EEE Transactions on Power Systems, Vol. 8, No.

3, August 1993

1219

An Efficient Load Model for Analyzing
Demand Side Management Impacts
S. Rahman
Senior Member

Rinaldy
Student Member

Bradley Department of Electrical Engineering
Virginia Polytechnic Institute & State University
Blacksburg, VA 24061,USA
e-mail : SRAHMAN @VTVMl .cc.vt.edu

ABSTRACT
The main objective of implementing Demand Side
Management (DSM) in power systems is to change the utility's
load shape - i.e. changes in the time pattern and magnitude of
utility's load. Changing the load shape as a result of demand side
activities could change the peak load, base load andor energy
demand. Those three variables have to be explicitly modeled into
the load curve for properly representing the effects of demand side
management. The impact of DSM will be manifested as higher or
lower reliability levels. This paper presents an efficient technique
to model the system load such that the impact of demand side
management on the power system can be easily and accurately
evaluated. The proposed technique to model the load duration
curve will facilitate the representation of DSM impacts on loss-ofload probability, energy not served and energy consumption. This
will provide an analytical method to study the impact of DSM on
capacity requirements. So far iterative methods have been applied
to study these impacts. The proposed analytical method results in
a faster solution with higher accuracy. It takes only 18 seconds on
an 80486 PC to solve each case study involving different peak and
base loads, and energy use.
Keywords: Capacity expansion planning, demand side
management, load shape impacts, load duration curve, loss-of-load
probability, energy not served.

1.0 Introduction
In the era of load growth and increasing constraints on new
and existing generation capacity, Demand Side Management
(DSM) options are being considered all over the world as possible
bridges between these two apparently conflicting requirements.
The high variability of load from one day to another, and from one
hour to the next, may provide significant opportunities for demand
side management. DSM provides a workable solution to some of
the major problems confronting the electric utility today. There is
a great deal of uncertainty in future demand, fuel prices,
construction cost, availability and cost of power from other
utilities, independent power producers, and the regulatory

9 2 SM 422-6 PWRS A paper recommended and approved
by the IEEE Power System Engineering Committee of
the IEEE Power Engineering Society for presentation
at the IEEE/PES 1992 Summer Meeting, Seattle, WA,
July 12-16, 1992. Manuscript submitted January 21,
1992; made available for printing June 4 , 1992.

environment. This is leading electric utilities toward incorporating
DSM concepts in their resource planning as discussed in several
papers [3,4,6,7 and 113.
Utility programs falling under the umbrella of DSM
include load management, identification and promotion of new
uses, strategic conservation, electrification, customer generation
and adjustments in market share. The DSM activity can be
characterized as a two level process [I] :
Level I

- Load shape objective

Level I1 - End use
- Technology altervatives
- Market implementation methods
The load shape objective can be realized in various ways.
Six generic load shape objectives illustrate the range of
possibilities: peak clipping, valley filling, load shifting, strategic
conservation, strategic load growth, and flexible load shape. Once
load shape objectives have been selected, an appropriate set of
DSM programs needs to be identified.
Load Duration Curve (LDC) is the vehicle through which
DSM impacts are incorporated into power system planning and
operation. Models of the LDC is one of the most important tools
in the analysis of electric power systems. It has been utilized for
various purposes, such as estimating the operating cost of a power
system, predicting the amount of energy delivered by each unit,
and calculating reliability measures. Due to the nature of
electricity, and dependence of our society on its uninterrupted
supply, reliability is one of the most important design criteria of an
electric power system [ 171. Furthermore, reliability assessments
are a necessary part of power system studies used to assist in
managerial decisions regarding the adequacy and reliability of the
system as discussed in [18].

As the needs and opportunities for DSM are discussed, two
issues become obvious. First is, how to quantify the effects of
these myriad of options such that proper credits can be provided.
This would ensure that the utility is not unfairly subsidizing one
group of customers at the expense of the other. Second is, how to
represent the impact of demand side management activities in
capacity expansion planning models. In the absence of an
analytical model, there is currently no straight forward load model
for use in capacity expansion and production costing models.
There are DSM screening models available, though not much
research has gone into the modeling of the effects of DSM on the
system load for production costing models. A proposed approach
to addressing these issues is discussed in the following sections.

