An Efficient Approach for Short Term Load Forecasting

Short term load forecasting (STLF), which aims to predict system load over an internal of one day or one week, plays a crucial role in the control and scheduling operations of a power system. Most existing techniques on short term load forecasting try to improve the performance by selecting different prediction models. However, the performance also rely heavily on the quality of training data. This paper proposes a novel short term load forecasting approach based on training data selection. There are two main characteristics of the proposed method that distinguish it from the previous studies. The first characteristic is that the load curve of a time interval before the target hour is regard as the benchmark of training data instead of the cluster center of all historical data used in previous studies. The second characteristic is the load curves are normalized for comparison to the benchmark. Thus the load curves having similar trend with the benchmark can be selected for training. Experiments conducted on the real load data show significant improvement over the baseline method.