1、 AbstractEvaluating the effects of users demand response(DR)accurately has been increasingly important with the promotion of demand response.However,there exists a relatively large error in the generic baseline calculation method because it ignores the operating characteristics of air conditioning.T
2、his paper establishes an air conditionings DR effects evaluation model and proposes a baseline calculation method based on the difference between indoor and outdoor temperature.It is shown in case study that the method proposed in this paper can improve accuracy of baseline calculation effectively.I
3、ndex TermsAir conditioning(AC),baseline calculation,demand response,effects evaluation I.INTRODUCTION ASELINE load refers to the power that a user consumes per hour without performing demand response.Regression analysis,Kalman filter theory 1,fractal theory 2,neural networks 3 and other algorithms a
4、re used in early stage to provide a theoretical basis for load forecasting 4 with users demand response.Besides,ILDC(Inverted Load Duration Curve,ILDC)is firstly presented by peak and off-peak load as well as total electricity quantity in 5.The impact of a large scale of air conditioning on the reli
5、ability of power supply is analyzed in 6-7,in which the baseline load prediction method is also proposed.On the application level,PJM proposed a baseline calculation method based on load power and date type 8,which greatly improved the practicability of baseline calculation results.The baseline load
6、s calculation of different Manuscript received July 10,2018.This work was supported by State Grid Shanghai Municipal Electric Power Company under Grant 52096017000B.J.Yang is with Economic and Technical Research Institute of Shanghai Municipal Electric Power Company,Shanghai 200001 China(phone:13764
7、610768;e-mail:).C.Gao is with School of Electrical Engineering,Southeast University,Nanjing,Jiangsu 210096 China(e-mail:).X.Ding is with School of Electrical Engineering,Southeast University,Nanjing,Jiangsu 210096 China(e-mail:).types is carried out in three aspects including data selection,calculat
8、ion method and modified algorithm in the Power-saving Measurement Verification Technical Requirements of Users Demand Response published by CSEE in 2014.However,there exists a relatively large error in the generic baseline calculation method because it ignores the operating characteristics of air co
9、nditioning.This paper proposes a baseline calculation method based on the generic method combined with the operating characteristics of air conditioning.It is shown in case study that the method proposed in this paper has high practicability.At the same time,it can improve accuracy for baseline calc
10、ulation effectively.II.AIR CONDITIONINGS BASELINE MODEL A.Air Conditionings Basic Model The model mostly used in the simulation of thermodynamics in a room with air conditioning is ETP(equivalent thermal parameters,ETP)9,10.The relationship between indoor temperature and cooling capacity is shown in
11、 formula(1)-(2):+1+1=()dtt t t t t tRCin out AC out AC inT T Q R T Q R T e(1)AC AC ACQP(2)where Ttin is the indoor temperature at the time of t;Ttout is the outdoor temperature at the time of t;QtAC is the cooling capacity at the time of t;R is the equivalent thermal resistance;C is the equivalent t
12、hermal capacitance;P AC is the electric power consumed by air conditioning;AC is energy efficiency ratio.The relationship of frequency,cooling capacity and electric power is presented in formula(3)and(4):=AC ACP n f m(3)2AC AC ACQ a f b f c(4)where f AC is the frequency of compressor;n and m are con
13、stant coefficients of linear function(3);a,b and c are constant coefficients of quadratic function(4).Effects Evaluation Strategy of Air Conditionings Participation in Power System Regulation Jianlin Yang1,Ciwei Gao2,Xiaoye Ding2,Mingxing Guo1,Fei Fei1,Sisi Ma2 1.Economic and Technical Research Inst
14、itute of Shanghai Municipal Electric Power Company,Shanghai 200001 2.School of Electrical Engineering,Southeast University,Nanjing 210096 B 2018 China International Conference on Electricity Distribution Tianjin,17-19 Sep.2018CICED2018 Paper No.201805280000148 Page1/52211f AC is often decided by the
15、 starting frequency and the difference between indoor temperature and setting temperatureT=T in-T s.T is detected by temperature sensor every time interval.f AC varies in f AC,min,f AC,max,so the relationship between real-time frequency and the frequency at the last moment can be denoted as:,maxmi1n
16、,0ACAC n AC nACf N Tn T Nf f K TKf T n(5)where f AC,min f AC,n f AC,max and K is a constant.The relationship of cooling capacity and electric power is established through f AC in formula(1)-(5),by which the load model in an air conditioning room is presented in the way of converting the demand for c
17、ooling capacity to the demand for electric power.B.Relationship Between AC Load and Temperature The relationship between AC load and temperature is shown in formula(6)-(8)in a stable condition.24()(,)=2out sAC out sTTb b a cRP T T n ma(6)out sT T T(7)122()(4()ACdP TnTb a cd T R R(8)It is considered
