Hourly Load Forecasting of Power System over Northeastern Thailand Using Artificial Neural Network

Thailand has limited resources to generate the electric power. In developing countries, the electrical demand is growing continuously. Power system planning is very important issues. There are three major components of the electrical power system namely, generation, a high voltage transmission line, and a distribution system. The hourly power variables, at Bamnet Narong substation in northeastern Thailand is fed to artificial neural network models for the short-term load forecasting. The training dataset in ANN is from measurement data of Supervisory Control and Data Acquisition (SCADA) system such as true power, reactive power and voltage source. The training methods with optimized weights in the ANNs, Levenberg-Marquart algorithm (LMA) and Gradient Descent algorithm (GDA), are applied. The results have shown that the ANN with LMA is more effective than GDA.