Solar Thermal Plants Integration in Smart Grids

Solar Thermal Plants Integration in Smart Grids E. F. Camacho ∗A. J. del Real ∗∗ C.Bordons ∗A. Arce ∗ ∗Dept. System Engineering and Automation, University of Seville School of Engineering (email: eduar[email protected], bor[email protected], [email protected]) ∗∗ IDENER R&D (e-mail: alejandro.delre[email protected]) Abstract: Solar energy penetration has been increasingly growing in recentyears. Since solar energy is intermittentits integration in existing grids is difficult. This paper deals with the optimal integration of solar power plants in grids. The paper proposes amodification of energy hubs whichallows to solvethe optimization problem with amixed integer programming algorithm in adistributed way.An introductory simulation study case is given. 1. INTRODUCTION Distributed energy resources, comprising distributed power generators and energy storage units, can playan importantrole in supporting key policy objectives of combating climate change, increasing the amountof electricitygenerated from renewable sources, and enhancing energy saving. Distributed generation of electricity,via solar (thermal or photo-voltaics), wind turbines, or combined heat and power plants, has a goodchance of penetration in the electricityinfrastructure network in the future (Houwing et al. [2007]). Large-scale diffusion of distributed energy resources will havea deep impact on the functioning of the electricityinfrastructure: It will bring radical changes to the traditional model of generation and supply as well as to the business model of the energy industry (Houwing et al. [2007]). In order to copewith this, the electrical power grid is undergoing amajor renovation, that will help to meet the power qualityand power availabilitydemands of the 21st century.The new power grid, whichis also called as the smart grid, aims to integrate the recenttechnological advancements in the Information and Communication Technology (ICT) field to the power engineering field. The presentsmart grid implementations focus on smart meter based utility-to-meter and utility-to customer communications. Although these features provide significantimprovements on the customer managementside, in the following decades, grid managementwill beone of the major ICTdominantfields. The future of power grids is expected to involvean increasing level of intelligence and integration of new information and communication technologies in every aspect of the electricitysystem, from demand-side devices to wide-scale distributed generation to avarietyof energy markets. However, the most digitized, sophisticated grid in the world will not bereally smart if it does not have the capabilityto put renewable energy online, and make the system more efficient, more reliable and more flexible. Smart grid technologies are presently undergoing rapid developmentin an effort to modernize legacy power grids to copewith increasing energy demands of the future (Amin and Wollenberg [2005]). High speed bi-directional communications networks will provide the framework for real time monitoring and control of transmission, distribution and end-user consumer assets for effectivecoordination and usage of available energy resources. Furthermore, integration of automation into all levels of power network operations enables smart grids to rapidly self regulate and heal, improvesystem reliabilityand security,and more efficiently manage energy delivery and consumption (Garrity [2009]). With the integration of advanced control algorithms, the smart grid concept will allowto optimize the electrical system bymeans of the integration of distributed power generation, storage and consumption, absorbing agreater amountof renewable generation while keeping or even improving the currentqualitystandards. Regarding such an integration, this work deals with solar thermal power integration, whichhas agreat interest due to its potential contribution to the energy mix in the following years, and also due to its massivestorage capabilities. The paper is organized as follows: section 2presents a brief introduction to solar thermal technology; section 3is related to smart grid management, describing a general modeling and optimization framework in order to fulfil optimal energy grids management; section 4presents acase study based in the aforementioned optimization framework, showing some interesting aspects related to solar thermal power plants integration in energy networks; finally,section 5is dedicated to the concluding remarks. 