome s.
Table 16: Peak load o building clus e s
R11
R15
R16
R17
R18
Spain R_BC
Numbe o houses
1
11
4
3
6
So
6.7
28.9
16.9
8.6
18.5
Smax
6.7
55.8
21.4
16.4
28.2
Sweden R_BC
Numbe o houses
1
11
4
3
6
So
8.7
65.5
31.7
20.4
39.9
Smax
8.7
90.4
33.8
25.8
47.6
Spain U_WC
Numbe o houses
2
22
8
6
12
So
8.3
52.6
28.6
16.2
31.2
Smax
13.7
124.0
42.0
35.1
62.8
Sweden U_WC
Numbe o houses
2
22
8
6
12
So
15.3
139.6
50.5
39.2
72.9
Smax
15.9
179.9
64.4
52.0
100.5
Pa hanassiou
Numbe o houses
1
-
4
1
-
So
5.7
57.0
25.0
5.7
25.0
Smax
15.0
72.0
55.0
15.0
47.0
CIGRE
Numbe o houses
-
-
-
-
-
So
-
-
-
-
-
Smax
15.0
52.0
55.0
35.0
47.0
0
50
100
150
200
R11 R15 R16 R17 R18
Clus e loads -
Smax [kVA]
Smax - Spain R_BC
Smax - Sweden R_BC
Smax - Spain U_WC
Smax - Sweden U_WC
Smax - Pa hanassiou
Smax - CIGRE
61 | P a g e
3. Vol age
• This wo k con i ms ha u al LV eede s a e mo e p one o ol age d op and de ia ion due o
he longe eede cables, as was ound in [46].
• The minimum ol age le els do dec ease when he addi ional HP load a e added o he g id,
howe e i seems ha he b eaching o he allowable de ia ion o ±7% om he nominal is mo e
p ominen in colde clima es whe e he hea ing load, and consequen ly he HP powe d aw, is
g ea e . F om all he scena ios wi h 100% HPs, only he Swedish u al wo s -case scena io
exceeded he accep able ol age ange. I should be kep in mind ha hese esul s a e alid o
he benchma k simula ed, and could be di e en o a speci ic eal g id wi h di e en
con igu a ions and cables.
• The cha ac e is ic powe d aw in cycling ope a ion po ayed by he VS AAHP model de eloped
in his wo k can ha e a conside able e ec on he ol age le els and peak powe loading o he
ans o me and should no be neglec ed in s udies abou he in e ac ion be ween HPs and he
LV elec ic g id.
• The HPs in his wo k a e hea d i en and he e o e he sudden hikes in powe consump ion as
hey u n on only occu s in low-load o de os cycling condi ions, howe e i could hinde mo e
complex g id d i en a i -based con ol s a egies i all HPs a e designed o u n o du ing high
a i s and back on as soon as he a i s a e low. The e o e, a coo dina ion be ween he HP would
be sugges ed in o de o p o ec he g id equipmen and p olong i s li e ime.
• The e ec s o local DERs o balance he unde ol age should be u he explo ed using he co-
simula ion pla o m p esen ed in his wo k.
4. T ans o me Loading
• The isk o he ans o me is dependen on he hea ing load he HPs a e o supply and he
ene gy densi y o he eede . In all Spanish scena ios, he hea ing and cooling loads a e manageable
by he HP and bo h he u al and u ban g ids can accommoda e a pene a ion a e o 100% HPs.
In he Swedish case he comple e swi ch o HPs educed he peak ans o me loading and he
u ban wo s -case scena io is he only es ed g id o su e om ans o me o e load.
• In cases whe e he HPs eplace con en ional elec ic esis ance hea e s, he load on he
ans o me is educed. On he o he hand, in cases whe e he HPs eplace a non-elec ic based
hea ing sys em, he load on he ans o me will inc ease.
P a g e | 68
1–8, 2004.
[81] A. Teje o, J. O iz, and A. Wagne , “Nea ly Ze o Ene gy buildings – Impac o use beha io ,”
Ka ls uhe In i u u Technologie, 2017.
