ENERGYCON 2014 • May 13-16, 2014 • Dubrovnik, Croatia

Data Pre-Processing and Representation for Energy Calculations in Net Metering Conditions Gianfranco Chicco, Valeria Cocina, Andrea Mazza, Filippo Spertino Politecnico di Torino, Energy Department Corso Duca degli Abruzzi 24, 10129 Torino, Italy {gianfranco.chicco,valeria.cocina,andrea.mazza,filippo.spertino}@polito.it Abstract—The different representations of the input data for energy system studies need to be conducted to the same time step for the purpose of carrying out power system calculations. Availability of data with different time steps, or even provided with irregular timing, needs a pre-processing phase to unify the representation within the same time step. This paper describes a dedicated pre-processing tool that accepts input data with different characteristics and provides a regular representation of the output. A specific application in which the differences between the time steps are relevant is analysed by considering a grid-interfaced system with local generation and local load, carrying out the input/output energy calculations and the assessment of the energy-related economics in net metering conditions. Keywords: Load pattern, Data representation, Pre-processing, Net metering, On-site power exchange.

kind of pre-processing at present is not available in the common solvers for power system analysis. The possible differences in the time step at which the input data are available may impact on the results of the energy calculations from systems interfaced with the grid and containing both local generation and local load. For these systems, two typical modes of connection are the on-site power exchange mode, in which the generation and the load are metered separately with two unidirectional meters, and the net metering mode, in which a bi-directional meter is used to separate the contributions of the positive net power (drawn from the grid) and of the negative net power (injected into the grid). Detailed information on the net power is important when an economic value is given to this information, depending on the amount of excess power injected into the grid and on the corresponding economic rate with which this power injection is rewarded in the time interval in which the I. INTRODUCTION electricity production occurs. These aspects have gained One of the key aspects of the data input for various tools importance with the diffusion of feed-in tariffs [1-10]. carrying out calculations for power system analysis is the This paper provides a twofold contribution to the studies difference among the way input data are available. For on data representation for energy system calculations. The example, the data gathered from the field could be provided at first contribution is the description of the characteristics and different time steps (1 minute, 5 minutes, etc.), or even in an usage of a pre-processing tool developed for passing from the irregular way (e.g., reducing the time step in time periods with initial data representations of the patterns involved in the larger variation in the pattern amplitude). This multiplicity of analysis to a uniform data representation at a given time step. conditions on the input data has to be solved before starting The second contribution is the study of the impact of different the calculations. In order to process energy systems data, one data representations and time steps on the calculation of the of the relevant points is to keep the meaning of the energy- total energy drawn from the grid or injected into the grid in based quantities. For this purpose, the points of a load pattern net metering conditions. In all the cases, the initial data provided as average power values at a given time step can be provided are assumed to be free of bad data, considering that transformed into the points of a load pattern provided at the time series available have been already checked for the another time step only if the total area of the load pattern presence of bad data and possible problems have already been (representing the energy consumption) is preserved. identified and corrected. For power system analysis, the presence of distributed The terminology used in this paper reflects the fact that generation and resources is typically calling for performing a most of the data considered represent the average power in a dedicated study spanning a specified time period. For this given time step and have to be used for the purpose of purpose, a unifying time step for the analysis has to be defined. calculating energy (i.e., with an integral meaning), not for In the presence of data available at different time rates, the reconstructing the waveform of a signal through data choice of the time step may be an option for the operator. In sampling. As such, the term sampling rate is not used in this any case, a dedicated pre-processing tool has to be available to paper, and the time interval referring to the data representation prepare the data in such a way to make them available at the is identified as averaging time step [11]. same analysis time step. This pre-processing tool needs to The next sections of this paper are organized as follows. address a number of issues referring to the meaning of the Section II describes the tool developed for creating regular data handled (e.g., for entries varying in a stepwise or in a data point sequences starting from data provided in different continuous way during time) and on the timing of the data formats and with different timings. Section III recalls the (provided at a regular time step or in an irregular way). This concepts of net metering and on-site power exchange. Section

