Short Term Load Forecasting Using Artificial Neural Network

This paper presents our research work on the Short Term Load Forecasting(STLF) Using Artificial Neural Networks.Short term load forecasting (STLF) with lead time from a few minutes to sevendays, play a key role for the ecouomic and secure operation of power systems.Various models for STLF have been proposcd in the last few decades. With thedevelopment of ANN, a new method of the STLF has been proposed. The forecastmodel of ANN bas been successfully applied to STLE ANN does not rely on humanexperience but attempt to learn by itself the functional relationship between systeminputs and outputs through a training process. Initially the ANN is trained by historydata. After training, the ANN can be used to forecast the load of power system onlywith inputs.This paper describes the experiences we gained during the research of loadforecasting based ANN tecbnique. The details are as the following:1. The training of an ANN is the most important step before it works. We try to finda better way to train ANN.2. In the process of rescarch we find that using difference step for each unit of ANNcan work more emciently3. Using this method, we create an ANN model for STLF4. To train the ANN, we use a special technique-interrelation factor method to formand select history data. In this way we get the faster training speed. The1996～1998 history data of TianJin China was used to be tested fOr forecasting, andthis forecast method is proved to be feasible.5. We also use the ANN to forecast Electricity Price, and we have got somesatisfactory results.6. We have done some research about the problems that appcared in process of usingANN, and we have gained lots of important expericnces of ANN and STLF...