Comparison of ARIMA and SVM for Short-term Load Forecasting

Abstract

To ensure a stable and reliable operation of a power system network, load forecasting on a short-term basis is very crucial. The two most important requirements of short-term load forecasting are accurate forecasting and speed. It is really necessary to study as well as analyze the load characteristics and to find out the primary factors responsible for obstructing accurate load forecasting. Auto-Regressive Integrated Moving Average (ARIMA) method is most frequently used because it needs only information regarding the historical loads to predict the load and no other assumptions are required to consider. This paper compares the forecasting ability of ARIMA and Support Vector Machines (SVMs) model with the help of the Mean Absolute Percentage Error (MAPE) and Mean Square Error (MSE). After the comparison, it is found that for a non-linear pattern SVM performs better than the ARIMA, whereas ARIMA gives a better show for the approximation of linear type of load.