[82] Roin e, “Ven ajas e incon enien es de los p incipales ipos de cale acción.” [Online]. A ailable:
h ps:// oin e.com/ ipos-de-cale accion/. [Accessed: 07-Sep-2018].
[83] M. Figue uelo, “Cómo calen a la casa y no mo i en el in en o,” elEconomis a, 2014. [Online].
A ailable: h p://www.eleconomis a.es/emp esas-
inanzas/consumo/no icias/6314446/12/14/Como-calen a -la-casa-y-no-mo i -en-el-
in en o.h ml. [Accessed: 07-Sep-2018].
[84] El Mundo, “El 66% de españoles no dispone de un sis ema de cale acción que sea adecuado pa a
su i ienda,” El Mundo, 2017. [Online]. A ailable:
h p://www.elmundo.es/economia/ i ienda/2017/11/16/5a0d66c6468aeb4d7d8b4641.h ml.
[Accessed: 07-Sep-2018].
[85] R. To anin, “Ene gy lexibili y s a egies o esiden ial buildings in he Medi e anean clima e,”
Uni e si a Poli ècnica de Ca alunya – Ba celona Tech; Ca alonia Ins i u e o Ene gy Resea ch -
IREC, 2018.
[86] Eu opean Hea Pump Associa ion (EHPA), “Eu opean Hea Pump Ma ke and S a is ics Repo
2015,” p. 4, 2015.
[87] B. La son, B. Hannas, P. S o m, and D. Baylon, “Duc less Hea Pump Cold Clima e Pe o mance
E alua ion,” 2013.
[88] ITEC (Ins i u de Teconologia de la Cons ucció de Ca alunya), “No ma Reglamen à ia
D’Edi icació Sob e Aïllamen Tè mic,” 1987.
[89] L. G. Swan, V. I. Ugu sal, and I. Beausoleil-Mo ison, “A da abase o house desc ip ions
ep esen a i e o he Canadian housing s ock o coupling o building ene gy pe o mance
simula ion,” J. Build. Pe o m. Simul., ol. 2, no. 2, pp. 75–84, 2009.
[90] Eu opean Commission, “EU Building Da abase,” 2014. [Online]. A ailable:
h ps://ec.eu opa.eu/ene gy/en/eu-buildings-da abase. [Accessed: 22-Aug-2018].
[91] J. O iz, A. Fonseca i Casas, J. Salom, N. Ga ido So iano, and P. Fonseca i Casas, “Cos -e ec i e
analysis o selec ing ene gy e iciency measu es o e u bishmen o esiden ial buildings in
Ca alonia,” Ene gy Build., ol. 128, pp. 442–457, 2016.
[92] M. Kegel, R. Sunye, and J. Tamasauskas, “Li e Cycle Cos Compa ison and Op imisa ion o
Di e en Hea Pump Sys ems in he Canadian Clima e,” P oc. eSim, pp. 492–505, 2012.
[93] J. Tamasauskas, M. Kegel, and R. Sunye, “An analysis o deep ene gy e o i s a egies in he
exis ing Canadian esiden ial ma ke ,” in 13 h Con e ence o In e na ional Building Pe o mance Simula ion
Associa ion, Chambé y, F ance, 2013, no. Augus 2014, pp. 169–176.
[94] Swedish Re ige a ion & Hea Pump Associa ion, “PULSEN 2018 - annual su ey o e aile s and
ins alle s o hea pumps o consume s,” 2018. [Online]. A ailable: h ps://sk p.se/ak uell -o-
opinion/s a is ik/pulsen/2018-eng. [Accessed: 30-Aug-2018].
[95] I. Sa o i, J. O iz, J. Salom, and U. I. Da , “Es ima ion o load and gene a ion peaks in esiden ial
neighbou hoods wi h BIPV: bo om-up simula ions s . Velande ,” in Wo ld Sus ainable Building 2014
Con e ence, 2014, pp. 17–24.
[96] J. O iz, F. Gua ino, J. Salom, C. Co che o, and M. Cellu a, “S ochas ic model o elec ical loads
in Medi e anean esiden ial buildings: Valida ion and applica ions,” Ene gy Build., ol. 80, pp. 23–
36, 2014.