2. SOLAR THERMAL PLANTS The most abundant, sustainable source of energy is the Sun, whichprovides over 150,000 terawatts of power to the Earth; about half of that energy reaches the Earth surface while the other half gets reflected to outer space bythe atmosphere. Only asmall fraction of the available solar energy reaching the Earth surface would beenough to satisfy the global expected energy demand. Although most renewable energies derivetheir energy from the sun, bysolar energy werefer to the direct use of solar Proceedings of the 18th World Congress The International Federation of Automatic Control Milano (Italy) August 28 - September 2, 2011 978-3-902661-93-7/11/$20.00 © 2011 IFAC 4939 10.3182/20110828-6-IT-1002.02791 Fig. 1. DISS Parabolic Trough Solar Colector at the PSA Fig. 2. Abengoa PS-10 Solar Power Tower radiation. One of the greatest scientific and technological opportunities weare facing is to develop efficientways to collect, convert, store, and utilize solar energy at affordable costs. There are twomain drawbacks of solar energy systems: a) the resulting energy costs are not yet competitiveand b) solar energy is not always available when needed. Considerable researchefforts are being devoted to techniques whichmayhelp to overcome these drawbacks, control is one of those techniques. While in other power generating processes, the main source of energy (the fuel) can bemanipulated as it is used as the main control variable, in solar energy systems, the main source of power whichis solar radiation cannot be manipulated (Camacho et al. [1997]) and furthermore it changes in aseasonal and on adaily base acting as a disturbance when considering it from acontrol pointof view. Concentrating solar thermal (CST) systems use optical devices (usually mirrors) and sun tracking systems to concentrate alarge area of sunlightinto asmaller receiving area. The concentrated solar energy is then used as aheat source for aconventional power plant. Awide range of concentrating technologies exist. The main concentrating concepts are: a) parabolic troughs (see Fig. 1), b) solar dishes, c) linear Fresnels, and d) solar power towers (see Fig. 2). The main purpose of the concentrating solar energy is to produce high temperatures and therefore high thermodynamic efficiencies. Parabolic trough systems are the most used CSP technology.Aparabolic trough consists of alinear parabolic mirror that reflects and concentrates the received solar energy onto atube(receiver) positioned along the focal line. The heat transfer fluid is pumped trough the receiver tube and picks up the heat transferred trough the receiver tube walls. The parabolic mirror follows the Sun bytracking along asingle axis. Regarding the problems that arise due to the the intermittency and unpredictabilityfluctuations associated to renewable energy resources, massiveenergy storage should betaken into account. Some technologies related to electricitystorage are under intense research: batteries, water pumping, super–capacitors, compressed air, fly wheels, superconducting magnetic energy storages, etc. Among of the most promising of them are those based on hydrogen production and utilization, but none of the aforementioned storage technologies are currently at acommercial stage. Nonetheless, some modern thermal power plants are incorporating technologies that store energy in athermal reservoir for alater reuse, whichcan beemployed to balance energy demand between daytime and nighttime. Abovesome interesting emerging technologies, molten salt has been proposed as ameans to retain ahigh temperature thermal storage for later use in electricitygeneration. Suchthermal reservoirs can store relatively high amounts of energy in periods whichcan reach6-8 hours. Storing energy in photovoltaic plants is more costly and difficult. Batteries and Hydrogen fuel cells havebeen used for this purpose in some experimental installations but there are no photovoltaic plants with asignificantstoring capabilities. 