Study On The Model Of Medium-Long Term Load Forecasting For Power System

Load forecasting for power system is one of the important tasks of power utilities. It usually can be divided into ultra-short term load forecasting, short term load forecasting and medium-long term load forecasting. Medium-long term load forecasting, which is judged by year, is the important basis of power scheme, production and operation. Unflattering load forecasting can boost the security and stability of power system, effectively reducing the cost of generating electricity, ensure power requirement, reinforce power reliability, and increasing the economical and social benefit of power system.First of all it discusses the purposes and development dynamic of power medium-long term load forecasting, and the comparison of all kinds of methods which are used in load forecasting. Furthermore, it introduces the generality, basic principle, basic process and error analysis of load forecasting.Next, it simply analyses load characteristics of some area, in this case, it can hold the rule and development trend of load characteristics, and it can produce effective conditions for posterior load forecasting process. Otherwise, it introduces principle and process of regression forecasting, grey forecasting and optimization combination forecasting in detail. It mainly introduces monobasic quadrate regression model, exponential regression model and multicomponent regression model, and then calculates with monobasic quadrate regression model and exponential regression model; it propounds grey residual model for the limitation and research of grey model. Through residual analysis for GM(1,1) model, it can boost forecasting accuracy of grey model, With new information and equal dimensional GM(1,1) model residual analysis, it can ensure new information available, it explores modeling process of GM(1,2); and introduces fixed weigh method of optimizationcombination forecasting model in detail------that is to choose weigh withaverage precision of each model.Then, it makes use of monobasic quadrate regression model, exponentialregression model, grey GM(1,1) model, new information and equaldimensional and improved grey residual forecasting model to forecastmedium-long term load combining with historical load datas year 2000 to2005 of some area, and chooses weigh with average precision of each model,that is fixed weigh method of optimization combination forecasting model toforecasting, and then compares with forecasting results of single model. As aresult, it indicates that the accuracy of optimization combination forecastingmodel is higher than single model, it is suitable for medium-long term loadforecasting. Moreover it establishments program for computing process, in this case, it can boost forecasting speed, then it forecasts load and maximum load of some area from year 2006 to 2010 with above methods to proffer condition for load management of some area.Finally, it generalizes all of works, brings forth a lot of new ways to improve load forecasting and propounds the way to develop load forecasting software, in this way, it is favor to calculate for load model, and it can make others see the forecasting results clearly, give user a kind of visual and clear effects.