Swarm Grid: Collective synchronization of electricity grid devices

Transcription

1 Swarm Grid: Collecive synchronizaion of elecriciy grid devices TR/ROBOLABO/ Developed by ROBOLABO Auhors: Manuel Casillo-Cagigal Eduardo Maallanas Álvaro Guiérrez Las updae: April 11, 2014

2 Swarm Grid: Collecive synchronizaion of elecriciy grid devices Manuel Casillo-Cagigal, Eduardo Maallanas, Álvaro Guiérrez ETSIT, Universidad Poliécnica de Madrid, Madrid, Spain Absrac The elecriciy grid has o respond o large power variaions in relaively shor periods of ime. This behavior causes an inefficiency operaion and design of he elecriciy sysem. The Demand Side-Managemen echniques are used o modify he power demand, reducing or smoohing is shape. Moreover, he elecriciy grid is a very huge sysem composed for millions of devices. There are problems associaed wih he densiy and disance beween elemens. The developmen of new echnologies has increased he complexiy of he grid in he las years. In his framework, he Smar Grid arises. The inclusion of communicaion echnologies ino he grid allows a beer monioring and conrol of he sysem. Smar Grid sysems mus be able o manage housands or even millions of devices o increase he efficiency of he elecriciy grid, bu hey have also o saisfy he classical grid consricions. The amoun of informaion and calculus of he Smar Grid managemen sysem should be reduced. For hese reasons, our sudies abou he elecriciy grid synchronizaion are based on a disribued design. Specifically, hey are focus on he Swarm Inelligence heory. The Swarm Inelligence heory is inspired on biological sysems, concreely on he self-organized behavior of several species. The elemens or agens follow very simple rules sensing he local environmen and wih a cerain random degree. For hese reasons, hese algorihms are good candidaes o solve he problem of he elecriciy grid synchronizaion, creaing wha we have called he Swarm Grid. The aggregaed consumpion of he grid is a ime varian funcion wih peaks and valleys. The elecriciy grid generaion has o follow his funcion in order o supply he power demand. An example of he ypical aggregaed consumpion for a weekly day in Spain is shown in Figure 1. Therefore, he elecriciy grid has o respond o large power variaions in relaively shor periods of ime. This behavior causes an inefficiency operaion and design of he elecriciy sysem. On he operaion side, here is a poor performance because of generaion power swings: generaion mus always supply power demand, herefore abrup changes in he demand cause generaion power swings, ogeher wih increases in operaion sress and losses. On he design side, he elecriciy grid mus be sized o supply he peak power demand: he power difference beween valleys and peaks implies ha he elecriciy grid mus be over-sized. Several power plans and lines are designed o supply he power demand during he peak hours and hey are no used during he res of he day. One echnical soluion o minimize hese problems is o flaen he power demand curve, reducing he difference beween peaks and valleys. The echniques used o modify he power 1

3 Figure 1: Aggregaed consumpion of a ypical weekly day in Spain. Source: Red Elécrica Española. demand are called Demand-Side Managemen (DSM) echniques. They are used according o four differen goals: i) o reduce he power consumpion, ii) o reduce he power consumpion during peak hours, iii) o increase he power consumpion during valley hours or iv)o displace he consumpion from peak o valley hours. These echniques are commonly focused on he house or indusrial common consumpion managemen (e.g. lighing, compuers, washing machines). Nowadays, given recen echnological advances in decenralized generaion and sorage, DSM should also include new local generaion and sorage echnologies like phoovolaic energy, lead-acid/ni-cd baeries and elecrical vehicles. In previous work carried ou by he Renewable Disribued Generaion and Inelligen Conrol research group, he developmen of DSM echniques has been focused on he operaion of a single insallaion in he GeDELOS-PV sysem framework. GeDELOS-PV sysem is a real example of he added-value of PV elecriciy ha combines he las generaion of PV gridconneced sysems, small-scale elecriciy sorage, DSM sraegies and a complee monioring and acuaion sysem [1]. Is main objecive is o saisfy he final user elecriciy needs while opimizing local energy self-consumpion, hrough he combinaion of wo complemenary sraegies focused on he load profile (daily paern can be modified o beer mach he expeced PV generaion paern) and local sorage managemen (only surplus of generaed elecriciy is sored for laer use in hours of low/zero generaion) [2, 3, 4]. Through GeDELOS- PV sysem he ineracion wih he elecriciy grid is reduced, saisfying one of he DSM goals. Despie he energy exchanges beween he elecriciy grid and a single insallaion are reduced, i is no synchronized wih he res of he elemens of he grid. Hence, a sysem o coordinae he whole grid is necessary if he advanages arising from he combinaion of local generaion, sorage and load demand are o be exended o muliple users. The elecriciy grid is a very huge sysem composed for millions of devices. There are problems associaed wih he disance beween elemens, e.g.: a power plan has o supply several devices ha are hundreds of kilomeers away. Oher problems are associaed wih he disri- 2

