Data-dependent systems approach to short-term load forecasting

A recently developed stochastic modeling and analysis methodology, called data-dependent systems (DDS), is introduced. The forecasting application of a univariate DDS model is illustrated for the actual hourly load data for a small community (Curtis, NE, USA). An accurate forecast for peak values of the load is provided by the conditional expectation of the statistically adequate model ARMA. The dynamics of this model and the possibility of applying multivariate DDS models to short-term load forecasting are also discussed.