0885-8950/93$03.00 0 1992 IEEE

5) 1 A. The currently available models can not handle the variation of energy. . WASP uses Fourier series expansion to express the ILDC [IO]..70 Baseload : 744.1. especially after incorporating demand side management alternatives.. 12 monthly load duration curves are represented in Table 1. E = energy. they have to go back to the numerical data to show the changes in system load. Such accurate reliability assessments are only possible through the use of probabilistic models of the system load curve. h = total hour in the period .4a3a5+4a~a~+120a12a~a4+132a~+ a1 180alza:a.1220 2. This implies that the expression for the new load shape has to be derived each time there is any change in these three parameters. This model mathematically represents the amount of time (normalized) a certain level of load is present. base load (B) and total energy (E)..45 Peakload : 3 209.the ratio of the minimum to the maximum load during the period.3. = One of the important benefits of DSM is the reduction of generating capacity required to serve system load while maintaining a satisfactory level of reliability.36a~%2a.21ala:a3) a1 1 A. such as Wien Automatic System Planning Package (WASP). where: It is more convenient in probabilistic simulation work to use the load duration curve with the ordinate and abscissa reversed [9]. A direct inversion of the LDC expression to the ILDC causes the resulting mathematical expression to give unacceptable errors. base and peak load in the mathematical expression for ILDC directly..a13a5. X = 1oad.= --a? a2 + (5al+a3 . an accurate assessment of the effects of DSM on system reliability performance is needed. This is a more accurate model of the load shape for use in planning and operation activities. for: B<X<P. The data : July 1990 : Total energy : 3 977 202. = 0. the area under the load duration curve represents the total system energy requirement. This has been named the "VPI model". An LDC is commonly used to represent the system load over an extended period of time.. The model represents the Inverted Load Duration Curve (ILDC) directly as a function of peak load (P). peak load and energy served.and A. The analytical model of an inverse of the load curve is presented in the following.00 Total hour MWh MW MW Hour . . which try to model the LDC with a fifth-order polynomial obtained by fitting the hourly load data [9]. in order to show how well the VPI model represents the actual load of a system under different conditions. a data point is taken from Table 1 for demonstrating the closeness between the VPI model and the actual load data. Following is a discussion on how the impacts of DSM may be represented in the load curve. - 1 a. a test case is formulated and analyzed under various conditions.2-a~a. One of them is Snyder [12]. In order to represent the ILDC including the DSM impacts.2.2. P B = peakload. The model is : Now.7 The equation (1) above is a direct mathematical representation of ILDC in terms of peak load.a12a4.. and for : X > P.3a3a. = 7 (n+2). = -&3a~a2a6+8a. The new model As a sample. 2.. Snyder model represents the LDC with a fifth-order polynomial. There are several models that try to express the LDC mathematically. The actual load and generation data from a Virginia utility are used to test the model. f(x) = 1.. and the ratio of the average load to the maximum load.45 : 7 873. The impacts of DSM activities will manifest themselves in three different ways -. This form of the curve is called Inverted Load Duration Curve (ILDC) which can be used to estimate the loading of various generation units by placing equivalent capacities of the units on the curve. regenerate the LDC and calculate the ILDC again. The DSM induced changes in system load (total energy. = baseload. Two widely used indices of reliability are: loss-of-load probability (LOLP) and energy not served (ENS).72%3%a3a4-12a~a~- A. base and peak load) generate new load shapes. Thus the availability of a closed form analytical expression for ILDC will be a significant step forward to study the impacts of DSM.. base load and total energy.Sa:) a1 1 A. where the coefficients of the polynomial are closely related to two quantities -. In such cases it is difficult to represent the changes to peak load.changes in base load. As with a chronological hourly load curve.n = 1.+84ala23a3-a14a6-~a12a~a4 a1 28a12a2a: . There are other models. Accordingly.0. for : X < B . Test case study of the VPI Model 2. This is a slow process and is subject to errors. = 7(6a:a2a4 + 3aI2a: + 14a: .42a. = y(7a:a2a5 +7a. See appendix A for the derivation of the model. Proposed Analytical Model For ILDC Once the expected impacts of demand side activities on the system load are estimated. base load and energy usage in the load curve.. ways have to be found to represent these in the load curve model.. Table 2 represents the generating unit data for the test system.

