Dynamically adding que ies o
p oduc ion moni o ing a no cos .
The P oblem: A Da a-Rich, Insigh -Poo G id:
•Da a O e load:Sma me e s (AMI) and PMUs
gene a e massi e, high- eloci y da a s eams.
•Reac i e Moni o ing:Ale s igge ed a e ol age
h esholds a e b eached, which is oo la e.
•Poo Selec i i y:F equen ale s o empo a y, sel -
co ec ing e en s lead o unnecessa y diagnos ics.
•Compu a ionally Was e ul:T adi ional sys ems e-
p ocess aw da a o e e y new ques ion o que y.
The Vision: A P oac i e, Fu u e-P oo Sys em
•Fas & Scalable:Mus be able o p ocess massi e
da a s eams om sma me e s, PMUs in eal- ime.
•E icien S o age & Compu a ion:Enable long- e m
his o ical/ empo al analysis wi hou p ohibi i e cos s.
•High-Accu acy & Con ex -Awa e:Le e age he
g id's hie a chy o unde s and spa ial co ela ions.
Mo i a ion
Ou key idea is o combine he con ex ual accu acy o hie a chical analysis wi h
he pe o mance bene i s o da a summa iza ion o achie e high-accu acy, high-
pe o mance and low-cos analy ics.
1. Model he Topology: We i s map he g id's physical hie a chy, es ablishing he
ela ionships be ween asse s (me e s → junc ion boxes → ans o me s).
2. Summa ize Once: A one-pass summa iza ion algo i hm (e.g., Disc e e Wa ele
T ans o m) is applied o he ime-se ies da a o c ea e a pe sis en , mul i-scale
summa y ha is o de s o magni ude smalle han he aw da a.
3. Run Any hing: This compac summa y becomes he single sou ce o mul iple
analy ics. I allows o dynamic, on- he- ly que ies ac oss any ime-scale o spa ial-
le el wi hou e e needing o e-p ocess he aw da a.
Insigh s
The Hie a chyScope F amewo k
Spa ial (Hie a chical) Co ela ion
•Sma me e s a e no independen ; hey a e g ouped by
physical connec ions (e.g., dis ibu ion ans o me s).
• A sys emic issue (e.g., ans o me aul ) will mani es as
a co ela ed pa e n ac oss a g oup.
•An indi idual issue (e.g., aul y me e ) will be an ou lie
wi hin i s g oup.
Da a Summa iza ion
• E icien ly p oduce compac ep esen a ion o he da a.
• Ene gy/Resou ce e icien , as /scalable, con igu able.
• Use-cases: moni o ing, co ela ion analysis, e c.
Re e ences
A. J. an Rooij, V. Gulisano, and M. Papa ian a ilou, “LoCoVol : Dis ibu ed De ec ion o B oken Me e s in Sma G ids h ough S eam P ocessing,” in P oceedings o he 12 h ACM In e na ional Con e ence on Dis ibu ed and E en -based Sys ems, in DEBS ’18.
New Yo k, NY, USA: Associa ion o Compu ing Machine y, June 2018, pp. 171–182. doi: 10.1145/3210284.3210298.
B. W. R. Pun e , O. Papape ou, and M. Ga o alakis, “OmniSke ch: E icien Mul i-Dimensional High-Veloci y S eam Analy ics wi h A bi a y P edica es,” P oc. VLDB Endow., ol. 17, no. 3, pp. 319–331, No . 2023, doi: 10.14778/3632093.3632098.
C. I. My ilinis, D. Tsoumakos, and N. Kozi is, “Wo kload-awa e wa ele synopses o sliding window agg ega es,” Dis ib Pa allel Da abases, ol. 39, no. 2, pp. 445–482, June 2021, doi: 10.1007/s10619-020-07307-w.
Vinh Quang Ngo, Jo is an Rooij and Ma ina Papa ian a ilou
Summa ies
Sma me e s
PMUs
Da a Summa iza ion
Que y di e en le els o he
hie a chy (e.g., child o some
ans o me _id, subs a ion_id)?
Que y based on some augmen ed
a ibu es (e.g., pos al_code,
cus ome _ ype, has_sola ) ?
Que y using di e en
ime-g anula i y (e.g., g oup by 10m,
1h om las yea , las 2 yea s).
High-Vol age T ansmission
230kV
Dis ibu ion T ans o me s
240V
Junc ion Boxes
Household sma me e s
Junc ion Box A
Junc ion Box A1 Junc ion Box A2
SM A11 SM A21 SM A22
Da a Analy ics
Moni o ing
Subs a ion T ans o me s
11kV
Subs a ion T ans o me s
69kV
…
…
MSCA Doc o al Ne wo k RELAX-DN - EU
Ho izon 2021-27 -Relaxed Seman ics
Ac oss he Da a Analy ics S ack
Swedish Elec ici y
S o age and
Balancing Cen e
Gö ebo g Ene gi
Hie a chyScope: Spa io-Tempo al Da a Summa iza ion o E icien
Vol age Fo ecas ing and Anomaly De ec ion in Hie a chical Powe G ids
Pos e
