Day Ahead Load Forecasting using an Artificial Neural Network & Elman Recurrent Network

Ellen Banda and Komla A. Folly


Artificial Neural networks, computational intelligence, elman recurrent neural network


The forecast of load for a period of the next minute up to a week is defined as short term load forecasting. It is required for generation commitment and dispatching as well as assisting Power System Operations engineers with the analysis of network contingencies. Conventional methods such as Box Jenkins and Regression Based methods were previously applied to short term load forecasting. It was found that these methods were unable to adapt to the dynamics of a power system resulting in large forecasting errors. Computational Intelligence techniques were then developed to improve on the inefficiencies of the conventional methods. This paper presents two day ahead short term load forecasting models using an artificial neural network and an Elman recurrent neural network. Weather and non-weather models are developed for comparative purposes. The models are then applied to actual data obtained from a power utility in South Africa.

Important Links:

Go Back