ELECTRICITY CONSUMPTION FORECASTING BASED ON MULTIPLE REGRESSION ANALYSIS

This study was to forecast monthly average of electricity consumption, the response variable, in Trat province, Thailand with stepwise method of multiple linear regression technique using the three particularly explanatory variables; monthly average number of electricity customers, monthly average rainfall and monthly average temperature. Data was divided into training and validation data sets. The first one was used to build the forecast equation then all assumptions of multiple linear regression analysis were testified by Anderson-Darling statistic (AD), Durbin-Watson statistic (DW), Breusch-Pagan statistic (BP) and Variance inflation factor (VIF). The second one was used to validate the performance of prediction with adjusted coefficients of determination, standard error of regression, mean absolute percentage error, mean absolute error and root mean squared error. The results showed that the number of electricity customers and temperature were significant factors to forecast electricity consumption in Trat province. The performance of prediction accuracy successfully displayed the adjusted coefficients of determination of 0.730, standard error of regression of 93,325.10 kWh, mean absolute percentage error of 4.08%, mean absolute error of 92,116.8812 kWh and root mean squared error of 131,836.3008 kWh.