A Multi-Parameter Regression Based Time-Series Load Forecasting Model With Fuzzy Ideology

Barnali Goswami, Department of Electrical Engineering, Assam Engineering College; goswamibarnali@yahoo.co.in

Rashmi Rekha Borgohain, Department of Electrical Engineering, Assam Engineering College; rashee.225@gmail.com

Abstract- Load Forecasting is an essential task which finds extensive use in power system planning and decision making. It has become important for the electric industry in deregulated environment nowadays. A suitable time series based Auto-Regressive (AR) model is proposed wherein the temperature effects is included for the purpose of accurate forecasting. Fuzzy Load
forecasting plays an important role in the operation and management of power systems.

The results are therefore checked using real time data of State Load Dispatch Centre (SLDC) and the final result is improved by the use of fuzzy ideology. Index terms- forecasting, deregulated environment, Auto-regressive, SLDC, fuzzy ideology

I. INTRODUCTION The electric power utilities face many economical and technical problems in operation, planning, and control of an electric energy system in order to supply high quality electric energy to the customer in a secure and economic manner. For the purpose of optimal planning and operation of the power system, modern optimization techniques are being applied with the expectation of considerable cost savings. In achieving this goal, the knowledge of future power system load is the first prerequisite. Load forecasting is a process of predicting the future load demands. It is an important component for energy management system. Precise load forecasting helps the electric utility to make unit commitment decisions including decisions on purchasing and generating electric power, load switching and infrastructure development. Besides playing a key role in reducing the generation cost, it is also essential to the reliability of power systems. Short term load

forecasting (STLF) plays an essential role for electric power system planning. It is important for power system planners and demand controllers in ensuring that there would be enough generation to meet the demand. Load serving entities use load forecasts for system security, to schedule generator maintenance and to plan the most cost-effective merit order dispatch.. Short-term load forecasting (STLF) can help to estimate load flows and to make decisions that can prevent overloading. Timely implementations of such decisions lead to the improvement of network reliability and to the reduced occurrences of equipment failures and blackouts. Load forecasting is also important for contract evaluations and evaluations of various sophisticated financial products on energy pricing offered by the market. Over the last decade, as electricity markets have deregulated, the importance of load forecast accuracy has become even more evident. Without an optimal load forecast, utilities are subject to the risk of over- or under- purchasing in the day-ahead market. Short Term Load Forecasting (STLF) refers to forecasts of electricity demand (or load), on an hourly basis, from one to several days ahead. The amount of excess electricity production (or spinning reserve) required to guarantee supply in the event of an underestimation, is determined by the accuracy of these forecasts .Conversely ,overestimation of the load leads to sub-optimal scheduling (in terms of production costs) of power plants (known as unit commitment).As electricity demand is closely influenced by these climatic parameters, there is likely to be an impact on demand patterns. Every hour load demand depends on these crucial role parameters. Load forecasting can be divided into three major categories: Long-term electric load forecasting(LTLF) considers load peaks and consumed energy, on a yearly basis, for several years ahead. For example, long-term forecasting is relevant for the planning of new electricity utilities, and inaccurate predictions have important financial cost used to supply electric utility company

management with prediction of future needs for expansion, equipment purchases, or staff hiring Medium term load forecasting (MTLF) covers the horizon from one month up to a few years ahead. used for the purpose of scheduling fuel supplies and unit maintenance Short-term forecasting, used to supply necessary information for the system management of day-to-day operations and unit commitment. In this paper, the work presents a study of short term hourly load forecasting with the help of using previous two days data and the prior electricity demand data are treated as time series. The proposed model is processed using historical data.

