Prediction of hydropower generation based on artificial neural network: case of hydropower plants in Cameroon

To meet the country's energy needs, Cameroon has built hydroelectric dams. Unfortunately, these old dams are now insufficient to produce the expected installed capacity. This drop, caused by aging production equipment and electrical or mechanical faults in some of its protection components, is leading to an increasingly inadequate supply of electrical energy. The aim of this study is to predict the hydroelectric power generation capacity using machine learning techniques. The solution developed based on artificial neural networks can make it possible to continuously control energy production and instantly meet changes in demand from the southern interconnected grid. Based on the results obtained, a model based on the multilayer perceptron validated with data from 2024 by the Mean Absolute Percentage Error, with a value of 0.071, and a mean square error of 0.065723 as performance criteria was designed. The prediction algorithm obtained regression fitting results of the highest accuracy with training value R = 1, validation value R = 1, test value R = 1 and overall regression with value R = 0.99187. The proposed model was used to forecast hydroelectricity production. These energy forecasts are indicative data that can help decision-makers know in a timely manner the unforeseen changes in electricity production.