Soft computing application for electrical short term load forecasting

Load forecasting is an important element for power system planning, scheduling, economic dispatch, unit commitment, and maintenance. In practice electrical load is having nonlinear behaviour. Soft computing methods have the ability to give better performance in dealing with the nonlinearity. Soft computing techniques have been successfully applied to a number of scientific and engineering problems during recent years. In this work, one of the soft computing techniques, fuzzy logic, is chosen for electrical short term load forecasting. Fuzzy logic has the ability to model the uncertain and ambiguous data. The similar days of the forecasting day are found from the historic data. The preferred similarities are based on a criterion that uses temperature, humidity and day type variables. Then fuzzy logic is applied for the correction of these similar days that gives good results for short term load forecasting with less Mean Absolute Percentage Error (MAPE). The solution technique proposed in this work has been applied on a real data set and results are presented. The results achieved with utilization of soft computing technique are very good in quality with less than 3% of MAPE. The paper also presents a comparison of the forecasted results achieved with and without using the soft computing technique and shows that soft computing application improves the quality of forecasted results.