978-1-4799-2448-6/14/$31.00 ©2014 IEEE

434

ENERGYCON 2014 • May 13-16, 2014 • Dubrovnik, Croatia

IV shows some examples for a system containing a local load and a local grid-connected photovoltaic (PV) plant, investigating on the effects of load and generation pattern representations at different time step on the results of the monthly energy calculations for positive and negative net power. Section V contains the conclusions. II. PRE-PROCESSING FOR UNIFORM TIMING OF THE DATA REPRESENTATION The specific function called Create pattern has been developed and implemented in the DERMAT tool for power system analysis, in the context of the European project SiNGULAR [12]. The basic structure of the DERMAT tool has been presented in [13]. The Create pattern function is directly called by the main procedure of DERMAT in the data input section.

‫ܘ‬୘்ǡఛ೚ ൌ  ሺܶǡ ߬௜ ǡ ߬௢ ǡ ǡ ‫ܘ‬୘்ǡఛ೔ ሻ

For the irregular patterns ( < 0), the points of the pattern in the input vector are introduced as successive pairs of values (time quantity). When  = 0 the time-independent parameter is a scalar value. For the regular patterns ( > 0), the time entries are not indicated, and the points represent the sequence of quantities represented at regular time intervals. Start

Createpatternfromdifferenttypesofinputdata

Patterntypecodes

A. The uniform time step pattern creation function

Irregular

Regular

The Create pattern function indicated here as  ሺǤ ሻ implements the creation of the pattern points from a given pattern type, input values and time step for the output pattern. The flow chart is shown in Fig. 1.

(1)

Stepwise

Stepwise

Linear interpolation

Linear interpolation

TABLE I. PATTERN TYPE CODES pattern type code  -2 -1 0 1 2 3 4 5 6

pattern type description

SelectInitialtimepoint(attheinitial point,midpointorendpointofthe firsttimestep)

Irregular with stepwise structure Irregular with linear interpolation Time-independent parameter Regular with stepwise structure and starting hour at the initial hour of the first time step Regular with linear interpolation and starting hour at the initial hour of the first time step Regular with stepwise structure and starting hour at the end of the first time step Regular with linear interpolation and starting hour at the end of the first time step Regular with stepwise structure and starting hour in the middle of the first time step Regular with linear interpolation and starting hour in the middle of the first time step

End

Fig. 1. Flow-chart of the Create pattern function.

The function inputs are the total period of analysis T (in minutes), the time step ߬௜ (in minutes) for regular input patterns (not considered for irregular patterns, in this case the value provided is ignored; for the sake of notation, this case is indicated by replacing ߬௜ with 0), the time step ߬௢ (in minutes) for the construction of the output pattern with regularly distributed points, the code of the pattern type  (see coding in Table I), and a row vector ‫ܘ‬୘்ǡఛ೔  containing the list of numerical values with different formats for regular or irregular patterns (the superscript T denotes vector transposition). The load pattern point created is included in the row vector ‫ܘ‬୘்ǡఛ೚ . The function output is the row vector ‫ܘ‬୘்ǡఛ೚ containing the output pattern with regularly distributed points at the fixed time step ߬௢ defined for the analysis. The Create pattern function is formulated as:

978-1-4799-2448-6/14/$31.00 ©2014 IEEE

435

The time steps ߬௜ and ߬௢ can be arbitrarily different with each other, that is, there is no reason for having them multiples of each other. This guarantees the flexibility of application of the tool. However, a consequence of this flexibility is that a correction factor has to be introduced when the pattern represents an average power, in order to preserve the total energy corresponding to the pattern in the period of analysis. The total energy Wi of the input pattern in the period of analysis is first calculated. The total energy Wo is then ෥୘்ǡఛ೚ determined starting from calculated from the pattern ‫ܘ‬ ୘ ‫் ܘ‬ǡఛ೔ on the basis of geometric considerations only, i.e., by maintaining the power level of the last initial point on the lefthand side of the current time instant (in the case of stepwise representation) or with suitable linear interpolation between the two initial points located on the left-hand side and on the right-hand side of the current time instant. The correction factor  is calculated as the ratio  ൌ ܹ௜ Τܹ௢ , and the final pattern with total energy ܹ௜ is obtained as: ෥୘்ǡఛ೚ ‫ܘ‬୘்ǡఛ೚ ൌ ‫ܘ‬

