Accuracy of Electric Power Consumption Forecasts Generated by Alternative Methods: The Case of Hawaii

A number of alternative methods can be used to generate forecasts of electricity consumption. However, it is seldom possible to compare the accuracy of forecasts generated by different statistical methods. Hawaii provided a unique opportunity for such a comparison because of the availability of the Hawaii Energy Demand Forecasting Model (HEDFM). HEDFM is an econometric-based simulation system designed to provide detailed annual consumption forecasts for various fuel types from 1978 to the year 2005 for the State of Hawaii. This article evaluates the accuracy of electricity consumption forecasts using various commonly employed methods and compares them with forecasts from HEDFM. Traditional forecasting methods based on average historical growth rates or the historical relationships between electricity consumption and key economic and demographic variables such as de facto population and per capita income did not perform well. Quadratic exponential smoothing outperformed all other methods. HEDFM, an econometric model, tracked reasonably well. In particular, simulation results have shown that econometric models such as HEDFM can produce accurate forecasts if independent variables can be predicted with a reasonable degree of certainty.