[97] J. Salom, J. Widén, J. Candanedo, and K. B. Lindbe g, “Analysis o g id in e ac ion indica o s in ne
ze o-ene gy buildings wi h sub-hou ly collec ed da a,” Ad . Build. Ene gy Res., ol. 9, no. 1, pp. 89–
106, 2015.
[98] L. Thu ne , A. Scheidle , F. Sch, J. Menke, J. Dollichon, and F. Meie , “pandapowe - an Open
Sou ce Py hon Tool o Con enien Modeling , Analysis and Op imiza ion o Elec ic Powe

69 | P a g e
Sys ems,” IEEE T ans. POWER Sys ., 2018.
[99] J. Salom, A. J. Ma szal, J. Widén, J. Candanedo, and K. B. Lindbe g, “Analysis o load ma ch and
g id in e ac ion indica o s in ne ze o ene gy buildings wi h simula ed and moni o ed da a,” Appl.
Ene gy, ol. 136, pp. 119–131, 2014.
[100] M. G. Flammini and G. P e ico, “In e ac ion o Consume s , Pho o ol aic Sys ems and Elec ic
Vehicle Ene gy Demand in a Re e ence Ne wo k Model,” in 2017 In e na ional Con e ence o Elec ical
and Elec onic Technologies o Au omo i e, 2017.
[101] H. Wilkening and C. T oye , “Modelling Residen ial Load P o iles and i s Applica ions,” 2016.
[102] L. Thu ne and A. Scheidle , “Pandapowe - Con enien Powe Sys em Modelling and Analysis
based on PYPOWER and pandas Lead,” 2017.
[103] M. Kuma , “THREE PHASE CIRCUITS: POWER DEFINITIONS AND VARIOUS
COMPONENTS,” in Powe Quali y in Powe Dis ibu ion Sys ems, Na ional P og amme on Technology
Enhanced Lea ning (NPTEL), 2012.
[104] P. Riede e , V. Duc eux, and W. Keilholz, “Coupling o T nsys Wi h Simulink – a Me hod To
Au oma ically Expo and Use T nsys Models Wi hin Simulink and Vice Ve sa,” in Ele en h
In e na ional IBPSA Con e ence Glasgow Sco land, 2009, pp. 1628–1633.
[105] N. Alibabaei, A. S. Fung, and K. Raahemi a , “De elopmen o Ma lab-TRNSYS co-simula o o
applying p edic i e s a egy planning models on esiden ial house HVAC sys em,” Ene gy Build., ol.
128, pp. 81–98, 2016.
[106] M. Jones, “COUPLING TRNSYS AND MATLAB FOR GENETIC ALGORITHM
OPTIMIZATION IN SUSTAINABLE BUILDING DESIGN,” in BauSim2010, 2010.
[107] E. Widl and W. Mülle , “Gene ic FMI-complian Simula ion Tool Coupling,” in 12 h In e na ional
Modelica Con e ence, 2017, pp. 321–327.
[108] K. Ede , E. Widl, G. Con ad, B. Beigelbock, and F. Judex, “CO-SIMULATION FOR CONTROL
DESIGN - A CASE STUDY FOR CROSS-DOMAIN COLLABORATION,” in ASHRAE and
IBPSA-USA SimBuild 2016, 2016.
[109] E. Widl, F. Judex, K. Ede , and P. Palensky, “FMI-based co-simula ion o hyb id closed-loop
con ol sys em models,” P oc. 2015 In . Con . Complex Sys . Eng. ICCSE 2015, 2016.
[110] E. Widl, W. Mulle , D. Bascio i, S. Henein, S. Haue , and K. Ede , “Simula ion o mul i-domain
ene gy sys ems based on he unc ional mock-up in e ace speci ica ion,” P oc. - 2015 In . Symp. Sma
Elec . Dis ib. Sys . Technol. EDST 2015, no. Sep embe 2015, pp. 510–515, 2015.
[111] A. Elsheikh e al., “MODELICA-ENABLED RAPID PROTOTYPING VIA TRNSYS,” in 13 h
Con e ence o In e na ional Building Pe o mance Simula ion Associa ion, 2013.