3.90 3977202. T = the period of the Fourier series. and the inverted LDC’s (ILDC) for VPI.15 2596. For conciseness.70 2812.10 2679..05 7873.00 2740.. f(Xi) = A ~ ( ~ ..l ) + A ~ ( ~ . the data for all 13 generating units are used for the results in Table 3. Generating Unit Data Nuclear Coal Coal Hydro # of Size Unit - (MW 2 2 2 1 1 1 2 2 1200 1000 800 1000 800 600 800 1200 The performance of the VPI model is compared to several available models.40 3077.45 6802.676184 0.85 2399.45 2767919.25 2594.) ..l ) ~ + A 3 ( ~ . The ILDC representation for the Snyder model is as follows.. ELDC Comparison 1 Figure 1.661192 0... VPI model (equation (2)) and the actual. The close match between these curves show that the VPI model can be used to express the ILDC as a function of the peak load..55 5033. For this purpose one needs to represent the equivalent load duration curve (ELDC) which is the result of convolution of LDC and generator failure rates.71 1613 0.71 1613 0.55 2927772. The cumulant method is .10 2763..70 6600.80 5525. The equation for ELDC is given in equation (3)..05 3754270..2.704106 0.05 6524. Total unit = 13 Using eqn.l ) ~ + A q ( x . Availability ..00 7764. are plotted in figure 1.(4) 0.(4) and (5) respectively. and total energy demand.15 3209. The ILDC representation for the WASP model is as follows. k = 0. ( l ) .46 3427458.50 3500956. . Comparison of the VPI Model to Others leak Load Base Load Total Energy (MW) 7382.100 (max.700571 0.75 2842369.. one can see that total energy under the VPI model is closer to the actual data than total energy of WASP. as well as the plots in figure 2. base load. WASP and Snyder models are given in equations (1). the system loss-ofload probability (LOLP) and energy not served (ENS).71 1613 0. In Table 4.80 2873. WASP.25 5686.75 3060.1.60 8 9 10 11 12 (MW) (MWh) 372890123 2571077.( A g5)( ~ . which is widely used as a reference. ..00 5731.947025 where : ak = the coefficients for the kth harmonic of the Fourier series.. the LOLP values computed using four different models are shown. ILDC Comparison From Table 3. The next step is to check how well the VPI model represents the total energy content.l ) ~ where : Ai = the coefficients of the polynomial.45 3055495.60 3735832. Load Data I I 2.30 3114.1 ) ~ +.95 Table 2. However. the model of ILDC is found as shown below : Four ILDC’s. generated by inverted Snyder model. These include the cumulant method which provides a direct representation of LOLP without the convolution process. the analytical expression for ELDC is derived for three units only. I JULY 195 I I 1 JULY 199 Figure 2.55 7011.1221 Table 1..65 291 7258.20 6604.