other factors such as weather, day type, and customer class. 2) Time Series- Time series methods are based on the assumption that the data have an internal structure, such as autocorrelation, trend, or seasonal variation. Time series forecasting methods detect and explore such a structure. Time series have been used for decades in such fields as economics, digital signal processing, as well as electric load forecasting. Time series method is a basic method to process random sequence, and is widely applied in STLF for power system. In this method, power load data is viewed as a fluctuant data sequence with time point, considering the causal relationship between power load and temperature. Auto regressive model (AR model), moving average model (MA model) and auto regressive & moving average model (ARMA model) are three main models. ARMA (autoregressive moving average), ARIMA (autoregressive integrated moving average), ARMAX (autoregressive moving average with exogenous variables), and ARIMAX (autoregressive integrated moving average with exogenous variables) are the most often used classical time series methods. ARMA models are usually used for stationary processes while ARIMA is an extension of ARMA to non-stationary processes. ARMA and ARIMA use the time and load as the only input parameters. Since load generally depends on the weather and time of the day, ARIMAX is the most natural tool for load forecasting among the classical time series models. . Considering the calculating velocity in STLF, AR model is selected in the work 3) Fuzzy Logic-Fuzzy logic is a generalization of the usual Boolean logic used for digital circuit design. An input under Boolean logic takes on a truth value of 0 or 1. Under fuzz y logic an input has associated with it a certain qualitative ranges. For instance a transformer load may be small, medium and high. Fuzzy logic allows one to (logically) deduce outputs from fuzzy inputs. In this sense fuzzy logic is one of a number of techniques for mapping inputs to outputs (i.e. curve fitting). Among the advantages of fuzzy logic are the absence of a need for a mathematical model mapping inputs to outputs and the absence of a need for precise (or even noise free) inputs. With such generic conditioning rules, properly designed fuzzy logic systems can be very robust when used for forecasting. Of course in many situations an exact output (e.g. the precise 12PM load) is needed. After the logical processing of fuzzy inputs, a defuzzification process can be used to produce such precise outputs. Fuzzy logic is therefore used in

II. LOAD FORECASTING TECHNIQUES Load forecasting has become one of the most significant aspects of electric utility planning. The economic consequences of improved load forecasting approaches have kept development of alternate, more accurate algorithms at the forefront of electric power research. Thus the significance of the subject in power systems has drawn alarming interests of many researchers, and to date a number of load forecasting approaches have been developed Generally, load forecasting models can be classified into two categories: time-of-day models and dynamic models. Time-of-day model is a non-dynamic approach and expresses the load at once as discrete time series consisting of predicated values for each hour of the forecasting period. The second classification involves the dynamic model that recognizes the fact that the load is not only a function of the time of the day, but also of the load most recent behavior. The various methods used for load forecasting are similar day approach, regression models, time series, neural networks, expert systems, fuzzy logic, statistical learning algorithms etc. and their classification is in terms of their degrees of mathematical analysis used in the forecasting model. A. Short-term Load Forecasting Methods A large variety of statistical and artificial intelligence techniques have been developed for short-term load forecasting. Some of them which finds use in this paper are as follows1) Regression Methods- Regression is the one of most widely used statistical techniques. For electric load forecasting regression methods are usually used to model the relationship of load consumption and

this paper to improve the result obtained from time series method. III. TIME SERIES MODEL Time series can be defined as a sequential set of data measured over time, such as the hourly, daily or weekly peak load. The basic idea of forecasting is to first build a pattern matching available data as accurate as possible, then obtains the forecasted value with respect to time using established model. Generally, series are often described as having following characteristic:

U1, U2 are model(AR) parameter.. ak = error term. The model parameter U1 and U2 are calculated by method of least square. For a set of linear equation, AU = B Where, A = k*n matrix UT = [U1 ,U2,..Un], the unknown vector. BT =[b1,b2,.b n], the measurement vector

A= [Y2 Y3

Y1 Y2

T2 T3

T1 T2

X (t ) T (t ) S (t ) R(t )
(1)

t ... 1,0,1,2,...

Y365 Y364

T365 T364]365*4

Here, T(t) is the trend term, S(t) the seasonal term, and R(t) is the irregular or random component. At this moment, we do not consider the cyclic terms since these fluctuations can have duration from two to ten years or even longer which is not applicable to short-term load forecasting .We have such assumptions to make things a little easier for the moment: 1) The trend is a constant level; 2) The seasonal effect has period s, that is, it repeats after s time periods. Or the sum of the seasonal components over a complete cycle or period is zero.

And, B= [Y3 Y4 Y366]365*1

The estimated value of U is given by, U= ( ATA)-1* AT *B

U= [U1 U2 U3 U4] Y366= U1Y365+U2Y364+U3T365+U4T364

predicted. Thus, the load of the 366th day can be

S (t j) 0
j 1

(2)

A) Forecasting Evaluation Method Till now, many forecasting methods have been utilized which are classified into two basic types: qualitative and quantitative methods. Qualitative forecasting methods generally use the opinions of experts to predict future load subjectively. Such methods are useful when historical data are not available or scarce. These methods include: subjective curve fitting, Delphi method, and technological comparisons. Quantitative methods include: regression analysis, decomposition methods, exponential smoothing, and the Box-Jenkins methodology. 1) Formation of AR(2) ModelFor AR (2) process, the transfer function is given by,

2) Algorithm for the regression based time series model with Fuzzy ideology The algorithm for the model can be shown in the flow chart as followsSTART

Read Load data and Temperature Form the AR (2) model

Y k= {1/ (1-U1Z-1-U2Z-2)}*a k Hence, Y k=U1Yk-1+U2Yk-2+a k,

Where, Y k= k th observation.