[112] S. Rohjans, E. Widl, W. Mülle , S. Schü e, and S. Lehnho , “Co-Simula ion o Complex Ene gy
Sys ems wi h MOSAIK and FMI,” Au oma isie ungs echnik, ol. 62, no. 5, pp. 325–336, 2014.
[113] C. Ande sson, J. Akesson, and C. Fuh e , “PyFMI : A Py hon Package o Simula ion o Coupled
Dynamic Models Facul y o Enginee ing,” Ma h. Sci., ol. 2, 2016.
[114] Dassaul Sys èmes, “FMPy,” 2018. [Online]. A ailable: h ps:// mpy. ead hedocs.io/en/la es /.
[Accessed: 10-Aug-2018].
[115] J. Dicke and P. Schegne , “Residen ial load models o ne wo k planning pu poses,” in Mode n
Elec ic Powe Sys ems (MEPS), 2010 P oceedings o he In e na ional Symposium, 2010, no. Sep embe .
[116] MODELISAR conso ium, “Func ional Mock-up In e ace o Model Exchange and Co-
Simula ion,” 2014.
[117] Aus ian Ins i u e o Technology (AIT), “The FMI++ Lib a y - Ve sion 1.0 - Documen a ion,”
2015.
[118] S. Sche ke, “mosaik Documen a ion, Release 2.0,” p. 102, 2014.
P a g e | 70
Appendix I: CIGRE Eu opean Residen ial LV G id
CIGRE and PyPowe LV Residen ial G id
The ollowing p o ides he main elemen s o he Eu opean esiden ial LV eede benchma k
desc ibed in CIGRE and modelled in Pypowe . The ans o me is desc ibed in Figu e 45, he
inpu loads a e p o ided in Figu e 46, and he cables in o ma ion and model a e ound in Figu e
47 and Figu e 48.
Figu e 45: T ans o me - CIGRE da a and PyPowe model
Figu e 46: Loads - CIGRE da a and PyPowe model
71 | P a g e
Figu e 47: Geome y o unde g ound lines - CIGRE da a and PyPowe model
Figu e 48: Lines - CIGRE da a and Pypowe model
P a g e | 72
CIGRE Powe Flow Resul s
Figu e 49: CIGRE powe low esul s I
73 | P a g e
Figu e 50: CIGRE powe low esul s II

P a g e | 74
Appendix II: Coincidence Fac o
The e a e se e al app oaches o es ima e he coinciden peak load o an agg ega ed clus e o
dwellings, such as he use o coincidence o simul anei y ac o s, he Velande 's o mula, and
p ocedu e based on s a is ical analysis [95].
Coincidence Fac o
The coinciden ac o (𝑐) is de ined as he coinciden peak demand 𝑃𝑚𝑎𝑥(𝑛) o a g oup o
cus ome s wi hin a speci ied pe iod o he sum o hei indi idual maximum demands wi hin he
same pe iod and can be be ween 0 and 1 [115]. I assuming ha all cus ome s ha e he same peak
demand (𝑃max1) , o ha an a e age pe cus ome is applied, hen he coincidence ac o can be
calcula ed as:
𝑐= 𝑃𝑚𝑎𝑥(𝑛)
𝑛∙𝑃max1= 𝑃0
𝑃max1+ (1 − 𝑃0
𝑃max1)∙ 1
√𝑛 = 𝑐∞+(1−𝑐∞)
√𝑛 (𝐴.1)
The ac o 𝑐∞ depends on he elec i ica ion le el o he households.
The coincidence peak demand o a g oup o simila cus ome s can hen be calcula ed wi h he
simple equa ion: 𝑃𝑚𝑎𝑥(𝑛)=𝑛∙𝑐∙𝑃max1 (𝐴.2)
In Figu e 51 can be seen ha he peak load pe cus ome calcula ion as he agg ega ion le el
inc eases is in luenced by he coincidence ac o
Figu e 51: Peak load con ibu ion pe cus ome as he agg ega ion le el inc eases o di e en coincidence
ac o s [115]
75 | P a g e
Velande ’s Fo mula
Velande ’s o mula is widely used in Scandina ia and i used o calcula e he maximum demand
pe cus ome 𝑃max1 based on he annual ene gy consump ion:
𝑃max1= 𝑘1∙ 𝐸+ 𝑘2∙√𝐸(𝐴.3)
Whe e he ac o s 𝑘1 and 𝑘2 a e empi ical coe icien s and some sample alues can be ound in
[95], [115] . The coincidence peak demand o 𝑛 cus ome s can be calcula ed as:
𝑃𝑚𝑎𝑥(𝑛)=𝑛∙𝑘1∙ 𝐸+ 𝑘2∙√𝑛∙𝐸 (𝐴.4)
The ac o s used in hesis a e gi en in Table 17, and E is assumed o be app oxima ely 16,500
kWh/(m2y) [95].
