AN ARTIFICIAL NEURAL NETWORK MODEL FOR ELECTRICAL DAILY PEAK LOAD FORECASTING WITH AN ADJUSTMENT FOR HOLIDAYS

Abstract

Short-term load forecasting (STLF) is of a great importance to power generation utilities. Decisions relevant to economical operation, unit commitment, and security assessment and enhancement rely on STLF. This thesis presents a neural network model for daily peak load forecasting. The input variables of the model have been selected using the correlation coefficient methodology. So the model uses.only three input variables. The minimum number of input variables reported in the literature is five. In addition, a new technique for . selecting the training vectors is introduced. Moreover, The model presents a unique adjustment algorithm to compensate the negative impact of holidays' forecasts. Also, the model uses an adjustment technique for Sundays and Mondays forecasts as these showed higher error than the rest of weekdays. Nevertheless, the model is simple, fast, and accurate. The mean percent relative error of the model over a period of one year is 2.066% including holidays.

Keywords: Short-term load forecasting, Daily Peak Load Forecasting, Artificial Neural

Networks, Holidays Forecast.