Determine the forecasted load

k=3, 4,..N
Is error1<=3%?,if yes stop else go to next step

STOP

Fuzzy model to determine error in forecast load of AR model and actual load Is error<=3% time-scale of the available electrical demand data and temperature data is given in Table1 and Table2 respectively.. TABLE I DATA TIME-SCALE AND RANGE Range 01/01/2012-24/12/2012 Time of scale Hourly No of data points 8616(359 days)

STOP

3) Forecasting errors Unfortunately, all forecasting situations involve some degree of uncertainty which makes the errors be unavoidable.

The forecast error for a particular forecast X t

with respect to actual value

TABLE II WEATHER DATA TIME-SCALE AND RANGE Range 01/01/2012-24/12/2012 Time of scale No of data points Hourly 8616(359 days)

X t is:

et X t X t
To avoid the offset of positive with negative errors, we need to use the absolute deviations.

| | et || X t X t

(4)

Hence, we can define a measure known as the mean absolute deviation (MAD) as follows:

MAD

e
t 1

X
t 1

X t
(5)

Again, Mean Absolute Percentage Error (MAPE) is defined as: MAPE=MAD*100% (6) Another method is to use the mean squared error (MSE) defined as follows:

V.RESULTS & DISCUSSION The method is applied for short term load forecasting for the 359 days time span in order to forecast the last week i.e.52nd week of the year 2012 and it is implemented in MATLAB and graphs are plotted with number of samples representing hourly duration for the whole day in the x axis and corresponding actual load and forecast load on y axis. From the Fig.1, it is observed that forecast load is very close to actual load and the calculated % error was found to be negligible and within limits. The curve between the predicted and actual load for the last day of 2012 i.e. 31st December,2012 is shown in the figure 1. The red line indicates actual load and the blue line indicates predicted load demand.

MSE

et
t 1

(X
t 1

)2 X t
(7) 900 800 700 600 500 400 300 200 100 0 1 3 5 7 9

load(MW)

Where, n is the number of predicted values. The RMSE is the square root of the MSE and is a measure of the forecasting accuracy. IV. METHODOLOGY A)Data Set Details The data used are fifty one weeks and two days i.e. 359 days hourly load demand measured in Megawatt (MW) from January 01, 2012 to December 24, 2012. They are gathered from State Load Dispatch Centre (SLDC), Assam, India .Also .the temperature data in Celsius for the same period as load data are available from the Regional Meteorological department t(RMC) located at the Lokpriya Gopinath Bordoloi International Airport,Borjhar, Assam . The range and

