Design and Implementation of an On-Line Load Forecasting Algorithm

This paper presents a short-term load forecasting algorithm for Energy Control Center applications. Regression techniques are combined with Auto Regressive Integrated Moving Average (ARIMA) models to produce hourly MW load values for the upcoming seven days. Results of several studies are discussed along with a description of the on-line implementation. The Load Forecasting algorithm discussed in this paper produces hourly MW load values for the remainder of the current day and the upcoming six days. The forecasted values are to be used for unit commitment and short-term maintenance planning Since the implementation of the algorithm is part of an Energy Control Center computer system, it was developed to take into account the following factors: 1) New forecasted values should be computable at any time, reflecting the most recent on-line system data. 2) The amount of on-line data needed to produce the forecast values should be minimal. 3) The forecast computations should not require an excessive amount of time or resources (core). 4) Data entered by the dispatcher should be minimal. 5) The dispatcher interface must be clear and simple, i.e., the implementation must not require that the user understand the forecasting technique.