LOAD FORECASTING USING FUZZY INFERENCE

Abstract

Forecasting is the process of making statements about events whose actual outcomes have not yet been observed. Power companies uses load forecasting technique to anticipate the amount of the power needed to supply the demand, which helps them utilizing the power generation system more efficiently. Therefore it is important to forecast the future load requirement accurately for efficient operation of power system. Some of the approaches of load forecasting are based on similar day selection. Loads of past similar days are used to forecast load of future day. The main goal of this work is to study fuzzy inference system based approach applied after considering various factors which impacts in similar day selection methods used for the short term load forecasting process. This report work analyzes the impact of similar weather and time factors in selecting similar day. Based on the results obtained, the combined approach of maximum-minimum temperature and classification of days in seven unique types we get better results for the forecasting purpose. Fuzzy logic is used to modify the load curves of the selected similar days. The comparison of results obtained in various methods is done by comparing forecast error. The forecast error is the difference between the actual value and the forecast value for the corresponding period. Some of the examples of measuring the aggregate errors are MAE (Mean Absolute Error), MAPE (Mean Absolute percentage Error). I use MAPE to compare results in various approaches.