3. SMARTGRID MANAGEMENT 3.1 Grid Mathematical Formulation Hybrid energy hubs (see Fig. 3) are defined as interfaces between energy producers, consumers and the transportation infrastructure, introducing ageneral steady– state modeling and optimization framework for energy systems including multiple energy carriers and describing steady–state power flowcouplings between differentenergy infrastructures and/or network participants. In particular, hybrid energy hub modeling framework enables integration of an arbitrary number of energy carriers and anytechnology for transmission, conversion, and storage of energy can beconsidered. Moreover, such ageneral formulation ensures high flexibilityin terms of modeling detail and accuracy,where more approximate flowmodels can beused as well as detailed steady–state power flowequations. As discussed above, the core of the formulation presented herein is composed of three basic features: •inputs and outputs, •conversion, and •storage. Asingle converter unit can beseen as shown in Fig. 4. Suchan elementconverts, at time instantk,ageneric r input flowuL i,r(k)of ageneric hub ibelonging to anetwork Ninto ageneric poutput flowyi,p(k), where superscript L is associated to hub variables related to converters. Input– 18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011 4940 (a) (b) external network external network network hub branch inputsoutputsnetwork hub Fig. 3. (a) network hub sketchand (b) network composed of an interconnection of network hubs (a) (b) γ i,p-r Li,p yL ui,1 L ui,r L ui,n L i,1 y i,p y i,n y rp ui,r L Fig. 4. (a) Flowconverter with single input and output and (b) converter cluster in anetwork hub output conversion is defined through the so–called coupling factors γL i,p−r,whichcorrespond to converter’s steady– state conversion efficiency between input and output flows: yi,p(k)=γL i,p−ruL i,r(k).(1) Aconverter set can beexpressed through the so–called converter coupling matrix ΓL i,whichdescribes the mapping of the flows from the input to the output of ahub and is composed of aset of converter coupling factors γL i,p−r:    yi,1(k) . . . yi,np(k)    |{z } yi(k) =   γL i,1−1. . . γL i,1−nr . . ..... . . γL i,np−1. . . γL i,np−nr    |{z } ΓL i    uL i,1(k) . . . uL i,nr(k)    |{z } uL i .(2) For their part, storage devices in hubs are composed of an interface and an internal storage, as shown in Fig. 5. Specifically,the interface can beseen as aflowconverter, whichmodulates ageneric sstorage interface input flow uE i,s(k)into another generic storage interface output flow ˘uE i,s(k). Suchan output flowcarrier is then stored in an internal ideal storage. Mathematically,the storage interface is modeled analogously to aconverter device, with steady–state input and output flowvalues being related to eachother through the relation: ˘uE i,s =ei,s(k)uE i,s(k),(3) ui,s E ui,s E ei,s xi,s Fig. 5. Storage in network hubs ei,s(k)being the efficiency of the charge/discharge interface sof hub ibelonging to network N,whichdescribes howmuchof the flowexchanged with the system affects the storage. Suchafactor depends on the direction of the exchanged flowor, in other words, the storage can be charged or discharged: ei,s(k)=½e+ i,s if uE i,s(k)≥0(charging/standby) 1/e− i,s else (discharging),(4) with e+ i,s and e− i,s being the charging and discharging efficiency,respectively.Notice that, for the sakeof simplicity, storage performance is assumed to beconstant. Following the formulation, and from adiscrete–time point of view, internal storage state xi,s at instantk+1depends on the state at previous instantkand on the total exchanged flow˘uE i,s(k)during the period∆Tbetween k and k+1, assuming ˘uE i,s(k)to remain constantduring ∆T: xi,s(k+1) =xi,s(k)+ k+1 Z k ˘uE i,s(t)dt = xi,s(k)+˘uE i,s(k)∆T.