(2)

In the DERMAT tool, each pattern is introduced after the pre-processing by providing as input data the pattern name, the time step ߬௢ , and the pattern data ‫ܘ‬୘்ǡఛ೚ . The same format

ENERGYCON 2014 • May 13-16, 2014 • Dubrovnik, Croatia

is used u to introduuce determiniistic load patteerns, as well aas the evo olution in timee of the voltaage factor (i.ee., the multipllier of the slack voltagge used to take t into account the vooltage variiations occurrring in the upsstream networrk) and the paatterns of probabilistic p rrepresentationns of loads (sttandard deviaations, or parameters oof the probabilistic represeentations) or other entrries. The steepwise structures are usseful to reprresent variiables that caan assume only discrete vaalues in time (e.g., the tap position oof the transforrmer on-load tap changer, oor the susceptance of a power facctor compensaation system with swiitchable capaccitor banks). Some exampples of patternn created with h this functioon are repo orted below. The first exaample is a reg gular patternn with step pwise structurre, starting att the initial minute m of thee first tim me step (patterrn type code  = 1). The red circles arre the inittial hourly datta (߬௜ = 60 min). The blue dots d are the ppattern poin nts reported to the analysiss time step ߬௢ min. The efffect of the pattern creaation is show wn in Fig. 2 for ߬௢ = 755 min (corrrection factoor  = 1.00022) and in Fig. 3 for ߬௢ = 5 min (corrrection factor  = 1). In thhe figures, thee horizontal aaxis is timee (hours) andd the vertical axis a contains the average ppower with hin the relevaant time stepp. In Fig. 2 and Fig. 3, thhe last poin nt is includedd to complete the representtation and doees not con ntribute to the calculation off the total enerrgy Wo.

n type code  = 2). The reed circles aree the time step (pattern h the same vaalues initiaal hourly data (߬௜ = 60 minn), taken with used in the previo ous example. The blue dots are the patttern pointts reported to the analysis ttime step ߬௢ min. m The effecct of the pattern p creatiion is shownn in Fig. 4 for f ߬௢ = 75 min (corrrection factor  = 1.0024) and in Fig. 5 for ߬௢ = 5 min (corrrection factor  = 1).

Fig. 4. Initial and finall daily patterns w with linear interpo olation structure for ߬௜ = 60 min m and ߬௢ = 75 min. m

Fig. 5. Initial and finall daily patterns w with linear interpo olation structure for ߬௜ m and ߬௢ = 5 min. = 60 min

Fig. 2. Initial and finnal daily patternss with stepwise sttructure for ߬௜ = 660 min and ߬௢ = 75 min.

Furthermore, F considering c irrregular input load patterns, the repreesentation witth stepwise sttructure is sho own in Fig. 6 for ߬௢ = 75 min (correection factor  = 0.9980) an nd Fig. 7 for ߬௢ = 5 min (correction n factor  = 1), with th he same typee of repreesentation of the initial dat ata and constrructed patternns as the one o used in thee above exampples.

Fig. 3. Initial and finnal daily patternss with stepwise sttructure for ߬௜ = 660 min and ߬௢ = 5 min.

Cases with ߬௢ > ߬௜ are of interest when w data gatthered with h relatively hhigh time resolution (low w ߬௜ ) are useed for calcculations to be carried out at a lower tim me resolution (e.g., to reduce the computation time of procedures p ruun at successive time ssteps in a giveen time intervaal). Another exaample is a regular pattern with linear inteerpolation struucture, startingg at the initiall minute of thee first

978-1-4799-2448-6/14/$31.00 ©2014 IEEE

436

Fig. 6. Initial (irregularr) and final dailyy patterns with steepwise structure for ߬௜ m and ߬௢ = 75 min. m = 60 min

Finally, F for irregular input lload patterns the t representaation with linear interpo olation structuure is shown Fig. F 8 for ߬௢ = 75 min (correction faactor  = 0.99998) and Fig. 9 for ߬௢ = 5 min (corrrection factor  = 1).

