ASSESSMENT OF THE POWER CURVE FLATTENING METHOD: AN APPROACH TO SMART GRIDS

Size: px
Start display at page:

Download "ASSESSMENT OF THE POWER CURVE FLATTENING METHOD: AN APPROACH TO SMART GRIDS"

Transcription

1 ASSESSENT OF THE POWER CURVE FLATTENING ETHOD: AN APPROACH TO SART GRIDS S. CARILLO APARICIO F. J. LEIVA ROJO Giacomo PETRETTO Gianluca GIGLIUCCI SmartGrids Endesa Red I+D Endesa S.A. Enel Ingegneria e Ricerca s.p.a Italy Enel Ingegneria e Ricerca s.p.a Italy susana.carillo@endesa.es Franciscojavier.leivar@scpec.net giacomo.petretto@enel.com gianluca.gigliucci@enel.com A. HONRUBIA-ESCRIBANO L. GIENEZ DE URTASUN A. ALONSO HERRANZ. GARCIA-GRACIA CIRCE (Centre of Researc for Energy Resources and Consumption) Spain CIRCE Spain CIRCE Spain CIRCE Spain aonrubia@fcirce.es lauragdu@fcirce.es adalonso@fcirce.es mggracia@fcirce.es ABSTRACT Tis paper presents an assessment of several metods to analyse power curve flattening in power systems. Several functions are commonly used to evaluate te flattening of te power demand curve. However, due to te ig deployment of Smart Grids, an evaluation of te suitability of te power curve flattening parameter is needed in current power systems. Results are tus of a valuable interest for bot Distribution System Operators and Transmission System Operators, wo will be able to assess teir Smart Grid Projects wit a iger degree of accuracy. Te developed work is based on te Smartcity álaga Project led by ENDESA and ENEL. I. INTRODUCTION Power demand curves describe power consumptions according to te time of te day. As an example, Figure sows te power curve for December, 5 t 203 of Spain obtained from te Spanis Transmission System Operator, Red Eléctrica de España (REE). In tis figure, current demand (depicted by yellow color) sows te instantaneous value of electricity consumption at te specified time of te day. A peak value over 32 GW and a valley minimum consumption wit more tan 0 GW of difference wit te maximum can be seen. Demand forecast (green color) is calculated by REE based on istorical electricity consumption data from similar periods, adjusted by working and seasonal patterns, as well as te economic activity. Finally, red color sows te sceduled production of te generating units wic ave been allocated te responsibility for supplying te electricity required to meet te forecasted demand. Altoug te consumption profile sown in Figure represents only one day, it reflects te common sape of daily power demand profiles in winter for Spain. Similar profile sapes are found in developed countries [], [2]. Tis common profile is usually composed of two peaks: te first one usually occurs between 0:00 and 4:00, and te second one is usually placed in te late evening (between 20:00 and 22:00). Also, te daily power demand profile usually contains a valley region (also known as off-peak region), wose lowest value is Figure. Electricity demand on te Spanis peninsula during December 203, te 5t. commonly located early in te morning, between 03:00 and 05:00. Te design of power system networks implies tat tese ave to be dimensioned for a safe load supply during peak load ours [3], wic occur a few ours a day as can be deduced from Figure. Neverteless, tis is an uneconomical situation since networks are tus oversized for te rest of te year. Under tis framework, te implementation of appropriate incentives and efficient energy consumption sceduling algoritms is a primary goal for te Smart Grids, SGs, [4]. Basically, te SG concept integrates information and communication tecnology systems into te power system in order to meet several objectives, []. On te one and, from te customers point of view, SGs suppose an increase of economic savings, comfort, social responsibility and awareness. From te power system utilities point of view, te increase of system efficiency and cost effectiveness, as well as te improvement of reliability and power quality appear as one of te most important issues involved in SGs. One of te means used to acieve te previous goals is te flattening of te power demand curve. Tis implies te use of different cost-effective initiatives to reduce or move peak consumptions to valley ours, suc us batteries [3], demand response management [5], [6], new tariff structures [2] Taking previous ideas into account, te present paper evaluates several metods commonly used to assess te Paper No 0398 Page / 5

