Application of statistical and neural approaches to the daily load profiles modelling in power distribution systems

Load modelling is an essential task in the economic analysis, operation and planning of distribution systems. Particularly, when a demand side management system is taken into account on a deregulated energy market, the knowledge of load profiles is of the greatest importance. Forecasting of daily demand, based upon load models, uses comparable load research data for a different customer mix. For the given season and day of the week, the shape of a daily load curve depends mainly on the customer composition. Difficulties in defining objective customer classes significantly complicate the forecasting process. Usage of statistical clustering and neural network approaches makes possible to improve the load modelling accuracy. This paper presents load modelling methods useful for the long-term planning of power distribution systems. The theoretical statement is illustrated by examples which correspond to Polish and German distribution systems.