Hybrid short term load forecasting using ARIMA-SVM

In order to perform a stable and reliable operation of the power system network, short term load forecasting is vital. High forecasting accuracy and speed are the two most important requirements of short-term load forecasting. It is important to analyze the load characteristics and to identify the main factors affecting the load. ARIMA method is most commonly used, as it predict the load purely based on the historical loads and no other assumptions are considered. Therefore there is a need for Outlier detection and correction method as the prediction is based on historical data, the historical data may contain some abnormal or missing values called outliers. Also the load demand is influenced by several other external factors such as temperature, day of the week etc., the Artificial Intelligence techniques will incorporate these external factors which improves the accuracy further. In this paper a hybrid model ARIMA - SVM is used to predict the hourly demand. ARIMA is used to predict the demand after correcting the outliers using Percentage Error (PE) method and its deviation is corrected using SVM. Main objective of this method is to reduce the Mean Absolute percentage Error (MAPE) by introducing a hybrid method employing with outlier detection technique. The historical load data of 2014-2015 from a utility system of southern region is taken for the study. It is observed that the MAPE error got reduced and its convergence speed increased.