Series1 Series2

11 13 15 17 19 21 23

hours Figure 1

VI.CONCLUSION The load is predicted for a week but results here in this paper are shown for only the last day of the week considered. The actual demand and predicted demand for 24 hours on 31st December, 2012 are as follows: Actual Demand (MW) 700 680 700 690 680 700 740 741.2 720 700 680 650 640 682 720 790 800 780 750 850 820 800 780 800 Predicted Demand(With Fuzzy)(MW) 736.8907 735.0819 725.2348 720.9018 712.9607 712.0575 721.8476 731.1098 715.6618 709.9324 706.5421 697.6981 685.6432 687.6017 677.8159 691.0655 701.5211 716.7639 731.8449 735.8389 726.8605 722.0343 723.3885 720.0855 Predicted Demand(Without Fuzzy)(MW) 721.8907 720.0819 710.2348 705.9018 697.9607 697.0575 706.8476 716.1098 700.6618 694.9324 691.5421 682.6981 670.6432 672.6017 662.8159 676.0655 686.5211 701.7639 716.8449 720.8389 711.8605 707.0343 708.3885 705.0855 Accurate load forecasting is very important for electric utilities in a competitive environment created by the electric industry de-regulation. Accuracy of prediction is improvised through sending the forecasted values obtained from the AR(2) model again to a fuzzy inference system .Its forecasting reliabilities were evaluated by computing the mean absolute percentage error (MAPE) between the actual and predicted values. Here, for the 366th day of 2012 i.e. for 31st December we were able to obtain MAPE of 4.09% and RMSE of 0.0409 without fuzzy and MAPE of 2.03% and RMSE of 0.0204 with Fuzzy which represents a high degree of accuracy. The forecasting results for the whole week starting from 25th December to 31st December are found to have MAPE within limits The results suggest that Regression based time series model with the fuzzy ideology can perform good prediction with least error ACKNOWLEDGEMENT The authors are grateful to SLDC, Kahilipara (Guwahati) for providing the load data and also extend their gratitude to the Regional Meteorological Centre(RMC),Guwahati for providing us the temperature data.. The authors also thank the Department of Electrical Engineering, Assam Engineering College for the use of their resources. Finally, the authors are indebted to numerous people who assisted in making constructive suggestions which have improved the quality of this paper. REFERENCES [1]Moghram and S.Rahman, Analysis and evaluation of five short term load forecasting techniques, IEEE Trans. Power Syst., vol. 4, no. 4, pp. 14841491, Nov. 1989. [2] Mrs. J. P. Rothe, Dr. A. K. Wadhwani and Mrs. S. Wadhwani,Short Term Load Forecasting Using Multi Parameter Regression,(IJCSIS) International Journal of Computer Science and Information Security,Vol. 6, No. 2, 2009 [3] A. D. Papalexopoulos and T. C. Hesterberg, A regression-based approach to short-term load forecasting, IEEE Trans. Power Syst., vol.5, no. 4, pp. 15351550, Nov. 1990. [4] B.L.Bowerman, R.T.O'Connell, and A.B.Koehler, "Forecasting, Time Series, and Regression: An

Applied Approach," 4th ed. California: Thomson Brooks/Cole, 2005. [5] D.C.Montgomery, L.A.Johnson, J.S.Gardiner, "Forecasting & Time Series Analysis," 2th ed. New York: McGraw-Hill, 1990.IEEE Transactions on Power Systems, Vol. 8, No. 1, February 1993 [4]J.H.Chow,F.F.Wu,J.A.Momoh,Applied Mathematics for restructured electric power systems, New York: Springer, 2005 [5]G.Janacek, L.Swift, Time series: forecasting, simulation, applications, West Sussex: Ellis Horwood Limited, 1993. [6]I. Moghram and S.Rahman, "Analysis and evaluation of five short-term load forecasting techniques," IEEE Trans. on Power Systems, vol.4, no.4, pp. 1484-1491, Oct.1989. [7]J.Y. Fan and J.D. McDonald. A Real-Time Implementation of Short-Term Load Forecasting for Distribution Power Systems. IEEE Transactions on Power Systems, vol.9, no.2, pp.988-994, 1994. [8] J. H. Park, Y. M. Park and K. Y. Lee, Composite modeling for adaptive short term load forecasting, IEEE PES Summer Meeting, # 90 SM 378-0 PWRS, Jul. 1990. [9] M. T. Hagan and S. M. Behr, The time series approach to short-term load forecasting, IEEE Trans. Power Syst., vol. PWRS-2, no. 3, pp. 78579, Aug. 1987 [10] Lefteri H.Tsoukalas, Robert E.Uhrig, Fuzzy and Neural Approaches in Engineering, A wiley-Interscience publication, JohnWilley&Sons INC,1997 [11] Lefteri H.Tsoukalas, Robert E.Uhrig, Fuzzy and Neural Approaches in Engineering, A wileyInterscience publication, JohnWilley&Sons INC, 1997.. [12] R.A. Aliev, R.R.Aliev, Soft Comput ing and its Application, World scientific publication, 2001. [13] Tsekouras, G.J.,E.N. Dialynas, N.D. Hatziargyriou and S. Kavatza, 2007. A non-linear multivariable regression model for midterm energy of power systems. Electric Power Systems Res., 77: 1560-1568. [14] Wei, W.W.S., 2006. Time Series Analysis, Univariate and multivariate Methods, , Second Edition. Pearson Addison Wesley, New York, pp: 72166. [15] Taylor, J.W., 2008. An evaluation of methods for very short-term load forecasting using minute-byminute British data. International J. Forecasting 24: 645-658.