GENETIC ALGORITHM-ARTIFICIAL NEURAL NETWORK (GA-ANN) AND GIS-BASED WIND MAPPING FOR WIND ENERGY EXPLOITATION: CASE STUDY IN MALAYSIA

Abstract

Wind maps are required to determine wind resource over a given areas and they are an important component of wind energy exploration and exploitation. The intermittency of wind, geographical, and temporal variability, as well as the complex relationship between wind and their nature, have made accurate spatial wind speed modelling more difficult. The aim of this study was to contribute a novel and original solution to the problem of developing wind maps for wind energy exploitation in Malaysia. The main inputs of this study were 37 Malaysian Meteorological Department stations’ wind data and 3 installed wind masts’ data. The Genetic Algorithm-Artificial Neural Network model was applied in the Measure-Correlate- Predict method to substitute and fill missing data. Spatial modelling was conducted to establish wind maps by interpolating point sources of wind data and extrapolating the wind flow at 10-m and 50-m heights. The Genetic Algorithm-Artificial Neural Network model was also applied to training spatial modelling and to generate a nonlinear wind map. The results revealed that nonlinear wind map had addressed the overprediction issue of the wind maps in mountainous areas at the Cameron Highlands site, where the root mean squared error, and the mean absolute error decreased by 60.39% and 64.01% respectively. Overall, the nonlinear wind map improved simulated wind data by increasing accuracy and decreasing errors, up to 18.39% and 31.42% respectively. In conclusion, the results clearly prove that addressing the complex nonlinear relationship between the input parameters and output wind map decrease errors in the simulation of wind speed.