An approach towards medium term forecasting based on support vector regression

Abstract

This paper presents Medium term load forecasting for hourly electric loads prediction of the peak loads of different days or hourly loads throughout the day. Most of the researchers concentrate on Short Term Load Forecasting, there are less researches are going on in Medium term load forecasting. This paper presents a Support Vector Regression (SVR) model for Medium Term Load Forecasting and also it compares the results obtained from SVR Model with ANN model. The month wise hourly electrical load is predicted by using Support Vector Regression model and ANN model, the result shows that SVR model have great potential towards MLTF. In literature most of the models used for MLTF are ANN based. The Neural Network Model is providing comparatively better results, but it requires relatively large amount of training data for learning the data pattern. Therefore the memory requirement and the processing time are quite high; due to these problems, the real time application is not possible with these models. So more accurate model having negligible processing time is required, this leads to Support Vector Regression model.