CIRED Workshop - Rome, June 2014 Paper 0323

Transcription

1 FUTURE LOW CARBON TECHNOLOGIES, IMPACTS AND ENERGY STORAGE SOLUTIONS ON UK DISTRIBUTION NETWORKS Oghenetejr ANUTA Chrstan BARTECZKO-HIBBERT Neal WADE Newcastle Unversty UK Durham Unversty UK Newcastle Unversty UK ABSTRACT Investment and uptake of low carbon technologes s ncreasng due to government polces that are set to decarbonse the electrcty grd n the UK Understandng future demand and generaton n dstrbuton networks where these low carbon technologes wll preval s mportant n plannng and managng the networks. Ths paper assesses the mpacts of future demand and generaton n the form of heat pumps and solar photovoltacs, whch could be evenly dspersed or locally concentrated on a medum voltage dstrbuton network. The effectveness of usng energy storage to manage the resultng ssues arsng from such networks s nvestgated. INTRODUCTION Low carbon technologes (LCT) are projected to grow n the UK based on government targets and polces to enable 15% of demand to be produced from renewable sources by 2020 and to decarbonze the grd [1]. A report by UKERC dscusses ssues wth the evolvng UK power sector and dfferent scenaros to handle such ssues between now and 2050 [2]. Among the ssues are problems n the dstrbuton network caused by LCTs such as heat pumps (HPs) and solar photovoltacs (PV). Furthermore, Ref. [2] dscusses the change n the mndset of Dstrbuton Network Operators (DNOs) n the UK towards the assumptons that LCTs would be evenly dstrbuted. As DNOs do not have control over the dstrbuton of LCTs, t s concevable that hgh concentratons would be nstalled n areas that are already overstretched, thus leadng to bottlenecks n the dstrbuton networks. Such developments would requre a more actve mode of managng the networks as demand and generaton becomes more varable and bdrectonal power flows occur. LCTs provde benefts such as reduced losses and ncreased relablty n the transmsson and dstrbuton network dscussed n Ref. [3]. However, they could also lead to ssues locally on the network whch can lmt LCT uptake, such as voltage rse, thermal overloadng, and reverse power flows [3-5]. DNOs wll have to move from a passve, demand drven control and nvestment approach whch was sutable n the past, to a more actve and nnovatve approach nvolvng new technologes other than the tradtonal nvestment n wres and transformers. Technologes such as demand sde response and energy storage are seen as possble solutons to the future challenges caused by LCTs [6]. LOW CARBON TECHNOLOGIES AND ENERGY STORAGE Assessng low carbon technology mpacts on the dstrbuton network It was dscussed n [2] that network operators of transmsson and dstrbuton systems n the UK lack an understandng of the changes they would need to mplement beyond the year Future plannng and operaton wll be demand and generaton drven. The varable locatons and operatng patterns of demand and generaton LCTs brngs about uncertanty. For example solar PV generaton output s dctated by the vagares of the weather. Solar PV s expected to make up 2% of the 15% target for renewables and HPs are expected to be nstalled n 25% of domestc households by 2030 [7, 8]. DNOs currently employ determnstc load flow methods n evaluatng, plannng and operatng ther networks. Ths however dsregards the uncertantes or stochastc devatons that customers wth LCTs would present on the networks. To enable a more accurate representaton of the ssues that may occur on a dstrbuton network, the authors employ a probablstc load flow (PLF) approach. Furthermore, an assessment s carred out on the effect of varyng concentratons of these LCT customers on the network to understand the ssues and severty as concentratons vary. Energy storage solutons Energy storage systems mplemented n the dstrbuton networks can be used to manage and allevate the mpacts of LCTs. The applcatons of ESS nclude, voltage control, to manage overvoltage and undervoltage; power flow management to reduce losses, reverse power flows and thermal overloadng of overhead lnes, cables and transformers [9]. Ths paper mplements ESS to mtgate LCT mpacts on a representatve medum voltage (MV) dstrbuton network. SIMULATION METHODOLOGY The load and generaton profles used n ths study comprse domestc customers, HP loads and PV generaton. The smulated profles were created from anonymsed data gathered and processed under the Customer Led Network Revoluton (CLNR)project n the UK [10]. CLNR s conductng a seres of montorng trals usng over 9000 smart meters n resdental, ndustral and commercal locatons wthn the UK to Paper No 0323 Page 1 / 5

