Short term load forecasting by using neural network structure

Load forecasting has an extraordinary important role in planning and operations of power systems. Since the beginning of the electrical industries, load forecasting has received special attention and different methods have been presented on this subject. In this paper, a practical load forecasting method for load forecasting in Khorasan province electricity market in the time limit between March 2004 to July 2008 is presented. In the proposed method, a multilayer perceptron neural network is trained with the obtained data. The program considered, has been written in visual basic language in the excel environment in which excel environment has been used as an information bank data base. According to the high volume of the information needed for training the neural network, this stage would be a time consuming task. Therefore, the MATLAB environment has been used for fast execution and at the same time accurate forecasting of the load. Finally, the accuracy of the structure considered has been forecasted and tested .The results show that the maximum error resulting from the network real data at July 2008 has been about 4.94%.