02 2042369.01 110609. but the base and energy use are increased by 5% each.010105 0.61 -0.1 3 0.60 2571077.79 0.66 WASP df.09 0.46 194051.005546 0.15 2931254.91 76521. Improvements from the base case are apparent.013724 0.004295 6 1. is also expected to give the best estimate for energy not served.70 7.96 0 2.94 4059044.10 0.04 0.025052 0.90 Table 5.99 3512053.36 2044440.97 330942.15 421 9040.010423 12 0.60 4.002637 -0. ENS readings from the cumulant method are.000645 0.16].23 2571077.16 0. The reference case shows no DSM activity.33 2.1222 Table 3.36 7.005010 0.10 3515974.40 3735032.91 2069025.30 0.50 2917250.72 -0.36 00515.22 10 2.45 3427450.46 1.40 0. because it models the tip of ILDC most accurately (see figure 1).30 639506.19 0.23 3720901. it can be shown how convenient it is to study different load shape impacts. WASP VPI (GW (GWh) 1 2.039479 0. The resulting LOLP and ENS show significant improvement.40 2574772. Since some of the LOLP values from the cumulant method came out to be negative.16 773344.001677 11 0.70 11097.00 0.60 3695.001206 0.46 2042369.51 166.00 3427450.49 0. the peak load is kept constant. Table 5 provides a comparison of energy not served as computed by WASP.34 9 4.49 0. several of the cumulant generated LOLP values came out to be negative. The cumulant method [13] is also an efficient method to calculate the LOLP.001322 0.40 3735933.29 4 1.000039 0.05 2767919.012924 0. the speed at which these results are obtained is significant.07 2.26 0.04 2601607.90 2927772. it may not be very meaningful to compute energy not served (ENS) under these conditions.010663 0. In case 3 (load shifting).91 5. However. it demonstrates some instability when the largest unit in the system represents a significant percentage of the total load.22 10.009010 0.06 6.007630 0.51 2927772.69 10. the Snyder model. The convenience of easily demonstrating the reliability impacts of various DSM actions is thus established.010772 0.00 0.74 3003.01 1004 5 6 7 0.25 242646.69 3754270.52 3055495.78 2917250.014366 0. the energy use is kept constant.55 3500956. but the base load is increased by 5% and peak load is decreased by 10%.07 0.008750 Cumulant -0.36 3090251. In case 2 (peak clipping).55 0. 3. Resulting increases in the LOLP and ENS indices are as expected.0 Application of the VPI Model in DSM Activity Now that the accuracy of the VPI model in comparison to the other well established models has been demonstrated.47 11 0.97 455330. This method is used in the EPRI generation expansion model.03 2415900. However.00 3401.07 0.15.31 195197. not shown in Table 5.016000 0. a reference and three test cases are set up.51 3949121. Furthermore.05 2.004060 0.33 110149.15 106. However.12 0.45 3977202. In order to demonstrate the flexibility and the fast response time of the VPI model.89 34755.75 3735032.14 3 1.91 3904910.63 237.16 2042555.42 0.003775 0. Again 'we do not have a reference.00 3500956.63 0.000897 0.02 3977202. The three indices needed to define the load curve are shown.64 2767919.001107 0. the peak is reduced by 10% and energy by 5%.004405 3 4 0. LOLP Comparison Model No.000380 0.006499 0. Results show that the VPI model comes the closest to the Snyder model in terms of LOLP.90 7715. therefore. . B1.069394 0.03 63306.42 0. the Snyder model is known to provide the best LOLP. As this is the case with the example presented in this paper.92 2 1.04 0. Results of the VPI model come close to that of the Snyder model for this parameter as well.19 3696153. Snyder and VPI models.51 1.00 3037922.from actual 3729130.60 2053952.01 1915 0.65 3754270.006425 10 0.54 0.001001 0. Energy Not Served Table 4.004433 No. Energy Comparison (MWh) - No - 1 2 3 4 5 6 7 0 9 IO I1 I2 - - Actual Data I vpI Model VPI WASP Model clif.003399 0.001209 0.97 2771002.fron actual - 3720901.001433 5 2.57 41 91170.020106 0.95 3055495.001 296 0.29 0. because of its more accurate LOLP representation. VPI Snyder WASP 1 2 0. Finally in case 4 (valley filling).000072 discussed in references [13.71 2917091.14. EGEAS [5].000527 0. The VPI model was run on an IBM-PC Model 90 with a 80486 processor.021446 0.62 64. and compare the resulting LOLP and ENS indices. The solution time for each case study was mere 18 seconds.40 3427393.15 I .011203 9 0. There are no "actual values" for LOLP for use as reference.005077 0. These are shown in Table 6.01 14.Pl and E l represent the reference case values.006734 0.000757 0.14 3754270.00 100.29 7 5.0027 67 0 - 0.01 12 3.