2 flattening of te power demand curve. Several demand curves based on real measurements are used to study te influence of different weigting parameters over te flattening equation proposed. Te developed work is based on te Smartcity álaga Project led by ENDESA and ENEL. Te paper is structured as follows: After tis introduction, different metods to assess te flattening of te demand curve are discussed in section II. Section III presents te metodology and proposal developed in tis work and section IV provides te results dealing wit te use of tis metod. Finally, section V collects te conclusions. II. REVIEW OF POWER DEAND CURVE FLATTENING ETHODS One of te means commonly used to assess te flattening of te power demand curve is by te calculation of te system peak factor [2], also known as load factor [7]. Tis factor, called as in Eq. (), gives te rate between te maximum peak demand and te average demand during a specified period of time. In te present work te period of time considered is - (daily profile). max( P ) = () average ( P ) : Were, P represents te power P consumed at our. Hig load factors indicate a very low efficiency for te local distribution system capability to carry electricity, wereas low numbers indicate a ig efficiency. A typical load factor value for omes is 2 [8]. In addition, some oter factors ave to be considered in order to carry out a complete assessment of te power demand curve. In tis way, two additional factors are proposed in te present work. On te one and, te rate between maximum and minimum power consumption is identified as a factor directly associated wit te flattening of te power curve, Eq. (2). In fact, an ideal flat power demand curve would provide a value 2 =. On te oter and, it becomes also important to take into account te utilization factor. Tis factor describes te rate between te capacity of te line currently used and te capacity of te line in te case of maximum consumption during ours, Eq. (3). Actually, (3) provides knowledge about te oversized network degree due to te maximum peak. 3 = max( P ) 2 (2) min( P ) = P i= (3) max( P ) III. DESCRIPTION OF THE DESIGNED ETHODOLOGY Taking into account te equations described in Section II, and based on a daily time period, Eq. (4) is proposed as te only function to assess te flattening of te power demand curve, being FF te flattening factor. FF = W ( ) ( ) P max P max P = + + W2 W3 (4) min( P ) average( P ) max( P ) Tree weigts (W, W 2, W 3 ) associated wit te tree addends involved in (4), wic were presented in Section II, are included in te proposal. Once defined te equation proposed, five power demand curves ave been selected to assess te suitability of te tree weigts included in te equation: Curve : te first demand curve taken into account is te ideal situation, wic is defined as a completely flat demand curve (also called as constant power demand curve). Tis ideal power curve will be able to provide comparable results wit respect to te oter demand curves. Curve 2: real power demand curve. Concretely, te real profile considered in tis work is sown in Figure. Curve 3: tis curve is a modification of curve 2. Specifically, te second peak of curve 2 is removed. Tis is performed troug doing constant te power consumed at 8:00 until 23:00. Curve 4: tis curve is a modification of curve 2. Specifically, te valley region of curve 2 is removed. Tis is performed troug doing constant te power consumed from 02:00 to 09:00. Curve 5: tis curve is a modification of curve 3. Te valley region of curve 3 is removed. Terefore, after curve, curve 5 is te flattest power demand curve as tis curve beaves as if peak our consumptions are moved to off-peak ours. Figure 2 presents te previous considered list of power demand curves. Te ideal curve (curve ) is drawn wit black colour and eac power value is marked wit circles. It is clearly seen tat tis ideal curve as a constant value. Ten, curve 2 as a iger widt because tis curve represents te real daily power demand curve of te Spanis Power System sown in Figure. Curve 3 follows te profile of curve 2 until 8:00 because from tis our te consumption is assumed constant. Curve 4 as also te same profile as curve 2, but wit te exception of te constant consumption between 02:00 and 09:00. Finally, curve 5 results as te combination between curves 3 and 4, wic is clearly sown in Figure 2. Paper No 0398 Page 2 / 5