2 understand current and emergng load and generaton profles. The CLNR smart meter dataset s classfed by dfferent categorcal varables from household ncome to ruralty. Demand and generaton profles CLNR domestc metered data Wnter and summer data was consdered due to the extremes both seasons provde for generaton wth PV and demand wth HPs. A clusterng approach was used to determne the number of natural customer groups that exst n the CLNR dataset. K-means s an unsupervsed learnng algorthm that parttons data nto k-clusters. Clusterng was performed on over 7200 domestc customers. In the wnter month of January, 3 clusters dentfyng dstnct demand groups were shown to exst based on ther magntude of energy usage. Ths clusterng approach was repeated for the summer month of July based on the assumpton that customer types would not dffer. Each cluster n effect assgned probabltes to a gven customer type. For January, the proporton of customers wth the hghest demand was attrbutable to 7% of the total sample, 36% to the medum cluster and 57% to lower demand customers. All domestc customers on a partcular LV network were consdered to be of the same type and thus homogeneous. The domestc customer base for each LV network was derved through a weghted approach based on the aforementoned customer types and the avalable number of customers on that LV network. SOLAR and HP As part of the CLNR database two dstnct LCT groups of HPs and PV exst. For both HP and PV, data s converted to half hourly real power readngs. Because no nformaton of the property type and HP characterstcs are known a Monte Carlo method was used wth the customer s 48 half hourly profle. Ths analyss uses weekday data as the random varable. To ncrease the sample selecton, all customers n the database for that month wth a full weekday demand profle were added to the sample set. Ths was deemed approprate as HP operaton on a day by day bass were consdered as ndependent events. To enrch the sample set of PV generaton each customer s monthly maxmum power consumpton was computed. Taken over a 4 month perod from May to September there was suffcent data for a representatve sample set consderng the dfferent lattudes and PV nclnatons. Test network descrpton The test network used was obtaned from [11]. A sngle lne representaton of the IEEE 33 bus radal dstrbuton test network s shown n Fgure 1. It was adapted to represent a UK MV dstrbuton network based on the followng assumptons: Followng relablty requrements on MV networks, the MV transformer power ratng was selected so 50% of ts capacty meets peak demand whle the lne ratngs were chosen so 75% meets peak demand on the network. The network was assumed to have only resdental customers and an after Dversty Maxmum Demand (ADMD) of 1.5 kw for resdental households was used to determne the number of customers on each busbar [12]. Grd Supply Pont Fgure 1: IEEE 33 bus test network Monte-Carlo applcaton To deal wth uncertanty a Monte Carlo approach was appled to demonstrate the feasble types of load and generaton a network could experence over a 24 hour perod network realsatons were deemed suffcent to encapsulate voltage varatons across the network as n Ref. [13]. When consderng LCTs, the mnmum number of customers allowable on the frst randomly selected LV network was the specfed penetraton level on the network, ths was necessary as a base to start populatng the network wth the requred LCT. Afterwards to ensure good accountng on successve realsatons, the mnmum