A Comparison Between using a Neural Network and a Fuzzy Regression System to Predict the Values of Hydro Power System Variables

M. Boukadoum, H. Lounis, G. Mai, H. Sahraoui, and V. Siveton (Canada)


forecasting and prediction, knowledge based systems, machine learning, hydro power system.


We compared the performance of an extended Elman neural network vs. that of a tree-based fuzzy regression system when using a database of historical hydrological data to predict the natural contributions flow in a hydroelectric power generation network. The neural network was trained with the Resilient Backpropagation (RPROP) algorithm and the fuzzy regression tree consisted of a new design where input fuzzification is accomplished by using mathematical morphology and output defuzzification is done by a multilayer perceptron (MLP) trained with the backpropagation with momentum algorithm. The purpose of the comparison was to select the best prediction technique to be part of a software framework adapted to hydroelectric power system assessment. The framework uses variable prediction to support rule-based decision processes. Our results are that the best prediction accuracy is obtained with the extended Elman neural network.

Important Links:

Go Back