ENERGYCON 2014 • May 13-16, 2014 • Dubrovnik, Croatia

ve role [14,15]. Efficient daata gathering and managem ment activ enables this activee role. At the grid interfacee, if the prosuumer ( sepaarate is alllowed to operate in net mettering mode (avoiding meteers for the eneergy injected iinto and draw wn from the grid), g the way the local productionn and consu umption data are h an impact ct on the energ gy assessmentt and repreesented may have on th he related economics. Thesse aspects are addressed in this sectio on. A. Geeneralized analysis at differrent averaging g time steps Let L us consider ߬௢ as the bbase averagin ng time step and defin ne the row veectors containi ning the load pattern data ‫ܘ‬୘்ǡఛ೚ and the t generation n pattern data ܏ ୘்ǡఛ೚ . In n order to geeneralize the analysis, lett us define other o averaaging time stteps multiple of ߬௢ , that is, ݉߬௢ with the multiiplier m = 1,… …, M for the lload patterns,, and ‫߬ݒ‬௢ withh the multiiplier v = 1,…, M for thhe PV generaation patterns.. To enable the compaarison, all loaad patterns arre reported too the samee time step ߬௢ . The net lload pattern ‫܌‬୘்ǡ௠ఛ೚ ǡ௩ఛ೚ is then defin ned in a generaal form as:

Fig. 7. Initial (irregullar) and final daily patterns with stepwise s structuree for ߬௜ = 60 0 min and ߬௢ = 5 m min.

w linear interppolation Fig. 8. Initial (irreggular) and final daily patterns with min and ߬௢ = 75 min. m struccture for ߬௜ = 60 m

w linear interppolation Fig. 9. Initial (irreggular) and final daily patterns with min and ߬௢ = 5 min. m struccture for ߬௜ = 60 m

Thee pre-processiing proceduree indicated ab bove is appliied to each pattern of a dataset. Thee computation nal burden deppends s and on thee time on the number of points in thee initial data set ps ߬௜ and ߬௢ . F For example, considering a dataset withh one step yeaar of hourly ddata (߬௜ = 60 min) to be represented w with 1min nute data (߬௢ = 1 min), thee execution off the Matlab® code imp plementation oon an i7-3740QM CPU at 2.70 2 GHz and 8 GB RA AM has resulteed in the averrage run timee (determined from 100 0 runs) of 10..2 ms for the stepwise stru ucture and 21 .4 ms for linear interpoolation. Withh many patterns to be anallysed, n be parallelissed to the calculations oon the differennt patterns can be carried c out sim multaneously on different processor core s. IIII. ANALYSIS O OF LOAD AND GENERATION PATTERNS IN NET METERIN NG CONDITION NS With the difffusion of the distributed d energy resourcees, the num mber of locatiions at which the customer may either eextract pow wer for servingg the local loaad or inject po ower into the ggrid is incrreasing. The customer is becoming a producer/cons p sumer (or prosumer) paarticipating inn the electricitty business wiith an

978-1-4799-2448-6/14/$31.00 ©2014 IEEE

437

‫܌‬୘்ǡ௠ఛ೚ ǡ௩ఛ೚ ൌ ‫ܘ‬୘்ǡ௠ఛ೚ െ ܏ ୘்ǡ௩௩ఛ೚

(3)

The T analysis of o the net loadd pattern at diifferent averagging time steps providees some hints on the effectss of the averagging time step on the reesults of the ennergy calculattions [11,16]. n order to carrry out an exttended analyssis by consideering In different averaging time steps for loads and d PV generatiions, c of the total enerrgy from the net n load patterrn in the calculation the period p of obseervation T muultiple of ‫߬ܯ‬௢ is carried ouut by chan nging the averraging time sttep independeently for load and PV generation g pattterns. ሺ௠ǡ௩ሻ Let L us denote as ்ܹǡఛ೚ thhe total energy in the periood T calcu ulated

from

the

ቄ݀ ்ǡ݉ ݉߬‫ ݋‬ǡ‫ ݋߬ݒ‬ሺ݇ሻǡ ݇ ൌ ͳǡ ǥ ǡ

load

net ் ఛ௢

pattern

‫் ܌‬ǡ݉߬‫݋‬ǡ‫ ݋߬߬ݒ‬ൌ

ቅ, with the ind dividual points of

the net n load patterrn reported too the same av veraging time step ߬௢ th hrough the application of the Cre eate patte ern functtion. The totall energy is exppressed as: ሺ௠ǡ௩ሻ

