Africans in general and specially Beninese’s low rate access to electricity requires efforts to set up new electricity production units. To satistfy the needs, it is therefore very important to have a prior knowledge of the electrical load. In this context, knowing the right need for the electrical energy to be extracted from the Beninese network in the long term and in order to better plan its stability and reliability, a forecast of this electrical load is then necessary. The study has used the annual power grid peak demand data from 2001 to 2020 to develop, train and validate the models. The electrical load peaks until 2030 are estimated as the output value. This article evaluates three algorithms of a method used in artificial neural networks (ANN) to predict electricity consumption, which is the Multilayer Perceptron (MLP) with backpropagation. To ensure stable and accurate predictions, an evaluation approach using mean square error (MSE) and correlation coefficient (R) has been used. The results have proved that the data predicted by the Bayesian regulation variant of the Multilayer Perceptron (MLP), is very close to the real data during the training and the learning of these algorithms. The validated model has developed high generalization capabilities with insignificant prediction deviations.
A model of the higher heating value on a dry basis from the proximity analysis of agricultural wastes in Benin has been proposed in this article. This model was developed using agricultural residues such as shea shells and cakes, cotton and soybean stalks, corn cobs and peanut shells identified as part of the implementation of an experimental system. The validity of this model has been established for the Higher Heating Value (HHV) between 18.07 MJ/kg to 25.91 MJ/kg, Volatile Matter rate (%VM) 66.8% to 79.87%, Fixed Carbon rate 13.83% to 29.59%, and Ash content (%Ash) 3.47% to 6.3%. The model has an average absolute error of 2.79% and a bias error of 0.034%, significantly better than the most accurate literature prediction model, which offers a mean absolute error of 5.97% and –4.66% for the bias error. This work presents as well the first data from the proximity analysis of agricultural residues in Benin. These analyzes are carried out using a well-structured methodology that respects the standards and measures of simple random sampling forsample collection. Samples prepared under appropriate conditions are analyzed using standardized protocols for the agricultural wastes studied.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.