penetraton was updated to reflect the networks status n terms of customer type added to the quantty stll requred. The random number selected was gven by where s the penetraton over bus wth technology ndcates the mnmum requred for that part of the LV network and unty s the maxmum possble. The smulaton procedure s descrbed n Fgure 3. Once the network s populated wth the customer types a load flow analyss was performed and the condton of the network assessed. The network constrants that were montored durng the random load flow smulatons were the voltage at ±6% of nomnal voltage, reverse power flow at 10% of the on-load tap changer (OLTC) ratng, and MV transformer and lne thermal capacty at a threshold of 50% and 75% of thermal ratng. A base case smulaton was carred out to establsh the default Paper No 0323 Page 2 / 5

3 condtons on the network. The case study was carred out usng half hourly wnter and summer demand and generaton data, whch represents the two extremes for the operaton of HPs and PV, wth HPs beng used more frequently n the wnter for heatng and PV reachng ts full generaton potental n the summer. Fgure 2 depcts a representatve HP demand on fve of the busbars on the IEEE test network. compensaton and when the lmts of the ESS are reached (.e. based on the converter ratngs), real power s then appled. Determne penetraton of customer types on the whole network HP consdered Populaton of customer types (HP) on each busbar determned by Pseudo random numbers t mn, t p p,1 YES Sum of HP customers on the network satsfed Populate the remanng busbar capacty or all busbars wth domestc customers PV consdered? NO Monte-Carlo method performed to generate the accumulated loads for each customer type on each busbar Perform N load flow smulatons YES PV penetraton updated usng: t mn,t p,1 p NO Compute worst case result Implement ESS on network Perform load flow and collate results Fgure 3: Load flow smulaton process Fgure 2: Representatve busbar loadngs for accumulated heat pump demand The followng steps were taken for populatng the busbars wth the requred PV, domestc and HP profles: Random populaton of varyng penetratons for all customer types across the network usng penetraton lmts of 25 and 50%; Usng the same penetraton lmts outlned above a lne approach was explored to populate the network wth HPs and PV. Lne was determned based on demand of LV network demand at each MV network node to produce the worst scenaro. For HP technology ths resulted n the lne wth the hghest demand and for PV the lne wth the lowest demand was selected. Energy storage plannng and operaton Two locatons were explored for the ESS, wth one locaton chosen at the network mdpont and another at the locaton of the busbar wth the hghest voltage excursons. The maxmum power ratng of the ESS, whch was connected to the MV sde at the secondary LV substaton, was constraned based on lne capacty. The ESS was used on the worst case demand and generaton scenaros. Ths was consdered based on the smulaton wth the maxmum amount of losses and hghest magntude and number of excursons outsde of the network lmts. Thresholds were set for ESS to be nvoked to resolve any power flow ssues and mantan the voltage wthn constrants. The ESS s operated to resolve or reduce overvoltage and undervoltage excursons by provdng a combnaton of real and reactve power; ths s requred because of the low X/R ratos n dstrbuton networks. However, preference n the control opton s gven to reactve power Power flows were managed by sourcng real power durng overpower and peak perods, or by snkng real power durng perods of hgh generaton that lead to reverse power flow n order to reduce thermal overloadng on the transformer and lnes on the network. Managng voltage and thermal constrants wll n turn lead to a reducton n real power losses. RESULTS AND DISCUSSION Over all the smulatons, there were only ssues wth voltage and reverse power flow as the network was robust enough to handle the ncrease n demand and generaton caused by HPs and PV. The most severe case n whch over voltage and