்ܹǡఛ೚

೅ ೅ ഓ೚ ൌ σ௞ഓ௞ୀଵ ݀ ்ǡ௠ఛ೚ ǡ௩ఛ೚ ሺ݇ሻ

(4)

Let L us further consider thee total energy y in the periood T calcu ulated by conssidering separrately the possitive componnents and the negativee componentts of the net n load patttern, respeectively: ೅

ሺ௠ǡ௩ሻ ഓ೚ ෡்ǡఛ ൌ σ௞ୀଵ ݀ ்ǡ௠ఛ೚ ǡ௩ఛ೚ ሺ݇ሻǡͲൟ ƒš൛݀ ܹ ೚

(5)

೅

ሺ௠ǡ௩ሻ ഓ೚ ෙ்ǡఛ ܹ ൌ σ௞ୀଵ ݀ ்ǡ௠ఛ೚ ǡ௩ఛ೚ ሺ݇ሻǡͲ Ͳൟ ‹൛݀ ೚

(6)

The T entries of the total energ rgy calculation ns are collecteed in ሺ௠ǡ௩௩ሻ ሺ௠ǡ௩ሻ ෡ ்ǡఛ ൌ ቄܹ ෡்ǡఛ the matrices ‫்܅‬ǡఛ೚ ൌ ቄ்ܹǡఛ೚ ቅ , ‫܅‬ ቅ and ೚ ೚

ENERGYCON 2014 • May 13-16, 2014 • Dubrovnik, Croatia

ෙ ்ǡఛ ෙ ሺ௠ǡ௩ሻሻ ‫܅‬ hich has M row ws and M coluumns, ் ೚ ൌ ቄ்ܹǡఛ೚ ቅ, each of wh for m = 1,…, M aand v = 1,…, M. M B. Effect of the economic rattes associated d with positivee and rgy values negative energ The positive and negative components of the total eenergy ong impact oon the are relevant because they maay have a stro locaal system, eespecially whhen the posittive and neggative amo ounts of the nnet power are associated to different econnomic ratees. In this casse, the effect of the calculaation of the en energy com mponents depeends on the raates at which electricity e is soold or bou ught. A generral study woulld require thee knowledge oof the patttern of variatiion of the eleectricity rates. In this papeer, the maiin concepts are highlightted by considering fixed rates durring time, usiing a parametter given by the t ratio z bettween the rate ௦ (mu/M MWh) at which electricity is sold to thee grid and d the rate ௕ (m mu/MWh) at which electriccity is boughtt from the grid (mu stannds for monetaary units): ‫ݖ‬ൌ

௦

ൗ

௕

ሺ೘ǡೡሻ ೚

ௐ೅ǡഓ

ሺ௠ǡ௩ሻ ሺ ෙ ሺ௠ǡ௩ሻ ෡்ǡఛ ൌ ௕ ܹ ் ೚ ൅ ௦ ்ܹǡǡఛ೚

B. Averaging A timee step effect onn the total enerrgy calculationns The T results off the analysis with differen nt averaging time t stepss are the net monthly m energgy matrix ‫்܅‬ǡఛ೚ , whose entries are all a equal with h each other ((at the value 1.214 MWh),, the ෡ ்ǡఛ (Fig. 11) andd the posittive net energy y component matrix ‫܅‬ ೚ ෙ ்ǡఛ (Fig. 12, negaative net eneergy componnent matrix ‫܅‬ ೚ show wing the ab bsolute valuees of the negative eneergy ෡ ்ǡఛ and ‫܅‬ ෙ ்ǡఛ are comp ponents). The entries in the he matrices ‫܅‬ ೚ ೚ different with each h other.

