Research On Short-term Electrical Load Forecasting Based On Grey Theory

Accurate short-term electrical load forecasting is usually in favor of the timely and reliable decision of district electrical load supply and brownout scheme and peaceful industrial produce and social life. It makes the generator units'schedule and maintenance reasonable and economic, keeps operator of electrical grid stable and secure, releases unnecessary capacity of storage and reduces the cost of electricity generation.However, as the electrical load of power system can't be accurately controlled in nature, the best effective method of electrical load forecasting is to research historical data of the load, and explore suitable load forecasting means of the actual system with current situation and information.As a new practical theory, grey forecasting theory is widely used for electrical load forecasting. However, according to different prediction condition of different areas, there isn't a common forecasting model. The general GM (1,1) grey forecasting model has the distinct deviation under the circumstance of considerable data fluctuations. This is against the actual condition. Therefore, to improve the grey forecasting model's accuracy and applicability, this paper presents an optimized combination model based on the analysis of grey correlation coefficient and a multi-segmentation scheme under actual applications. Firstly, GM (1,1) presents good forecasting result within the smoothly upward or downward segment, while daily power load can be divided into several segments with peaks or valleys. Therefore, grey correlation segmentation and optimized combination are carried out in order to avoid the risk in which errors are introduced into the model and then gradually amplified for the inappropriate choice of original condition. Secondly, by choosing different aspects of original number and analyzing the data rules, the model decreases effects of uncertain factors on forecasting. Lastly, the model is brought into several self optimized algorithms and data preprocess in this paper. This optimized gray GM (1,1) forecasting model is proved by applications of the China Southern Power Grid Corporation Limited in GuiGang City of GuangXi Province. The model controls the forecasting average error at the level of about 3%, which greatly improves forecasting accuracy and completely satisfies the requirements of short-term electrical load forecasting in this area.