3 set of weigts (WG) te FFs calculated are constant. Actually, tis is an interesting result, because it implies tat te assessment of te demand curve flattening depends on te sape of te profile, wic is a necessary requirement to define a suitable demand curve flattening metod. Table 2. Flattening factors obtained wit tree identical power demand profiles. Figure 2. List of power demand curves used in tis work. IV. RESULTS Several analyses ave been carried out to assess te suitability of te tree weigting parameters included in te flattening demand curve equation, (4). IV.I Comparison of identical profile sapes Te first study is aimed to analyse te effect of te weigts over tree power demand curves wit identical profiles. Concretely, curve 2 (see Figure 2) as been selected as te reference curve, wereas te oter two profiles are obtained by just moving up and down, respectively, eac value of power consumed at eac our. Considering tese identical profiles, seven groups of weigts (also referred as weigt groups, WG, or set of weigts) ave been selected, as described in Table. As can be seen, te sum of te weigts is equal to in all te configurations, but te value of te weigt is canged between configurations. For example, te first set of weigts provides identical level of importance to te tree addends sown in (4). Table. Weigt groups (WP) used for calculation. W W 2 W 3 WG 0,33 0,33 0,33 WG 2 0,50 0,25 0,25 WG 3 0,25 0,50 0,25 WG 4 0,25 0,25 0,50 WG 5 0,90 0,05 0,05 WG 6 0,05 0,90 0,05 WG 7 0,05 0,05 0,90 Te previous list of weigts as been used to obtain te flattening factor, FF, defined in (4) for te tree demand curves wit identical sape considered. Results are presented in Table 2. As it can be seen, for any specified Curve 2 oved up curve 2 oved down curve 2 WG,2036,2036,2036 WG 2,2942,2942,2942 WG 3,25,25,25 WG 4,05,05,05 WG 5,56,56,56 WG 6,2306,2306,2306 WG 7 0,8686 0,8686 0,8686 IV.II Assessment of te influence of te weigt parameters value on te FF Tis section evaluates te influence of a wide range of weigts on te flattening factor calculation. Te five power demand curves depicted in Figure 2 will be used. Te criteria for te selection of te best set of weigts will be tat set of weigts tat order te demand curves by teir flattening. Tis means tat curve must be te first (tis is ideal, constant power curve), te second best sould be curve 5, curve 2 sould be te worst, and bot 3 and 4 curves sould be in te middle. Te order is given by te proximity to value, wic represents te constant flattening factor value given by te ideal flat power demand curve. Under te previous framework, and taking into account te group of weigts (WG) previously defined in Table, te calculated FFs are presented in Table 3. In order to obtain a grapic view of te results, Figure 3 summarizes te information contained in Table 3. Obviously, te FF of curve is constant equal to because tis is te ideal flat power demand curve. As can be clearly seen, te sevent set of weigts cannot be accepted because it provides almost te same flattening factor value for curves 3 and 5, wereas curve 3 is less flat tan curve 5. Table 3. Flattening factors obtained in te five curves wit seven set of weigts collected in Table. Curve Curve 2 Curve 3 Curve 4 Curve 5 WG,0000,2036,28,473,0739 WG 2,0000,2942,805,22,092 WG 3,0000,25,62,650,0763 Paper No 0398 Page 3 / 5