18、that variation of AC load is only related to the difference between indoor and outdoor temperature for the reason that basic parameters including a,b,c,n,m and R are constants.C.Baseline Calculation Method Based on Similar Temperature Difference Days A baseline calculation method based on similar te
19、mperature difference days is proposed in this section.Demand response similar days,whose outdoor temperature data are more similar to the DR day,are choosed from historical data.The steps are introduced as follows.1)Assuming that d n is the DR day and air conditioning is operating in a stable condit
20、ion.If d n is a weekday,let d n-1,d n-2,d n-i be the continuous i weekdays before d n.Set the load cycle as T 0,the corresponding temperatures of t n-1,1,t n-1,2,t n-1,j-1,t n-1,j as T out(n-1,1),T out(n-1,2),T out(n-1,j-1),T out(n-1,j),the setting temperatures as T s(n-1,1),T s(n-1,2),T s(n-1,j-1),
21、T s(n-1,j)and the historical loads as P(n-1,1),P(n-1,2),P(n-1,j-1),P(n-1,j).All the data mentioned above are known already.2)The outdoor temperature difference at t k in d n-1 is calculated as formula(9).Data in other weekdays are calculated likewise,by which we can obtain the temperature difference
22、s from t 1 to t j in every DR weekday.(1,)(1,)(1,)1,2,.,out sT n k T n k T n k k j(9)Where T(n-1,k)is the temperature difference at t k in d n-1;T out(n-1,k)is the outdoor temperature at t k in d n-1;T s(n-1,k)is the indoor temperature at t k in d n-1.3)Select the indoor temperature at the last mome
23、nt before DR program as the indoor temperature T s(n)in d n.The temperature difference at t k in d n is calculated as formula(10).(,)(,)()1,2,.,out sT n k T n k T n k j(10)4)The variance between the temperature differences in d n-1 and d n is shown as follows:211(1,)(,)1,2,.,jkn T n k T n k k j(11)5
24、)Range the variances calculated for each weekday in an ascending order and select the three ones in the front as DR days,denoted as similar days d 1,d 2 and d 3.The mean value of historical loads from t 1 to t j is calculated in formula(12).1 2 3(,)(,)(,)(,)1,2,.,3P k P k P kP n k k j(12)The mean va
25、lues of historical loads constitute the baseline load in the DR weekday,as shown in(13).()(,1),(,2),.(,1),(,)P n P n P n P n j P n j(13)6)If d n is weekend,let d n-1,d n-2,d n-i be the continuous weekends before d n.Then carry out the calculation as steps(2)-(5).D.Baseline Calculation Method Based o
26、n Temperature Difference Regression Analysis Linear regression analysis is used in this section to establish baseline calculation model during DR period based on temperature difference.The steps are introduced as follows.1)Assuming that d n is the DR day and air conditioning is operating in a stable
27、 condition.If d n is a weekday,let d n-1,d n-2,d n-i be the continuous i weekdays before d n.Set the load cycle as T 0,the corresponding temperatures of t n-1,1,t n-1,2,t n-1,j-1,t n-1,j as T out(n-1,1),T out(n-1,2),T out(n-1,j-1),T out(n-1,j),the setting temperatures as T s(n-1,1),T s(n-1,2),T s(n-
28、1,j-1),T s(n-1,j)and the historical loads as P(n-1,1),P(n-1,2),P(n-1,j-1),P(n-1,j).All the data mentioned above are known already.2)The gross load at t j in each weekday is calculated as follows:000()()()()()s s out ssP j P P j P K T j T jP K T j(14)where P(j)is the baseline load at t j;P 0 is the b
29、asic element part of AC load;P s(j)is the temperature difference sensitive load element at t j;K s is the coefficient of P s(j).2018 China International Conference on Electricity Distribution Tianjin,17-19 Sep.2018CICED2018 Paper No.201805280000148 Page2/522123)Set the sequence of historical tempera
30、ture difference sensitive load element as P s(1)=P s(n-1,1),P s(n-2,1),P s(n-i,1),historical temperature as T out(1)=T out(n-1,1),T out(n-2,1),T out(n-i,1),historical indoor temperature as T s(1)=T s(n-1,1),T s(n-2,1),T s(n-i,1).The sequence of sensitive elements,outdoor temperatures and indoor temp
31、eratures are shown in formula(15):(,1)(,1)(,1)(1)(1,1),(2,1),(,1)out sT n i T n i T n iT T n T n T n i(15)where T(1)is the sequence of historical differences between outdoor temperatures and indoor temperatures at t 1.4)Substitute P s(1)=P s(n-1,1),P s(n-2,1),P s(n-i,1)and T 1=T(n-1,1),T(n-2,1),T(n-
32、i,1)calculated in step 3 into formula(16)to obtain P 0 and K s.Then evaluate the effectiveness of regression analysis according to the significance level of regression function.(,)(,)()out sT n j T n j T n(16)5)Substitute P 0,K s and T(n,1)into formula(14)to obtain P(n,1)and P(n,1),P(n,2),P(n,j)like
33、wise.The sequence of baseline load is shown as follows.()(,1),(,2),.(,1),(,)P n P n P n P n j P n j(17)6)If d n is weekend,let d n-1,d n-2,d n-i be the continuous weekends before d n.Then carry out the calculation as steps(2)-(5).III.EVALUATION INDICATORS OF BASELINE A.Relative Tolerance Set the loa
