Short-term load forecasting by using artificial neural networks

Abstract

Load forecasting is a vital component for power systems. For any electric power organization, it is their main role to provide electric energy in an economical and secured manner maintaining the quality. In power systems, electric load forecasting is a very essential issue and has been studied widely so that to attain more precise load forecasting results. Since in power systems the next day's power generation must be scheduled every day, day-ahead short-term load forecasting (STLF) is an important daily task for power companies. The short-term load forecast represents the electric load forecast for a time span of few hours to a several days. This thesis uses the method of Artificial Neural Networks (ANN) to create a STLF model for Faisalabad Electric Supply Cooperation (FESCO). FESCO can be described as an institutional/industrial-type electric load. The ANN is a mathematical/computational tool that mimics the way human brain processes information. Several types of ANNs are revealed in this research, among them Feed-Forward (FF) neural networks have been used to create a STLF model. Two different methods ReliefF and Correlation analysis are presented and used for the selection of important input variables which will be used as inputs of ANN. The inputs given to the ANN are historical electric load and average air temperature data. Correlation method provides better results for the selection of input variables rather than the ReliefF method. The ANN uses the historical electric load and weather data as an input and provides a forecasted electric load at its output. Average air temperature data is used in this research in order to achieve better short-term load forecasting results. Furthermost ANNs in the literature are used to forecast day ahead electric load for a transmission-level system with resulting load forecast errors ranging from nearly 0.1% to 2.3%. This research indicates that an ANN can be used to forecast the smaller, more disordered load profile of an institutional/industrial-type power system and results in a similar forecast error range. In addition, the in-service constraints of the FESCO electric load will be investigated along with the weather profiles for the site.