(7)

The calculatiion of the total energy cost for the local systtem in the pperiod of anallysis, taking into account local gen neration and llocal load, in function of the t averagingg time step p can be carrieed out dependding on the paarameter z, wiith the aim m of identifyinng possible margins m of con nvenience of using som me values of aaveraging timee step with resspect to otherss. The quaantity calculaated for caarrying out the convennience asseessment is thee total energy cost: ‫ܥ‬

me step. The PV generatioon is repreesented with a 5-minute tim repreesented with its absolute value in order to show the interaactions existin ng among thee two patterns at some hourrs of the day. d

Fig. 10. 1 Load and PV V patterns with 55-minute averagin ng time step forr four conseccutive days.

(8)

wheere the secondd addend is neegative. The reelated case stuudy is presented in the nnext section. IV. CASE STTUDY APPLICA ATION WITH NET METERING G This section contains thee results of some calculaations c The T base averraging carrried out in nnet metering conditions. tim me step is ߬௢ = 5 min. The period p of anaalysis is one m month (Jun ne). A. Generation annd load patterrn data Let us considder a grid-connnected local system contaaining both h generation aand load. The generation pllant is a rooftoop PV plan nt with peak ppower 7.5 kW Wp. The PV data d are gatherred at irreegular time steeps and are processed p by using u the Cre eate pat ttern functtion to get a series of valu ues referring to 5min nute averagingg time step. The local loaad is composeed of the aggregated demaand of 10 residential r apaartments (the sum of the co ontract powerss is 30 kW W), with data ooriginated from m 1-minute averaging a timee step and d synthesized into a series of values at 5-minute 5 averraging tim me step [17,18]]. An examplee of the loaad and PV patterns for four con nsecutive dayss is reported in Fig. 10, both b patterns being

978-1-4799-2448-6/14/$31.00 ©2014 IEEE

438

Fig. 11. 1 Monthly enerrgy calculated frrom positive net power variationns for differeent averaging tim me steps.

Fig. 12. Monthly energ gy calculated froom the absolute values v of the neggative ower variations fo or different averagging time steps. net po

ENERGYCON 2014 • May 13-16, 2014 • Dubrovnik, Croatia

Starting from m any entry of the positive or o absolute neggative ሺ௠ǡ௩ሻ ෡்ǡఛ eneergy matricess (e.g., the generic entry y ܹ andd the ೚ ሺ௠ǡ௩ሻ ෙ corrresponding eentry ்ܹǡఛ ), the entriies having llonger ೚

aveeraging time sstep for eitherr load (multiplier higher thaan m) or generation g (m multiplier highher than v) result in lower vvalues for both the posiitive energy and a the absolu ute negative eenergy hose differencce remains coonstant and equal e to the value (wh ሺ௠ǡ௩ሻ ்ܹǡǡఛ೚ ). In particular, looking at thhe diagonal en ntries of the eenergy mattrices (e.g., too the cases inn which load and PV generration are represented w with the samee time step), Fig. 13 show ws the y by increasinng the variiations occurrring in the moonthly energy aveeraging time sttep. These results provide quanntitative evideence to the facct that redu ucing the averraging time sttep it is possib ble to capture more variiations in the net energy paatterns, in parrticular providding a bettter identificattion of the tiime intervals in which thhe net pow wer fluctuates around the vaalue zero. In order to shhow details onn the net poweer representattion at diffferent averaginng time steps, four averagiing time stepss have beeen chosen (i.e.., 5 min, 15 min, m 30 min an nd 60 min). Fiig. 14 sho ows the net loaad patterns for these averag ging time stepps in a tim me interval of 55500 min, illuustrating the smoothing s efffect of incrreasing the avveraging time step.

Fig. 14. Evolution of the net power forr four different av veraging time steeps (5 o the lower piccture, min, 15 min, 30 miin, 60 min, froom the upper to respecctively).

