Prediction of electricity production by small wind power using artificial neural networks

Introduction Wind energy is one of the most significant and rapidly growing renewable energy sources worldwide. It is a clean and environmentally friendly form of energy production, which emits no harmful substances or greenhouse gases during the power generation process. There has been a growing interest in research in the field of wind energy. In this article, an artificial neural network method is used to evaluate the forecasting of wind energy production from a small wind turbine (SWT) installed in central Poland, reflecting inland wind conditions. Methods A comprehensive set of algorithms and results from simulations are presented. An artificial neural network (ANN) is trained and verified using a large observation dataset. The model includes four input variables: wind speed and direction, rotor speed, air temperature, and one output variable - the power generated by the turbine. Among the available neural networks, Multilayer Perceptron was selected. Genetic algorithms were used to optimize the structure of the model. The Pearson correlation coefficient was used to assess the correspondence between the predicted values and the actual ones. The modeling was carried out in MATLAB, and coefficients such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) were used to evaluate the prediction error. Results and Discussion The learning and testing performance of the neural network model using back propagation with feedback was 96.3% and 97.0%, respectively. Additionally, a sensitivity analysis of the predictive model was performed. The neural network model presented in the article provides accurate predictions of the power generated by a wind turbine. The results obtained confirm the effectiveness of the use of MLP-type neural networks in tasks related to the prediction of energy production in small wind turbines in inland locations.