Combined Forecasting Method for Monthly Energy Consumptions based on Customer Types

P. Churueang and J. Chureemart (Thailand)


load classifications, decomposition methods, electric energy forecasting, seasonal ARIMA models.


This paper presents a time-series forecasting technique for monthly energy consumptions of an electric distrib-ution utility in Thailand. The forecasting is performed on individual customer classification, categorized by tar-iff rates. Classical decomposition technique is used as a main forecasting approach for every customer type. An additive or multiplicative decomposition function is cho-sen for each customer classification based on its time se-ries behavior. Trend-cycle components of the decompo-sition for all customer types are projected into the future by searching for parametric functions or seasonal ARIMA models that best describe the trend lines. The seasonality of every customer type is obtained by simple averaging method. The forecast trend-cycle and seasonal components are then put back into the decomposition function to form a forecasting model. It is concluded that different models for trend-cycle components are suitable for differ-ent types of customers resulting in different decomposi-tion forecasting models.

Important Links:

Go Back