1.6 1.4

monthly energy (MWh)

1.2 1

point nuumber

positive net energy

0.8

Fig g. 15. Zoom of th he net load patternn for different averaging time stepps.

to otal net energy

0.6

negative net energy

0.4 0.2 0 0

10

20

30

40 0

50

60

-0.2 -0.4

averagin ng time step (min))

Fig. 13. Net energy at different averraging time steps (diagonal termss of the net energy e matrix).

Fig. 15 show ws a zoom off the net load d pattern in a short me interval, ussing the four averaging tim me steps indiicated tim abo ove. It is interresting to notiice the evoluttion of the paatterns with h respect to thhe zero level. In the centraal part of the ffigure (fro om the point nnumber 251000 to 25500), iff the averagingg time step p is 5 min it is possible too obtain somee time periodss with possitive net loaad, while when w the aveeraging time step incrreases the perriods with positive net poweer are reducedd until the net power rem mains always negative.

978-1-4799-2448-6/14/$31.00 ©2014 IEEE

439

C. Averaging A timee step effect onn the total enerrgy costs Some S results of the investiigation on ho ow the averagging time step may afffect the totaal energy cossts are presented below w. These resu ults are show wn by consideering the casse in whicch the PV gen neration and lload pattern data d are analyysed with the same averraging time sttep. The T parameterr z defined inn Section IIII.B changes from f zero (the limit case in which noo reward is giv ven for the eneergy produ uced) to 2 (ii.e., the energgy produced is rewarded at a doub ble rate with reespect to the ccost of the eneergy bought). The resullts are shown in i Fig. 16. For F z = 1 th here is no diifference in the energy costs c depending on the averaging tiime step, as the t net energgy is alwaays the samee and the ppositive and negative eneergy contrributions comp pensate for eaach other. For F z > 1 (i.e., the generatiion is reward ded more thann the cost of the energy bought), it is advantageouss for the prosuumer duce the averaaging time steep, in order to identify betteer all to red the conditions leading l to ppositive or negative eneergy he ones shownn in Fig. 15). contrributions (as th

ENERGYCON 2014 • May 13-16, 2014 • Dubrovnik, Croatia

totalenergycost(mu)

1.5

REFERENCES

z 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2

1.4 1.3 1.2 1.1 1 0.9 0

10

20

30

40

50

60

averagingtimestep(min) Fig. 16. Effect of the averaging time step on the total energy cost.

Conversely, for z < 1 the higher detail achievable by reducing the averaging time step does not provide direct economic benefits. However, the reduction of the averaging time step may bring advantages when demand response programmes are associated with incentives to enhance the quality of the information gathered from the user, thus providing more detailed knowledge on the patterns to be analysed. V. CONCLUSIONS This paper has described a tool for constructing suitable patterns to be used in energy system analysis, taking into account the correction needed to these patterns (represented in stepwise or linear interpolation form) to obtain the same total energy in a specified time interval of observation. On the basis of this tool, a study has been carried out by considering different values of the averaging time rate to construct various load patterns for generation and load, in order to assess the energy calculations in net metering conditions. The results provide quantitative evidence on how lower values of the averaging time step correspond to higher fluctuations of the net power, and reducing the averaging time step means increasing the detail of information representation. The net metering issues considered in this paper do not show up in the case of on-site power exchange in which the electricity drawn from and injected into the grid are metered in a separate way. Under the net metering scheme with fixed energy rates (constant during time), the convenience of exploiting lower averaging time steps has been shown to appear only when the energy injected into the grid is rewarded at a higher rate with respect to the energy taken from the grid. ACKNOWLEDGMENT The research leading to these results has received funding from the European Union Seventh Framework Programme FP7/2007-2013 under grant agreement no. 309048, project SiNGULAR (Smart and Sustainable Insular Electricity Grids Under Large-Scale Renewable Integration).

