Multilayer Perceptron Application in Electricity Consumption Forecasting from Wind Power

Abstract

In the context of increasing global demand, and the importance of renewable energy sources. There is a need to focus on renewable energy management and explore more efficient sources of energy generation. In recent years, the rise of the transition from fossil energy to renewable energy and its decreasing cost have posed new forecasting problems of interest to researchers, namely, the knowledge in advance of the electricity consumption that wind power plants may generate in the future. Recognizing the need to improve renewable energy management as a strategy to search for better and more efficient sources of energy generation, a system for forecasting the generation of electricity consumption by wind power using neural networks is proposed that allows maximum utilization of the potential that can produce wind power. A backpropagation neural network is used as an artificial intelligence component to accurately predict electricity consumption. The network is a multi-layer perception (MLP) network that uses a reverse transcription learning algorithm to train a feed-forward neural network to perform a specific task. The Levenberg-Marquardt algorithm is used to improve network training and evaluate its performance. In this study, the suggested approach trains the feed-forward neural network for the specified task using a multi-layer perception (MLP) network and a backpropagation learning method. By modifying the weights of the connections between neurons, the number of neurons in the hidden layers is optimised in order to reduce the output error in comparison to the desired target values and achieve high accuracy. Support vector machine (SVM) and auto-regressive integrated moving average (ARIMA) models' performances are compared to those of the suggested model.