Table 17: Sample Velande 's coe icien s [95]
Simul anei y Fac o
Acco ding o he Spanish egula ion o designing he dis ibu ion g id he ollowing equa ion is
o be used o calcula e 𝑃𝑚𝑎𝑥(𝑛) o an agg ega ed clus e o dwellings:
𝑃𝑚𝑎𝑥(𝑛)=𝑃𝑚𝑎𝑥,𝑟𝑒𝑓∙𝑆𝐹(𝑛) (𝐴.5)
Whe e 𝑃𝑚𝑎𝑥,𝑟𝑒𝑓 is he e e ence powe load and is assumed o be 5.75 kW o a esidence wi hou
elec ic hea ing/cooling and 9.2 kW o a esidence wi h elec ic hea ing/cooling. The Simul anei y
Fac o (𝑆𝐹) depends on 𝑛 and he alues a e gi en in Table 18:
Table 18: Simul anei y Fac o o calcula ing he peak demand o an agg ega ed clus e o dwellings in Spain
[95]
P a g e | 76
CIGRE Load Coincidence Fac o
The coincidence ac o (𝑐) is a unc ion o he 𝑛 and he calcula ion o an equi alen appa en
powe load 𝑆𝑚𝑎𝑥(𝑛) can be pe o med by summing he indi idual loads 𝑆𝑖 , and mul iplying by
he coincidence ac o , as in he ollowing equa ion[36]:
𝑆𝑚𝑎𝑥(𝑛)=𝑐∑𝑆𝑖
𝑛
𝑖=1 (𝐴.6)
Whe e 𝑐=0.6(1+1 𝑛
⁄ ).
Fo he powe low calcula ion he agg ega ed esidence clus e will ha e an equi alen powe
ac o (𝑝𝑓) which is calcula ed as:
𝑝𝑓𝑒𝑞=𝑃𝑒𝑞
√𝑃𝑒𝑞
2+ 𝑄𝑒𝑞
2(𝐴.7)
Whe e 𝑃𝑒𝑞 and 𝑄𝑒𝑞 a e he equi alen ac i e and eac i e powe o he agg ega ed clus e and a e
calcula ed as:
𝑃𝑒𝑞= ∑(𝑆𝑖∙𝑝𝑓𝑖)
𝑛
𝑖=1 (𝐴.8)
𝑄𝑒𝑞=∑𝑆𝑖∙ sin (a ccos(𝑝𝑓𝑖))
𝑛
𝑖=1 (𝐴.9)
77 | P a g e
Appendix III: Co-simula ion Tools
The main ools and pla o ms ha ha e been examined o he co-simula ion en i onmen o his
hesis a e desc ibed in he ollowing sec ions.
Func ional Mock-up In e ace
Func ional Mock-up In e ace (FMI) is a gene al s anda d ha suppo s bo h model exchange and
co-simula ion. A simula ion model ha complies wi h he FMI s anda d is w apped in a Func ional
Mock-up Uni (FMU). The FMI o Co-Simula ion (FMI-CS) assumes ha he FMU includes a
sol e and hus may independen ly simula e when called. FMI o Model Exchange (FMI-ME), on
he o he hand, expec s he mas e algo i hm o sol e he simula ion model p o ided by he FMU.
The FMU is dis ibu ed as a zip- olde ha con ains he ollowing main elemen s [116]:
• modelDesc ip ion.xml ile: An XML ile con ains he de ini ion o all he main a iables in
he FMU such as ime s ep, inpu and ou pu s, as well as he capaci y o he FMU o handle
ad anced algo i hms such as a iable communica ion ime-s eps, signal ex apola ion, o
o he s.
