A comparative study of BPNN, RBFNN and ELMAN neural network for short-term electric load forecasting: A case study of Delhi region

Constant tariff scheme produces a large and continuously-changing difference between electricity cost and price. Consequently, the concern of power system planning and economic generation becomes significant. To overcome this problem accurate load forecasting is a field of immense importance. Conventional methods, i.e., Moving Average (MA) and Holt-Winter (HW) methods are inappropriate to forecast in highly non-linear electrical environment, as existing in Delhi region. In this paper, electrical Load (L), Temperature (T), Relative Humidity (RH) and atmospheric Pressure (Pr) of New Delhi, India are analysed and used to develop the load forecasting model. This paper presents the results of an investigation on various Artificial Neural Networks (ANNs), i.e., Back Propagation Neural Network (BPNN), Radial Basis Function Neural Network (RBFNN) and ELMAN Neural Network (ELMNN), together with specified conventional methods, due to non-linear mapping characteristics of electrical load. Day-Type (D) is additionally used as an input parameter to improve the forecasting accuracy. The investigation has shown that the ELMNN is more accurate than other ANN structures and conventional methods.