34、d cycle as T 0,the set of actual load data at each moment in d n is shown as formula(18).()(,1),(,2),.(,1),(,)P n P n P n P n j P n j(18)Let P j be the difference between baseline load and actual load.,jP P n j P n j(19)Then the relative tolerance of baseline load can be shown as:100%bPPPP(20)where
35、bP is the mean value of baseline load;P is the mean value of actual load.B.Mean Square Error(MSE)The MSE of baseline load and actual load is calculated as follows:2 2 212.jP P PPnj(21)IV.CASE STUDIES AND SIMULATION RESULTS A.AC Load Data Options on DR Day In this section,we select time period 10:00-
36、17:00 from July 1st to 13th as DR period.AC parameters are shown in TABLE I.Load data at 8:00 in the morning are used to calculate meteorology factor K.The outdoor temperatures and indoor temperatures in time period 10:00-17:00 from July 1st to 13th are presented in Fig.1.TABLE I AC parameters Param
37、eters Unit Value or distribution P set kW 3300 dt s 1 T max 28 T min 22 T out 35 f max Hz 110 f min Hz 0-5 T 0(23.4,1.122)R/kW(10,12)C kJ/(200,0.52)a/(-0.23,0.0252)b/(50,32)c/(200,502)n/(18,32)m/(50,102)Fig.1 Indoor and outdoor temperatures from 10:00 to 17:00 B.Baseline Calculation Based on Similar
38、 Temperature Difference Days(STD)The baseline loads based on STD are calculated in this section.The results are shown in Fig.2 with the black solid line representing the actual load,the black dashed line representing the baseline load based on generic method and the red solid line representing the b
39、aseline load based on STD.It is indicated in 2018 China International Conference on Electricity Distribution Tianjin,17-19 Sep.2018CICED2018 Paper No.201805280000148 Page3/52213Fig.2 that the results of STD,compared with those of generic method,are relatively accurate to measure the electric power A
40、Cs consume without DR programs,which can evaluate the load curtailment effectively.Fig.2 Baseline load by generic method and STD Besides,differences between the baseline load results by STD and actual load are shown in TABLE II.It is shown in Fig.2 and TABLE II that although the baseline load calcul
41、ated by generic method can reflect the average power-consuming condition approximately in a long period,there exist relatively large errors on the specific load value at each moment.The MSE is up to 135.67.The generic method cannot reflect the DR implementation effects in a short period accurately a
42、nd has an impact on the judgment of users DR execution,for the reason that DR program is usually carried out in the specific period on a DR day.On the contrary,the baseline load calculated by STD can reflect the actual power-consuming condition accurately at each moment.The MSE declined to 41.65,red
43、uced by 69.3%compared with the generic method.The difference between average load and actual load is merely 1.68%,which is relatively precise.TABLE II Results comparison between generic and STD methods Actual load/W Generic results/W STD results/W Average load/W 367.02 351.62 360.87 Relative toleran
44、ce 4.20%1.68%MSE 135.67 41.65 C.Baseline Calculation Based on Temperature Difference Regression Analysis(TDR)The baseline loads based on TDR are calculated in this section.The results are shown in Fig.3 with the black solid line representing the actual load,the black dashed line representing the bas
45、eline load based on generic method and the red solid line representing the baseline load based on TDR.It is depicted in Fig.3 that the TDR method can reflect ACs power-consuming condition without DR programs more accurately compared to STD.Fig.3.Baseline load by generic method and TDR Additionally,d
46、ifferences between the baseline load results by STD and actual load are shown in Table III.It is shown in Fig.3 and Table III that the baseline load based on TDR can fit perfectly with actual load,with the difference of only 0.21%compared with actual load.Meanwhile,the MSE descended to 9.131,reduced
47、 by 93.27%and 78.07%respectively compared with the generic method and STD.It can be seen from Table III that TDR progressed comparing with STD,because the selection of similar days are based on finite historical data while the interior relationship of temperature difference and AC load can be analyz
48、ed and extrapolated quantitatively by regression analysis.Table III Results comparison between generic,STD and TDR methods Actual load/W Generic results/W STD results/W TDR results/W Average load/W 367.02 351.62 360.87 367.79 Relative tolerance 4.20%1.68%0.21%MSE 135.67 41.65 9.131 V.CONCLUSIONS Bas
49、eline load calculation methods based on STD and TDR are proposed in this paper according to the deficiencies lying in the generic method.We obtained consumers load values and indoor temperature information by smart meters,based on which a regression analysis model is established.The results of case
50、studies show that the proposed method can predict the power-consuming condition effectively.We can draw the conclusions as follows:1)The generic method is based on the loads condition of similar days,often used to evaluate the power-consuming 2018 China International Conference on Electricity Distri