(5) Thus, vector xi(k+1) containing all the storage states xi,s(k+1) ∈{xi,1(k+1), . . . , xi,ns(k+1)}depends on the state vector at previous time step xi(k), on matrix ΛE containing all the interface efficiencies ei,s(k)∈{ei,1(k), . . . , ei,ns(k)}and on vector uE icontaining all the interface flowinputs ui,s(k)E∈{ui,1(k)E, . . . , ui,ns(k)E},nsbeing the total number of storage elements in ahub:   xi,1(k+1) . . . xi,ns(k+1)   |{z } xi(k+1) =  xi,1(k) . . . xi,ns(k)   |{z } xi(k) +   ei,1(k) ... ei,ns(k)   |{z } ΛE i(k)   uE i,s(k) . . . uE i,ns(k)   |{z } uE i(k) ,(6) also taking into accountthe charging/discharging values of the interface efficiencies, with ΛE i(k)=©ΛE+ i,ΛE− iª=   {e+ i,1,1/e− i,1} ... {e+ i,ns,e− i,ns}   (7) 18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011 4941 ui,q E ui,q E ei,q xi,q γ L i,p-r ui,m E ui,m E ei,m xi,m ui,r Lui,r Lyi,pyi,p Fig. 6. Multiple storage elements in anetwork hub As storage elements maybeconnected to both hub inputs and outputs (see Fig.6), amathematical transformation of the corresponding storage flowto either side of the hub has to beconsidered, in order to obtain an input– output description independently where the storage device is connected physically. As seen in Fig. 6, the flow˘uL i,r(k)entering the converter equals the hub input flowuL i,r(k)minus the left side storage flowuE i,q(k): ˘uL i,r(k)=uL i,r(k)−uE i,q(k).(8) Analogously,the output converter side can bedescribed as: ˘yi,p(k)=yi,p(k)+uE i,m(k),(9) where ˘yL i,p(k)is the flowleaving the converter, yi,p is the hub output flowand uE i,m(k)is the rightside storage flow. Expressing flowtransformations as in (1) yields to: ˘yi,p(k)=γL i,p−r˘uL i,r(k).(10) Substituting (8) and (9) in (10), both converter sides storage flows can beexplicitly included in the hub input– output description: yi,p(k)=γL i,p−r³uL i,r(k)−uE i,q(k)´−uE i,m(k)= γL i,p−ruL i,r(k)+γE i,p−ruE i,r(k),(11) where γE i,p−rare the coupling factors related to storage variables uE i,r(k). Following the discussion, vector yi(k) containing all hub outputs yi,p(k)∈{yi,1(k), . . . , yi,np(k)} can beextended in order to include storage elements through the so-called storage coupling matrix ΓE i.Sucha matrix describes howchanges of the storage flows affect the converter output flows, resulting the following equation:   yi,1 . . . yi,np   |{z } yi =   γL i,1−1. . . γL i,1−nr . . ..... . . γL i,np−1. . . γL np−nr    |{z } ΓL i   uL i,1(k) . . . uL i,nr   |{z } uL i +    γE i,1−1. . . γE i,1−ns . . ..... . . γE i,np−1. . . γE np−ns    |{z } ΓE i   uE i,1(k) . . . uE i,ns   |{z } uE i .(12) Summarizing equations (6) and (12) yields to acomplete state–state representation of ageneric hub i∈N: i j yi uj wout,i j= win,j i wout,j i= win,i j yj ui Fig. 7. Interconnections between hubs iand j xi(k+1) =xi(k)+ΛEuE i(k) yi(k)=ΓL i·uL i(k)+ΓE i·uE i(k), whichcan becondensed bydefining the complete converter interface efficiency matrix Λi,the complete coupling matrix Γiand the complete hub input vector ui(k)as follows: Λi(k)=£0ΛE i(k)¤ Γi=£ΓL iΓE i¤ ui=£uL i(k)TuE i(k)T¤T, whichresults in the next condensed state space representation: xi(k+1) =xi(k)+Λi(k)ui(k)(13) yi(k)=Γiui(k).(14) From this formulation it is clear that ageneric network hub ican befully described bythe set of matrices Hi={Λi(k),Γi},whichis the main advantage of the formulation proposed herein, as it offers asimple method to model arbitrarily complex systems. Regarding acomplete energy grid, let’s consider ageneric network composed of aset of N={1, . . . , nN}interconnected hubs (see Fig. 7), nNbeing the total number of hubs composing the network. The dynamics of subnetwork i∈Nis defined bythe following nonlinear time–invariant discrete–time state space model: xi(k+1) =xi(k)+Λiui(k) yi(k)=Γiui(k)+Πin,i win,i(k), Considering the interconnections among hubs, generic interconnecting variables between hub iand its neighbor jcan bedefined as seen in Fig. 7, where win,i←j(k) and wout,i→j(k)are, respectively,generic interconnecting input and output vectors of