Vo1. This model also provides a direct method to study the load factor impacts. Proceeding of 2nd WASP Conference. Schenk. C. load. V01.an Electric Utility Optimal Generation Expansion Planning Computer Code". "Wien Automatic System Planning Package (WASP) . December. CRC Press.W. base load and total energy demand. et al. Hirst. November 1990. and M.L. K. Keelin and C. W. Oak Ridge National Laboratory.S. Broehl. 0. 1981. J. June 1989.2 and 3. Vo1. Chan. 19-24 August. Proceedings of The Tenth Power Systems Computation Conference. G. Boca Raton. 1990. This model can now be used in capacity expansion planning models to study the effects of DSM under different conditions of weather. 7. R. "Load Duration Curve".02585 14.. July 15-20. et al. M. "Electric Energy Generation: Economics.19 ~ Load Shifting 0. Smith. J. Inc. November 1986. on Power Systems. 17. "Electric Generation Expansion Analysis System". on Power Systems. 0. "Key issues in Integrated Resource Planning for Electric Utilities". 1981. T. 1979. No. Proceedings of the International Symposium on Simulation.2. Jenkins.1.1 and 2. W.W.P. 16. Chan.D. 6.01291 6.W. "Method of Moments Applied to Evaluation of Expected Energy Generation and LOLP for Generation Expansion Planning". 15. EPRI/EA/EM-3597. The VPI model is more convenient and equally accurate (if not more) for expressing the system load curve. IEEE PES Summer Meeting. 1984. and S. Stremel. 10. Vol. Joy.89 Valley Filling 4. Research Project 2381-4. References 1. 12. 1988. R. E. Vardi and B. EPRI EA/EM-3597. April 1989. "Integrating Load Management with Supply-side Operation". as a function of peak load. 13.0. "Demand Side Management". .H.. 1981. Nuclear Technology.T.P. IEEE Trans. M. This results in a more realistic as well as comprehensive reflection of DSM impacts. Stremel and N.. EPRI EM-4815-SR. March 5-6. 1977.T. It can also be used to help the operator in the control center to study the effects of energy sale/purchase on system reliability. Proceedings of IEEE. "Determine of Reliability Worth for Power System Design Application".H. Proceedings of Electric Generating System Expansion Analysis Conference.C. Avi-Itzhak.F. This means that the ILDC does not have to be calculated by inverting the LDC. October 1986 and an Update September 1990. "The Application of the Methods of Moments in WASP-11".0.4. V01.T.03186 17. July 1974. Schenk.77. 3. "Use of Fourier Series in Power System Probabilistic Simulation". provides a straight forward means to study the effects of demand side management. and C. Massachusetts. "Handbook of Mathematical Sciences".F. ACTA PRESS. Summary and Conclusions The VPI model.1223 Table 6. August 1987.70 7873. Rau.J. EPRIReport. Snyder.3. Wacker. Vo1. et al. A. Reliability and Rates". IEEE Trans. Goldman. IEEE Trans. et al. Comparative Analysis of LOLP and ENS under Different Scenarios Base Load Peak Load Energy DSM Activity No 3209.24.. "Impact of Demand Side Management on Future Customer Electric Demand". Fleck. "Practical Methods to Integrate Load Management into Noma1 Operations of Power System Control Centers". June 1979. "Integrating Demand Side Management into Utility Planning". 2. Vorce. November 1974. J.22 Peak CIipping 0. 5. et al. Beyer. Nilsson. 8. 14. P.S. No.5. Modelling and Decision in Energy Systems. 11. This model is not only related to the peak load. Jenkins and T. . 18. Edvinsson. Proceedings of EGEAS conference.60 4. "Compensation for Cumulant Load Fit Discrepancies-A Computer Graphics Approach". Publishers. Feb. Hill and R. Jenkins and D. Florida. and program penetrations. Gellings and W. EPRI EL-4927.S. K.M. 9. which could create errors.45 LOLP 5.O. but also to the base load and the energy. March 5-6. The MZT Press Cambridge.01622 7.45 3977202. "The Cumulant Method of Calculating LOLP".. "Sensitivity of Cumulant Method of Calculating Generation System Reliability". The model is already in the form of ILDC. J. This is because: One does not have to deal with the hourly load data every time a change is effected in the load shape. on Power Systems. 1981. No.6.J. Gellings. August 1984. RPI 1529-2. Austria..

we get: The shape of the ILDC curve generated by the VPI model is as shown in figure Al.7) into eqn.(A.~CB ) [ ( 1 + =B) 1 fl(x) dx - 7 fZ(x) dx] . The VPI Model .. B P P B B Referring to figure A l .(A. P f(x) dx = exp(. and C = a variable as a function of energy. referring to eqn.1224 Appendix A so.5) B B B where.(A..l). (A. The mathematical expression of the model is shown in eqn.(A.. B B P-B The next step is to derive a formula for variable 'C' and find the relationship of 'C' to peak load. (A.2) f(x) dx B Let us define: By substituting eqns. where: B = base load..a). area under the curve f(x).. we get: Load Figure A l ..S). A(r). I). we have: and. is: P A(r) = ...6) and (A. P = peak load. base load and energy. Derivation of a n expression for the variable C Referring to equation (A.l.