under voltage occurred was at 50% penetraton levels n terms of customer numbers. Fgure 4 depcts the voltage extremes and system losses for HPs and PV respectvely compared to the base case. 25% penetraton of both HP and PV does not cause any voltage or thermal excursons. However, at a 50% penetraton ssues occurred on the network. Maxmum overvoltage for both PV cases occurred at 1pm. For HPs the perod wth the lowest voltage was consstent at 5pm. For the case of HPs, there were four undervoltage events and the network was operatng close to the lower voltage lmt over the 1000 smulatons for both cases. Lkewse, at 50% PV penetraton there was a hgher reverse power flow event for all smulatons and there were 5 cases of overvoltage for the case and 48 overvoltage events for the case wthout lne, ths s llustrated n Fgure 4. The hgher densty of overvoltages for the PV case s as a result of the room for wder nstallaton of PV at all remote ends of the network. On the contrary, for HPs the frequency of undervoltage for the scenaro showed a hgher frequency wth lower magntudes of undervoltage compared to that of the system wth. Paper No 0323 Page 3 / 5

4 Power Loss (MW) Voltage (P.U) Power Loss (MW) Voltage (P.U) CIRED Workshop - Rome, June 2014 as shown n Fgure Losses Maxmum Voltage Fgure 4: Densty plots of voltage percentage as a functon of nomnal voltage vs. real power of the network at 50% penetraton of HP (above) and PV (below). From the worst case results that were selected, there were 14 reverse power flow events reported for both cases of 50% PV penetraton wth and wthout lne. But the overvoltage devatons were extreme n the case wthout lne. For HPs, although there was one undervoltage event, there was a hgher amount of real power loss as the amount of HP was ncreased as shown n Fgure 6. The ESS was mplemented at two locatons n turn, one at the mdpont of the network and at the end of the feeder,.e. the most problematc busbar. The ESS operated at the end of the feeder was better at reducng the losses on the network. Results of loss reducton usng ESS are shown n Fgure 5 and Fgure 6. Fgure 5 and Fgure 6 shows the base case results (only domestc customers) alongsde the respectve PV and HP and domestc customer results for the two cases (25% and 50% penetraton levels). In each case the voltage and losses are computed for comparson purposes of the extreme events. The result from mplementng ESS to resolve network constrants on the dfferent cases s also depcted. The real power losses were reduced for the 25% cases when compared to the basecase but ncreasng the penetraton levels to 50% ncreased the losses. Ths s expected because of the hgh amounts of reverse power flow on the network. To curb the voltage excursons and mantan the network voltage wthn lmts, the ESS was used cooperatvely wth the OLTC. The operatng lmts for the ESS were thus set at 5% above and below the nomnal network voltage. In addton, the ESS was charged up durng perods of reverse power flow over a defned threshold of 0.75 MW. As a result of snkng reverse power on the network, losses were reduced below base case levels by 40% for the PV case and 32% for the case wthout lne 0 Base case max 25% PV & 25% PV & Fgure 5: Maxmum voltage and losses for PV base case, study cases and EES mplementaton Fgure 6 llustrates the ncrease n losses as the penetraton of HPs ncreases on the network. The losses ncrease even more as expected due to the feeder havng a hgher demand than respectve feeders on the network. The ESS was mplemented for the 50% HP case where there was a voltage excurson Base case mn 25% HP + Dom & 75% Dom no lne Losses 25% HP + Dom & 75% Dom Mnmum Voltage 50% HP + Dom 50 HP + Dom & & 50% Dom no 50% Dom wth lne lne 50 HP + Dom & 50% Dom wth lne Fgure 6: Maxmum voltage and losses for HP base case, study cases and EES mplementaton The ESS was also dscharged to reduce peak power flows on the feeder wth hgh concentraton of HPs wth a threshold set at 40% of the lne capacty ratng, ths reduced the amount of losses on the network by 36%. However, wth ESS effcency losses consdered assumng a round trp effcency of 90%, there was a net ncrease n losses of 5.5%. Ths negates the loss reducton benefts provded on the network. Fgure 7 shows the operaton of the ESS and the resultng drop n capacty utlsaton of the feeder wth the hghest utlsaton. A summary of the ESS power ratng and energy capacty used s shown n Table 1. ' ' Paper No 0323 Page 4 / 5

