Of Ann-based Short Term Load Forecasting Techniques

Load forecasting technology is a new field in which many countries research with great concentration in recent few years. Load forecasting technology plays an important role not only in the design and running in power system but also in the increase of economical benefit. Short-term Load prediction based on artificial neural network is a common but most efficacious method. So some forecasting algorithms attached to ANN begin to be a promising and important field in the development of prediction technology.The paper primarily explicated some algorithms about prediction in EMS. Firstly, the background and development of prediction technology are introduced and then some introduction of basic theory and research work have been done about how to apply ANN to prediction technology, during which BP network and RBF network are introduced importantly and then some improvement about the application of ANN to prediction technology is given. With an example the paper explicitly discusses the application of BP network in load prediction and has a deep research in pattern division of inputting, the selection on the number of the hidden layer, the modifying of weight, the adjustment of the speed of the study and etc. In the paper, a new and sufficient method about the selection of the training sample is proposed and also the division of inputting in festivals is operated with a new method by using interpolation. Besides, in the paper, the longest predicable time is studied theoretically and practically operated. The compare of the two kinds of network and their respective privilege and limitations is the emphasis in the paper.The paper cites chaos theory to predict technology. In the end the paper give some prospects and hypothesis on the prediction.