• Bina ies olde : All needed model equa ions o he access o co-simula ion ools a e p o ided
wi h a small se o easy o use C unc ions. These C unc ions can ei he be p o ided in sou ce
and/o bina y o m and implemen ing he ac ual in e ace (as C API).
• Resou ces: Fu he da a can be included in he FMU zip ile, especially a model icon (bi map
ile), documen a ion iles, maps and ables needed by he FMU, and/o all objec lib a ies o
dynamic link lib a ies ha a e u ilized.
To acili a e he handling o FMUs, he FMI++ oolbox [117] has been de eloped . Opensou ce ools
ha acili a e he use o CS and ME wi h FMI ha e been de eloped by [107] and a e eely dis ibu ed
o he public as FMU w appe s, as well as Py hon packages ha can be used o communica e wi h a
Py hon based sc ip “Mas e Algo i hm” (Table 19).
Table 19: F ee FMU w appe s o simula ion ool o ene gy ne wo ks
Tool
Websi e
Las upda ed
Powe Fac o y
h ps://sou ce o ge.ne /p ojec s/powe ac o y- mu/
17 Aug 2018
Ma lab
h ps://sou ce o ge.ne /p ojec s/ma lab- mu/
19 Jul 2018
TRNSYS
h ps://sou ce o ge.ne /p ojec s/ nsys- mu/
08 Feb 2018
FMpy
h ps://pypi.o g/p ojec / mpy/
4 Sep 2018
PyFMI
h ps://pypi.o g/p ojec /PyFMI/
14 Ma 2017
P a g e | 84
Figu e 62: Two households in he Spanish R_BC Cooling scena io
Daily T ans o me Loading
Table 20: Spain - Daily ans o me loading in e ms o % o capaci y
Spanish
Case s udy
Tou
[°C]
HP_100
[%capaci y]
HP_60
[%capaci y]
HP_30
[%capaci y]
HP_0
[%capaci y]
Day 1
Max
9.5
22.6
16.6
14.2
12.1
Min
0.0
1.0
1.0
1.0
1.0
Gap
9.5
21.7
15.7
13.3
11.1
A e age
4.2
7.9
5.5
4.0
2.7
Median
4.7
6.9
4.9
3.5
2.2
Day 2
Max
7.5
19.2
14.9
12.3
11.0
Min
1.6
1.0
1.0
1.0
1.0
Gap
5.9
18.2
14.0
11.3
10.1

85 | P a g e
A e age
4.8
8.0
5.6
4.1
2.8
Median
4.8
6.9
5.2
4.0
2.1
Day 3
Max
12.4
25.1
19.0
14.3
11.4
Min
1.2
1.0
1.0
1.0
1.0
Gap
11.2
24.1
18.0
13.3
10.5
A e age
6.0
7.7
5.2
3.7
2.5
Median
4.9
7.0
4.8
3.5
2.0
Table 21: Sweden - Daily ans o me loading in e ms o % o capaci y
Swedish
Case s udy
Tou
[°C]
HP_100
[%capaci y]
HP_60
[%capaci y]
HP_30
[%capaci y]
HP_0
[%capaci y]
Day 1
Max
-7.9
36.0
53.8
71.4
89.4
Min
-12.2
4.3
32.0
49.7
64.7
Gap
4.3
31.7
21.9
21.8
24.7
A e age
-9.4
18.2
42.6
62.4
80.2
Median
-8.8
17.6
42.0
62.6
80.5
Day 2
Max
1.9
27.7
48.7
64.0
80.9
Min
-8.7
5.0
12.7
18.4
23.7
Gap
10.6
22.7
36.1
45.6
57.3
A e age
-3.5
12.5
31.4
46.6
60.1
Median
-3.1
12.1
31.4
46.5
59.4
Day 3
Max
-5.1
65.7
89.7
98.4
110.0
Min
-19.9
7.7
27.7
42.0
55.5
Gap
14.7
58.0
62.0
56.3
54.5
A e age
-15.9
24.1
47.3
66.4
83.5
Median
-16.8
23.0
45.6
65.4
83.8