5 and Scottsh Power Energy Networks for fundng ther research on energy storage use n dstrbuton networks. Fgure 7: Peak power reducton on feeder wth hghest PV concentraton Table 1: Energy storage power ratng and energy capacty for cases Case Power Ratng Actons / Energy capacty 50% PV wth lne 1.2 MW/ 6 MWh Reverse power flow and 50% PV wthout lne 50% HP wth lne CONCLUSION 1.1 MW / 5.1 MWh 1.8 MW/ 11.9 MWh overvoltage Reverse power flow and overvoltage Overpower and undervoltage Hgher penetraton of HPs and PV can lead to voltage excursons, ncreased losses and reverse power flow. The authors have shown the mpacts of varable levels of penetraton and concentratons of HPs and PV on a network whlst demonstratng that the effectve use of ESS can be used n resolvng voltage ssues and reverse power flows that would be caused by these LCTs. Furthermore PV and HPs on the network lead to ncrease n losses, ths was worse for the cases wth hgh HP concentratons. Losses were reduced by reducng the peak power flows on the network usng ESS. When usng ESS, the net losses on a network could potentally ncrease because of the storage effcency losses, hence ESS operaton wll need to be optmsed to mnmse system losses durng daly ESS operaton. Wth large amounts of HPs and PV on the dstrbuton network, ESS could be a possble soluton outsde of tradtonal plannng methods to manage the problems that wll occur. ACKNOWLEDGEMENT CLNR s a collaboratve project between Northern Powergrd, Brtsh Gas, EA Technology and Durham and Newcastle Unverstes. The authors are grateful to project partners and to Electrcty North West Lmted REFERENCES [1] UK Department of Energy and Clmate Change, "Plannng our electrc future: a Whte Paper for secure, affordable and low carbon electrcty," ed, 2011, pp [2] N. Balta-Ozkan, Watson, T., Connor, P., Axon, C.,, L. Whtmarsh, Davdson, R., Spence, A., Baker, P., Xenas, D., Cpcgan, L. and Taylor,, and G., "Scenaros for the Development of Smart Grds n the UK - Synthess Report," UKERC, London2014. [3] R. Vral and D. K. Khatod, "Optmal plannng of dstrbuted generaton systems n dstrbuton system: A revew," Renewable and Sustanable Energy Revews, vol. 16, pp , [4] G. Pepermans, J. Dresen, D. Haeseldonckx, R. Belmans, and W. D haeseleer, "Dstrbuted generaton: defnton, benefts and ssues," Energy Polcy, vol. 33, pp , [5] A. A. Bayod-Rujula "Future development of the electrcty systems wth dstrbuted generaton," Journal of Energy, vol. 34 pp , [6] P. Warren, "A revew of demand-sde management polcy n the UK," Renewable and Sustanable Energy Revews, vol. 29, pp , 1// [7] Energy Research Partnershp (ERP), "The future role for energy storage n the UK," [8] Parlamentary Offce of Scence and Technology, "Solar Photovoltacs, POSTnote," ed: House of Parlament, [9] N. S. Wade, P. C. Taylor, P. D. Lang, and P. R. Jones, "Evaluatng the benefts of an electrcal energy storage system n a future smart grd," Energy Polcy, vol. 38, pp , [10] R. Wardle, Chrstan, Barteczko-Hbbert., Mller, Davd and Sdebotham, Lz., "Intal Load Profles form CLNR Interventon Trals," [11] M. E. Baran and F. F. Wu, "Network reconfguraton n dstrbuton systems for loss reducton and load balancng," Power Delvery, IEEE Transactons on, vol. 4, pp , [12] I. Rchardson, M. Thomson, D. Infeld, and C. Clfford, "Domestc electrcty use: A hghresoluton energy demand model," Energy and Buldngs, vol. 42, pp , 10// [13] M. Neameh, G. Hll, P. Blythe, R. Wardle, Y. Jalang, and P. Taylor, "Integratng smart meter and electrc vehcle chargng data to predct dstrbuton network mpacts," n Innovatve Smart Grd Technologes Europe (ISGT EUROPE), th IEEE/PES, 2013, pp Paper No 0323 Page 5 / 5