.(A. M-78. His areas of interest are demdnd side management..c 2 c2 c 2 ! .. power system where: a m = (m+2)! ... He serves on the System Planning and Demand Side Management subcommittees. Rinaldy was with the Cement and Oil industries in Indonesia from 1981 to 1984 and was involved with various design aspects of substation...D.14) where: A1.. we find: From eqm(A.... . degree in Electrical Engineering...c ......+2! 3! 4! n! Saifur Rahman has taught in the Department of Electrical = CO-2 Engineering. From equation (A.(A.. total hour. demand side management and environmental aspects in electrical industry.1225 = (P-B)f( C+1) + ex where : .17) in to the eqn.(A.l). .2.. the Bangladesh University of Engineering and . and the Long Range 0.1 L+L+-+(-+-+-+..3..... . Since January 1990 he has been a Ph.. His industrial experience includes work at the Brookhaven that the incremental changes are insignificant. we get : 7 exp(C) = .+x).8) we can see that $$is the function of C.+- e$- . and energy (E). The total energy is the area under the curve multiplied by total hour. (A...5) = a. His principal areas of interest are.8). because beyond at VPI.S. He has authored more than 150 technical papers and reports in these areas.13) System Planning.. Substituting eqn. The next step is to find out that C as a function of % P-B such as : .(A.(A.. and baseload. We shall use a limit of 7 form.... He also directs the Energy System Research Laboratory Let m = n .11) into eqn... Substituting eqn. degree (1978) is in Electrical Engineering from the Virginia Polytechnic Institute and 1 c c2 CO-2 State University... if we use series reversion for eqn. Now.. SM-83) graduated from the Bangladesh University of Engineering and Technology in 1973 with a B. transmission and distribution engineering... base load (B)..3! 4! New York at Stony Brook in 1975.(A...(AB) Ab(r) A(r) B = = = area at base load = 1 x B.ll) c2 c3 C" ( l + C + z + ~ + n=l Substituting eqn (A.. we find : Suifur Rahman (IEEE S-75..+-$ 1 1 C C 2 CO-2 M. = -+-+-+.. student at Virginia Tech.Sc.. (A. Cm .12) Technology. (A. the Texas A&M University and the Virginia P-B 2! 3! 4! n! Polytechnic Institute and State University where he is a full Professor.A2.A7 are expressed in eqn. His Ph.. area under the curve. Such.. the Load Forecasting and the Photovoltaics E l working groups of the IEEE Power Engineering Society.15) ...9) 2! . we get: 632) c+c'+.16) Rinaldy (S-91) graduate8 from Univetsity of Indonesia in 1981 with a BSc. degree in Electrical Science from the State University of .2.1.(A. and base load. degree in Electrical Engineering.. He obtained his . ..16) into eqm(A. m= 1..17) ..D.... (A.. Beyer (2): National Laboratory. peak load.lO) c = f($) We know that where: E= h= P= B= energy... New York and the Carolina Power and Computer Societies..14) and let n = m.17) we can see that variable 'C' is a function of peak load (P). (A..(A. such as: Energy = [Ab(') + A(r)] x hour . alternative energy system and expert system.(A. 13). He obtained his MS degree in Electrical Engineering from Michigan State University in 1988. Since 1985 he has been a lecturer of electrical engineering at the University of Indonesia in Jakarta.7 planning....