hub irelated to its neighboring hub j,with: wout,i→j(k)=Πout,i→jyi(k), where Πout,i→jis the output interconnecting matrix referred to the coupling of hub iwith respect to its neighbor j. 18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011 4942 3.2 Grid DistributedManagement AModel PredictiveControl (MPC), (Camacho and Bordons [2004])Due to the complexityof future energy networks, with an increasing number of consumers and producers, adistributed control effort is taken into account in this work. Specifically,aLagrange-Based Distributed Model PredictiveControl (Lag-MPC) formulation (Negenborn [2007]) has been chosen as the control framework utilized in order to solvethe network problem. The objective of suchaproblem is to minimize the total cost of supplying energy to satisfy the demand. Tothat end, eachenergy producer has assigned acost per power unit produced while, at the same time, eachconsumer has assigned an energy demand. The next algorithm implements the Lag-MPC discussed herein, whichwould haveto bedone, at eachtime step k,byeachlocal agenti∈Nin aparallel computation scheme: (1) Makeameasurementof currentstate b xi(k) (2) Compute the optimal control sequence e u∗ i(k). Todo so, perform the following steps: (a) Parameter initialization: p=1;ei≫1;λ1 in,i→j(k)=0;λ1 out,i→j(k)=0 (b) Solvethe following optimization problem: min exi(k+1),eui(k),eyi(k) ewin,i(k),ewout,i(k) φlocal,i³exi(k+1),eui(k),eyi(k)´ +X j∈Ni φinter,i³ewin,i(k),ewout,i(k)´, subject to subnetwork dynamics, constraints and initial condition. (c) Send e wp in,i→j(k)and e wp out,i→j(k)to neighboring agents j∈Niand collect e wp in,j→i(k)and e wp out,j→i(k)from them. In other words, you (an agenti)tells to your neighbors j∈Niwhat you would like to do e wp out,i→j(k)and what you would like them to do e wp in,i→j(k). Atthe same time, you receivewhat your neighbors want to do e wp out,j→i(k)and also what your neighbors want you to do e wp in,j→i(k). (d) Upgrade the Lagrange multipliers e λp+1 in,i→j(k)=e λp in,i→j(k)+ γc³e wp in,i→j(k)−e wp out,j→i(k)´ e λp+1 out,i→j(k)=e λp out,i→j(k)+ γc³e wp out,i→j(k)−e wp in,j→i(k)´. (e) Evaluate the stopping conditions: •p>p •ei=° °e λp+1 in,i→j(k)−e λp in,i→j(k)° °≤e, where pis the maximum number of iterations allowed and eis the maximum error allowed. If one of the stopping conditions is true, go to step (3). If not, moveon to the next iteration p←p+1 and return to step (b). external network external network Fig. 8. Energy network case study 0 5 10 15 20 25 0 5 10 15 20 25 30 35 40 Time (hours) Available solar power (p.u.) Fig. 9. Available solar power (p.u.) (3) Implementthe optimal control action u∗ i(k) (4) Start anew control cycle k←k+1and go to step (1) 4. CASE STUDY In this section, anetwork case study is studied in order to showan example of asolar thermal plantintegration in asmart grid. Specifically,the energy network shown in Fig. 8was simulated. Suchanetwork is composed of three interconnected citycenters, an industrial center, two wind farms, twothermal fuel oil plants and asolar thermal plant. The aforementioned network case study was simulated in MATLAB, also utilizing CPLEX solver. Suchanetwork was firstly divided into three sectors, thus considering three interconnected network hubs. Each hub was assigned to aLag-MPC control agent, and the distributed control algorithm described in the preceding section was performed. A24-step simulation time was considered, thus dividing acomplete dayinto 24 hours. Regarding the thermal solar plantand due to the solar irradiation data taken herein into account, total solar power shown in Fig. 9corresponding to asunnydayand expressed in power units (p.u.) was available. As can beseen in Fig. 8, the thermal solar plantis directly connected to acitycenter, whichis also feed by the twofuel oil power plants. In order to simulate the network, eachenergy producer offers acertain energy to the consumers eachtime step. Atthe same time, each 18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011 4943 Fig. 10. Power demand energy mix 0 5 10 15 20 25 0 10 20 30 40 Time (hours) Stored energy (e.u.) 0 5 10 15 20 25 −10 −5 0 5 10 15 20 Time (hours) Storage power input (+) / output (−) (p.u.) Fig. 11. Stored energy (e.u.) (upper graph) and storage power input and output (p.u.) (lower graph) consumer minimizes their total demanded energy cost by selecting the best offers eachtime step. In this example, investmentand operational costs of the thermal solar plantare supposed to besupported byfeed in tariffs and only fuel costs are considered. Since the thermal power plant“rawmaterial” is solar irradiation, suchaplantcan offer its energy production at acost of 0monetary units (m.u.) per power unit (p.u.) produced. In turn, the fuel oil power plants offer, respectively,power at acost of 2.5 and 3m.u. per p.u. As aresult, all solar power offered is boughtbythe citycenter. In order to showthe impact of energy storage in solar thermal plants, aenergy storage device (suchas molten salt) was firstly taken into account. Fig. 10 shows the resulting power demand mix of the citycenter placed next to the solar thermal plant. As can beseen, first nine steps are dominated byfuel oil power, as no power from sun is available. Contrarily,sun power is mainly demanded the following eighthours. Atthe same time, as can beseen in Fig. 11, sun power excess is stored in the thermal solar plantstorage device, resulting in an increase in the energy stored (Fig. 11 upper graph). Attime step 20, no more solar irradiation is available. However, as there is asensible energy amountstored in the molten salt device, the solar thermal power plantis able to continue injecting power to the grid. As aresult, power from the storage device is also included in the energy mix corresponding to the last part of the simulation. Finally,Fig. 12 shows acomparison between the described scenario with asolar thermal plantincluding storage 0 5 10 15 20 25 0 50 100 150 200 250 300 350 400 Time (hours) Accumulated cost (m.u.) without molten salt storage with molten salt storage Fig. 12. Interconnections between hubs iand j capabilities and without it. As can beseen, accumulated costs are notably higher without considering storage. Such acost incrementis due to the fact that the excess of sun power collected during the central daily time is lost when no storage capabilities are considered. The resulting energy mix corresponding to the last part of the daywould bethen mainly composed of costly fuel oil power, which is the reason of the theoretical accumulated energy cost increment. 5. CONCLUSION This paper deals with the optimal integration of solar power plants in grids. The paper proposed amodification of energy hubs whichallows to solvethe optimization problem encountered when integrating solar power plants in electrical grids with amixed integer programming algorithm in adistributed way.An introductory simulation study case was presented. ACKNOWLEDGEMENTS The author would liketo acknowledge the Spanish Ministry of Education, Junta de Andalucia and the European Commission for funding this project under grants DPI2008-05818, P07-TEP-02720 and HD-MPC. REFERENCES S. Massoud Amin and B. F. Wollenberg. Toward asmart grid: power delivery for the 21st century. IEEE Power Energy Mag,3(5):34–41, 2005. E.F. Camacho and C. Bordons. Model Predictive Control. Springer Verlag, London, 2004. E.F. Camacho, M. Berenguel, and F.R. Rubio. Advanced Control of Solar Power Plants.Springer Verlag, London, 1997. T. F. Garrity.Innovation and trends for future electric power systems. In Proc. PSC’09. Power Systems Conference,pages 1–8, 2009. M. Houwing, A. N. Ajah, P.M. Herder, and I. Bouwmans. Addressing uncertainties in the design and operation of residential distributed energy resources: Case study of a micro-CHP system. In 10th Conferenceon Process Integration, Modelling and Optimisation for Energy Saving and Pollution Reduction,Ischia Island, Italy,June 2007. R.R. Negenborn. Multi–Agent Model Predictive Control with Applications to Power Networks.PhD thesis, Technische Universiteit Delft, 2007. 18th IFAC World Congress (IFAC'11) Milano (Italy) August 28 - September 2, 2011 4944