978-1-4799-2448-6/14/$31.00 ©2014 IEEE

440

[1] A. M. Payne, R. D. Duke and R. H. Williams, “The impact of net metering on the residential rooftop PV market”, Conference Record of the Twenty-Eighth IEEE Photovoltaic Specialists Conference, pp. 1391–1394, 2000. [2] A. Campoccia, L. Dusonchet, E. Telaretti and G. Zizzo, “Feed-in Tariffs for Grid-connected PV Systems: The Situation in the European Community”, Proc. IEEE Lausanne Power Tech, pp. 1981–1986, 2007. [3] K. Sedghisigarchi, “Residential solar systems: Technology, netmetering, and financial payback”, Proc. IEEE Electrical Power & Energy Conference (EPEC), 2009. [4] N. R. Darghouth, G. Barbose and R. Wiser, “The impact of rate design and net metering on the bill savings from distributed PV for residential customers in California”, Energy Policy, Vol. 39 (9), pp. 5243-5253, September 2011. [5] C. Genao and A. St Leger, “Insight into demand response and photovoltaic source with Time of Day pricing”, Proc. IEEE PES Transmission and Distribution Conference and Exposition (T&D), 2012. [6] M. Fanjun and B. H. Chowdhury, “Economics of grid-tied customerowned photovoltaic power generation”, Proc. IEEE Power and Energy Society General Meeting, 2012. [7] T. E. Del Carpio-Huayllas, D. S. Ramos and R. L. Vasquez-Arnez, “Feed-in and net metering tariffs: An assessment for their application on microgrid systems”, Proc. Sixth IEEE/PES Transmission and Distribution: Latin America Conf. and Exposition (T&D-LA), 2012. [8] Y. Yamamoto, “Pricing electricity from residential photovoltaic systems: A comparison of feed-in tariffs, net metering, and net purchase and sale”, Solar Energy, Vol. 86 (9), pp. 2678-2685, September 2012. [9] T. E. Del Carpio-Huayllas and D. S. Ramos, “Economic evaluation of a residential solar microgrid connected to the distribution system”, Proc.IEEE PES Conference on Innovative Smart Grid Technologies Latin America (ISGT LA), 2013. [10] A. Poullikkas, “A comparative assessment of net metering and feed in tariff schemes for residential PV systems”, Sustainable Energy Technologies and Assessments, Vol. 3, pp. 1-8, September 2013. [11] G. Chicco, “Challenges for Smart Distribution Systems: Data Representation and Optimization Objectives”, Proc. 12th International Conference on optimization of electrical and electronic equipment (OPTIM 2010), Brasov, Romania, pp. 1236-1244, 20-22 May 2010. [12] European project SiNGULAR (Smart and Sustainable Insular Electricity Grids Under Large-Scale Renewable Integration), online: http://www.singular-fp7.eu/home/ [13] A. Mazza, G. Chicco, E. Bakirtzis, A. Bakirtzis, A. De Bonis and J. P. S. Catalão, “Power Flow Calculations for Small Distribution Networks under Time-Dependent and Uncertain Input Data”, IEEE Transmission & Distribution Conference, Chicago, IL, 14-17 April 2014. [14] G. K .H. Larsen, N. D. van Foreest and J. M. A. Scherpen, “Distributed Control of the Power Supply-Demand Balance”, IEEE Trans. on Smart Grid, Vol. 4 (2), pp. 828–836, 2013. [15] P. Goncalves Da Silva, D. Ilic and S. Karnouskos, “The Impact of Smart Grid Prosumer Grouping on Forecasting Accuracy and Its Benefits for Local Electricity Market Trading”, IEEE Trans. on Smart Grid, in press. [16] G. Chicco and A. G. Rusu, “Data sampling effects on net power analysis”, Scientific Bulletin of the Electrical Engineering Faculty, Valahia University of Targoviste (Romania),Year 7, No.1 (7), pp. 2125, 2007. [17] A. Cagni, E. Carpaneto, G. Chicco and R. Napoli, “Characterisation of the aggregated load patterns for extra-urban residential customer groups”, Proc. IEEE Melecon 2004, Dubrovnik, Croatia, vol. 3, pp. 951–954, May 2004. [18] E. Carpaneto and G. Chicco, “Probabilistic characterisation of the aggregated residential load patterns”, IET Generation, Transmission and Distribution, Vol. 2, No. 3, pp. 373–382, May 2008.