Application of fuzzy linear regression to load forecasting

Linear regression analysis is a common method for mid-term load forecasting. Because of the influence of many uncertainty factors, the historical data we collected and the future correlative variable data are often inaccurate; this method often contains notable error. In order to improve the accuracy of load forecasting, an improved fuzzy linear regression model, which named as weighted fuzzy linear regression (WFLR), is proposed in this paper. Linear programming is used to solve the fuzzy linear regression problem. The fuzzy linear regression (FLR) model is improved in two aspects. The weight of each term in the object function is determined according to the importance of each regression variable. The relevant value is determined according to the importance of each historical datum. The weighted fuzzy linear regression model is adjustable. It can take some qualitative and fuzzy factors into account easily. Practical calculation shows that the forecasting precision is improved with the method mentioned above. (6 pages)