Applications of noniterative least absolute value estimation for forecasting annual peak electric power demand

A noniterative least absolute value (LAV) technique for estimating the parameters of a selected electric load forecasting model is utilized. The selected forecasting model with the estimated parameters is employed in forecasting the demand of a given data set. The main feature of the LAV technique is its capability of rejecting any bad data in the parameters estimation process without any previous knowledge of their location. To illustrate the efficiency of the LAV technique in electric load forecasting, two types of applications are considered. In the first application, the adequacy of the LAV technique for estimating reliable electric load forecasting model parameters is illustrated. Results have shown that models with parameters estimated using the LAV technique generate better forecasting results than those using least-squares-technique-estimated parameters. In the second application, the efficiency of the LAV technique in estimating good forecasting model parameters for given bad data is demonstrated. The results have shown that the model with parameters estimated using the LAV technique produces quite reasonable forecasting results; whereas the model with least-squares-technique-estimated parameters generates completely unacceptable forecasting results due to the effect of bad data.