An adaptively trainable neural network algorithm and its application to electric load forecasting

Abstract

A training procedure that adapts the weights of a trained layered perceptron type artificial neural network to training data originating from a slowly varying nonstationary process is proposed. The resulting adaptively trained neural network (ATNN), based on nonlinear programming techniques, is shown to adapt to new training data that is in conflict with earlier training data with affecting the neural networks' response minimally to data elsewhere. The ATNN demonstrates improved accuracy over conventionally trained layered perceptron when applied to the problem of electric load forecasting.<>