the additonal step of generating ILDC from adjusted LDC can be avoided. The equation given in the discussion is a simplified version of the VPI model (eqn. P A ) : The a u t h o r s have p r e s e n t e d . Chronological load data from six utilities in seven states are used to compare the capability of the VPI and PECO (given in the discussion) models in representingthe total energy under the load duration curve. First of all. VA 24061-01 11) The authors thank Mr. i n my o p i n i o n . Blacksburg. and total energy. but not precisely. Figure A. And why not use the model given by the discusser. i n a l e s s c o m p l i c a t e d manner t h a n t h a t o f e q u a t i o n 1: L. and energy. e n e r g y d i v i d e d by period hours ) However. t h e r e i s no j u s t i f i c a t i o n f o r assuming t h a t t h e p r o p o s e d model c a n a c c u r a t e l y r e p r e s e n t a wide range o f u t i l i t i e s w i t h d i v e r s e l o a d shapes. Flomhart for LOLP Calculation ~ with Conventional Model with VPI Model Saiiur Rahman and Rinaldy (Electrical Engineering. by using the direct method (the VPI model). Thus If DSM studies are to be done using base and peak loads. M a n u s c r i p t r e c e i v e d S e p t e m b e r 1 4 . and it can represent a diverse set of load Curves from different climatological zones in the United States. t h e f o l l o w i n g f o r m c o r r e c t l y u s e s t h e same p a r a m e t e r s .- Jan 89 1 Alabama 4688785 Oct 89 5039859 5039893 5039329 Jan 89 j 10875882 10875881 10873910 Apr 89 10394917 10394919 10393570 (P-(E/h))/((E/h)-B) f(X) = ((P-X)/(P-B)) W h i l e t h e a u t h o r s ' c h o i c e o f shape happened t o h a v e r o u g h l y f i t t h a t o f one V i r g i n i a u t i l i t y . P h i l a d e l p h i a . r UGY ~. Today's e f f i c i e n t i n t e g r a t e d r e s o u r c e p l a n n i n g models allow analysts to rapidly. t h e s e p l a n n i n g models a c h i e v e s u p e r i o r accuracy w i t h l i t t l e increase i n c o m p u t a t i o n a l burden. Results from the PECO model are measurably different. The cumulative amount of energy utilized over a given period of time can still vary under this threevariable representation. I n comparison t o t h e p r o p o s e d model. m i n i m u m l o a d and a v e r a g e l o a d ( i . the base and peak loads will be represented exactly in both VPI and PECO models. Virginia Tech. Here are the resDonses. 1992. If the impacts of DSM on peak load. This extremely accurate representation provided by the VPI model for six different utility systems around the country establishes the fact that this model is generic. M a n u s c r i p t r e c e i v e d J u l y 29. this model will provide a more direct. if the inverted load duration curve (ILDC) representation of the load is needed. While the total actual energy readings vary from 1 to 11 million MWhrs. . these three parameters do n o t t o t a l l y d e f i n e t h e curve. it looses the chronological time information of the load. base load and total energy are to be examined. A l t h o u g h i t may b e a p p r o p r i a t e t o u s e the proposed ILDC to demonstrate basic r e l i a b i l i t y relationships f o r generic load m o d i f i c a t i o n s . Of course. Now the question is. l . the VPI model will provide adequate accuracy for screening purposes.1226 Discussion R . it is better to use a direct mathematicalmodel of ILDC than to get it from the load duration Curve (LDC). because these variables are used directly in these models. 340 Whittemore Hall. This is shown in Table A. represent the actual base and peak loads. though the total amount of energy will remain the same. ILDC is used for convenience and speed. This is because. and whether it is case specific to the location used in the paper. how well does the VPI model represent the ILDC. Thus it is expected that the simplified model will closely. m a t h e m a t i c a l forms w h i c h c o u l d have been selected. i n e q u a t i o n 1. and therefore time saving approach of evaluation. t h e shape o f t h e p r o p o s e d ILDC model has b e e n a r b i t r a r i l y s e l e c t e d by a d o p t i n g t h e f o r m o f T h i s i s o n l y one o f many equation A . and by design. . systematically evaluate individual DSM programs u s i n g a l g o r i t h m s w h i c h f i r s t m o d i f y the chronological system loads and t h e n d e v e l o p ILDC r e p r e s e n t a t i o n s f r o m t h e a d j u s t e d c h r o n o l o g i c a l loads. Noyes for his discussion of the paper. e . The fundamental point made in the paper is. 1992. Noyes ( P h i l a d e l p h i a E l e c t r i c Company. The functional difference between reliability evaluation using conventional practices and the VPI model are shown in the figure A. a g e n e r a l i z e d i n v e r t e d l o a d d u r a t i o n c u r v e (ILDC) w h i c h i s a f u n c t i o n o f o n l y t h r e e p a r a m e t e r s : peak l o a d . F o r example. t h e u s e o f s u c h a s i m p l i f i e d method o f analysis i n the detailed study of specific demand-side management (DSM) programs o n s p e c i f i c u t i l i t y systems w o u l d b e i n a p p r o p r i a t e and p r o n e t o significant errors. Once the chronological information is lost. the results from the VPI model are exactly the same. or within a few MWhrs. 1) given in the paper. and energy. the remaining variables of interest are base and peak loads.