4 WG 4,0000,05,0688,0657,036 WG 5,0000,56,323,3645,940 WG 6,0000,2306,027,2074,082 WG 7,0000 0,8686 0,945 0,8699 0,9456 Figure 3. Flattening factors obtained in te five curves wit seven set of weigts collected in Table. Wit te aim of studying in detail te influence of te value of te weigts over te flattening factor, an extensive study is carried out in te following. For tis purpose, a total amount of 66 sets of weigts are selected. Tese sets of weigts represent all te weigt values combination possibilities between 0 and using 0, steps. Additionally, te /3 set of weigts as been added in te last row of weigts to be used. Terefore, te total number of weigt groups considered is equal to 67. As an example, Table 4 summarizes some of te first fifteen weigt groups used in tis compreensive study. Table 4. Some of te weigt groups (WP) used for calculation. W W2 W3 WG 0,0 0,0,0 WG 2 0,0 0, 0,9 WG 3 0,0 0,2 0,8 WG 3 0, 0, 0,8 WG 5 0, 0,3 0,6 Terefore, using te defined 67 sets of weigts, te calculated flattening factors related to te five power demand curves defined in Figure 2 are sown in Figure 4. Te main effects of te weigt values can be deduced from tis figure. Firstly, it must be pointed out tat it as not been found a unique set of weigts tat allow te best power demand curve flattening identification. Neverteless, it must be remarked tat tere are some sets of weigts tat sould not be used because te obtained FF would be wrong. Specifically, te next sets of weigts sould be avoided to prevent misleading results: 5, 6, 5, and 23: Tese configurations provide almost identical flattening factor values for te four non-ideal power demand curves., 2, 3, 4, 7, 8, 9, 0,, 2, 3, and 4: Tese configurations provide almost te same flattening factor results for curves 3 and 5. Some of tese configurations even provide a better FF result for curve 3 tan for curve 5. V. CONCLUSIONS Tis paper as evaluated in detail te influence of te weigting parameters used to assess te flattening degree of te power demand curve. A real power demand profile and several modifications, as well as an ideal flat demand curve, ave been used. A wide range of sets of weigts ave also been considered. After te tests carried out, interesting conclusions ave been acieved: Te ideal power demand curve, wic is described by a constant demand at eac our, provides a flattening factor value equal to. Tis statement is valid wen te sum of te tree weigts is equal to. Terefore, te sum of te tree weigts must be always equal to in order to obtain valuable results. It as been proven tat te assessment of te demand curve flattening depends on te sape of te profile. Tis means tat profiles wit identical sape will generate te same flattening factor value. In fact, tis is a necessary requirement to define a suitable demand curve flattening metod. It as not been found a unique set of weigts tat allow te best flattening factor value. However, tere ave been detected several sets of weigts tat cannot be used because misleading results would be obtained in tis case. Finally, based on te equation to assess te flattening of te power curve proposed and taking into account te different motivations tat may ave eac Distribution System Operator (DSO) utility, several weigt groups could be used. Paper No 0398 Page 4 / 5

5 Figure 4. Flattening factors obtained in te five curves wit 67 sets of weigts (WG). ACKNOWLEDGENTS Te autors would like to express teir appreciation to ENDESA and ENEL Group, for contributing wit valuable comments under te Smarcity álaga Project framework ( REFERENCES [] Qiao Hui-ting; Li Yi-jie, 20, "Researc of peak and valley period partition approac on statistics," International Conference on Electric Utility Deregulation and Restructuring and Power Tecnologies (DRPT), [2] Public Utilities Commission of Sri Lanka, 202, Study report on electricity demand curve and system peak reduction. [3] Dupont, G.; Baltus, P., 2009, "Dimensioning and grid integration of mega battery energy storage system for system load leveling," IEEE Bucarest PowerTec, -6. [4] Pang, G.; Kesidis, G.; Konstantopoulos, T., 202, "Avoiding overages by deferred aggregate demand for PEV carging on te smart grid," IEEE International Conference on Communications (ICC), [5] aarjan, S.; Quanyan Zu; Yan Zang; Gjessing, S.; Basar, T., 203, "Dependable Demand Response anagement in te Smart Grid: A Stackelberg Game Approac," IEEE Transactions on Smart Grid, vol.4, [6].H. Albadi, E.F. El-Saadany, 2008, A summary of demand response in electricity markets, Electric Power Systems Researc, vol.78, [7] Hammon, R., 202, Flat Load Profile Homes: Improving ZNEs? ACEEE Summer Study on Energy Efficiency in Buildings. [8] Ceniceros, B., Vincent, B., 2006, New Homes wit Load Sapes to ake an Electric Utility Drool: How to Integrate ultiple DS Strategies to Acieve Long Term Energy Performance Goals. ACEEE Summer Study on Energy Efficiency in Buildings. [9] Koutsopoulos, I.; Tassiulas, L., 202, "Optimal Control Policies for Power Demand Sceduling in te Smart Grid," IEEE Journal on Selected Areas in Communications. vol.30, no.6, Paper No 0398 Page 5 / 5