Time series forecasting of utility load duration curves

The load duration curve is a primary tool used in the economic analysis of electric utility operations and for the purpose of planning new construction. In general equilibrium models and also in other models using linear programming, piecewise linear approximations to the load duration curve are required. This paper endeavors to develop a forecasting model for such discrete approximations using Box-Jenkins time series analysis. A model is structured and estimated for six, twenty, and fifty discrete approximations for two utility regions in the United States. The forecast results are extremely good, indicating that while a larger number of approximations perform marginally better, there is nothing to lessen the advantage of using only a six-step approximation.