4 Figure 2: Smar Grid managed by a swarm inelligence algorihm concepual scheme. buion of he consuming elemens, having areas wih very high or low densiy. Moreover, he developmen of new echnologies increases he complexiy of he grid e.g.: disribued generaion, local sorage sysems, elecrical vehicles. Wih he aim of solving curren elecriciy grid managemen problems and gradually inroducing susainabiliy and raional use of resources ino elecriciy neworks, he Smar Grid arises. Smar Grid sysems mus be able o manage housands or even millions of devices o increase he efficiency of he elecriciy grid, bu hey have also o saisfy he classical grid consricions. The elecriciy grid mus be robus because he producion and he way of a whole counry depend on i. This implies ha a Smar Grid managemen should no be cenralized in a single locaion: he sysem mus respond quickly o generaion or demand variaions because he elecriciy supply mus always be saisfied. Therefore, he amoun of informaion and calculus of he Smar Grid managemen sysem mus be reduced. For hese reasons, our sudies abou he elecriciy grid synchronizaion are based on a disribued design. Specifically, hey are focus on he Swarm Inelligence heory. The Swarm Inelligence heory is inspired on biological sysems, concreely on he self-organized behavior of several species. The elemens or agens follow very simple rules sensing he local environmen and wih a cerain random degree. Alhough here is no cenralized conrol srucure dicaing how individual agens should behave, ineracions beween such agens lead o he emergence of an inelligen global behavior, unknown o he individual agens. When hese algorihms are applied o engineering problems, hey achieve global soluions wih high robusness and a low informaion exchange. For hese reasons, hese algorihms are good candidaes o solve he problem of he elecriciy grid synchronizaion, creaing wha we have called he Swarm Grid (see Figure 2). 3

5 P 34GW P 30GW 14GW 7GW P (a) P (b) 27GW 25GW 20GW 24GW (c) (d) Figure 3: Operaion example of he swarm inelligence algorihm for Smar Grid managemen: a) wihou conrollable consumpion, b) 25% of conrollable power consumpion, c) 50% of conrollable power consumpion and d) 100% of conrollable power consumpion. We consider he devices conneced o he grid like elemens of a huge swarm. These elemens are able o sense informaion abou he elecriciy grid saus, in he same way ha insecs of a colony sense informaion from he environmen. Every elemen modifies sochasically is own consumpion and aking ino accoun he sensed informaion, follows a swarm inelligence algorihm. This algorihm indicaes how each device conneced o he grid has o demand power hrough a probabiliy disribuion. We have developed his algorihm and we have applied i ino an elecriciy grid simulaor. Our experimens are focused in he residenial secor, where several users demand power from he elecriciy grid acivaing differen elecrical appliances. When a user acivaes a device ha can be conrolled by he Swarm Inelligence algorihm, i decides when i mus be sared according wih he elecriciy grid saus. An example of he swarm inelligence algorihm operaion can be observed in Figure 3. We have esed how he increase of he power amoun conrolled by he algorihm affecs he grid. Firsly, he grid operaion wihou conrollable 4

6 REFERENCES REFERENCES power is shown in Figure 3a (he acual siuaion). Secondly, a 25% of conrollable power has been inroduced in he elecriciy grid and he resul is shown in Figure 3b: he difference beween peaks and valleys has been considerably reduced (41% in relaion o reference case a) by he use of he swarm inelligence algorihm. Finally, he amoun of conrollable power has been increased o 50% and 100% and he resuls are shown in figures 3c and 3d respecively (reducions of peak-valley consumpion of 74% and 96% respecively, relaive o reference case a). Acknowledgemens E. Maallanas is sponsored by he Spanish Minisry of Educaion wih a Ph.D gran (FPU- 2011). M. Casillo-Cagigal is sponsored by he Spanish Minisry of Educaion wih a Ph.D gran (FPU-2010). References [1] M. Casillo-Cagigal, E. Maallanas, A. Guiérrez, F. Monaserio-Huelin, E. Caamaño- Marín, D. Masa-Boe, and J. Jiménez-Leube, Heerogeneous collaboraive sensor nework for elecrical managemen of an auomaed house wih pv energy, Sensors, vol. 11, no. 12, pp , [2] M. Casillo-Cagigal, A. Guiérrez, F. Monaserio-Huelin, E. Caamaño-Marín, D. Masa- Boe, and J. Jiménez-Leube, A semi-disribued elecric demand-side managemen sysem wih PV generaion for self-consumpion enhancemen, Energy Conversion and Managemen, vol. 52, no. 7, pp , [3] M. Casillo-Cagigal, E. Caamaño-Marín, E. Maallanas, D. Masa-Boe, A. Guiérrez, F. Monaserio-Huelin, and J. Jiménez-Leube, Pv self-consumpion opimizaion wih sorage and acive dsm for he residenial secor, Solar Energy, vol. 85, no. 9, pp , [4] E. Maallanas, M. Casillo-Cagigal, A. Guiérrez, F. Monaserio-Huelin, E. Caamaño- Marín, D. Masa, and J. Jiménez-Leube, Neural nework conroller for acive demandside managemen wih pv energy in he residenial secor, Applied Energy, vol. 91, no. 1, pp ,