Support Vector Machine Regression for forecasting electricity demand for large commercial buildings by using kernel parameter and storage effect

In the framework of a competitive commercial world, having accurate energy forecasting tools becomes a Key Performance Indicator (KPI) to the building owners. Energy forecasting plays a crucial role for any building when it undergoes the retrofitting works in order to maximize the benefits and utilities. This paper provides accurate and efficient energy forecasting tool based on Support Vector Machine Regression (SVMR). Results and discussions from real-world case studies of commercial buildings of Colombo, Sri Lanka are presented. In the case study, four commercial buildings are randomly selected and the models are developed and tested using monthly landlord utility bills. Careful analysis of available data reveals the most influential parameters to the model and these are as follows: mean outdoor dry-bulb temperature (T), solar radiation (SR) and relative humidity (RH). Selection of the kernel with radial basis function (RBF) is based on stepwise searching method to investigate the performance of SVM with respect to the three parameters such as C, γ and ε. The results showed that the structure of the training set has significant effect to the accuracy of the prediction. The analysis of the experimental results reveals that all the forecasting models give an acceptable result for all four commercials buildings with low coefficient of variance with a low percentage error.