Study On Fuzzy Method For Long-Term Load Forecasting

The author researches the principle and methods applying fuzzy theories and modern optimizing technology to power load forecast. Combining fuzzy theory with the traditional regression method, the fuzzy linear regression model for long-term power load forecasting has been set up in this paper. The author uses social economic indexes such as population, GDP, gross output value of main national economy trade, etc. as the independent variables, and uses quantity of electricity as the variable, considering the effect of every social economic index on the demand of quantity of electricity. As a result, the author rejects the weak relevant independent variables in model with Stepwise regression method , revises the strange historical data with Ridge regression method, then uses Least Square method to solve regression coefficients. Investigating the characteristic of the array of time of historical data, it is clear that the degree of impact on model of different-year historical data shows large difference. Thus, the linear subjection function cumulating along with the array of year in the fuzzy center is adopt to the fuzzy independent variables. In order to stress the important position of the independent variables which are relevant to quantity of electricity strongly ,the author introduces regression weight coefficient to revises the routine fuzzy linear regression model , and presents the improved fuzzy linear regression model . On the base of above methods, the intact load forecasting program is written and applied to real engineering load forecast. The practical application indicates that the improved fuzzy linear regression model is superior to the routine fuzzy linear regression model. It has reduced human errors caused by artificial insufficient experience, and improved accuracy and practicability in load forecast. Studying further, the different relevance exists heavily among the real data to forecasting value of generalized linear regression method, the real data to forecasting value of Stepwise regression method and the real data to forecasting value of Ridge regression method. In view of the above the author establishes a kind of integrating optimized fuzzy forecasting model which is the optimizing combination of variety of forecasting models on fuzzy regression theory and think about variety of developing direction of forecasted target and constraints, it imitate more accurately than some above forecasting models on fuzzy theory.In this paper, the author resolves uncertainty of social economic indexes influencing load-forecasted value preferably. The improved fuzzy linear regression model and integrating optimized fuzzy forecasting model in the thesis have instructional significance to load forecasting problems. The real application of Shenzhen year-load forecasting proves that the result of the paper is accurate and practicable in projects preferably.