Forecasting Wind Speed Using Machine Learning ANN Models at 4 Distinct Heights at Different Potential Locations in Pakistan

Abstract

The explosive progression in population causes a rapid reduction in the resource of fossil fuel which is the basic supplier of energy in industry and household. This scarcity of fossil fuel is the reason for the costly produced energy. However, pollution is also one of the severe issues occurring due to the burning of gases. Therefore, different researcher worldwide drew their attention to clean and environmentally –friendly energy resources. Wind energy is a renewable source of energy and it is accumulated from renewable resources. Wind speed is one of the most significant parameters used to study the wind energy of any region. This paper presents the fitting of the Artificial Neural Network for the assessment of wind speed in different wind stations in Pakistan. Five Neural Network models have been fitted to the 10-minute mean wind speed data from 2016 to 2018 of each of four distinct heights in 12 different stations in Pakistan. Conventionally used statistical measures are utilized to assess the best-fitted model. The simplest model shows the minimum values of MSE and R2 amongst all other models. The model of one hidden layer with five neurons is the best-fitted model in 12 different stations with four distinct heights in Pakistan. We will be extending this work by applying some other soft computing algorithms such as a random forest with different optimization techniques such as genetic algorithm and swarm optimization algorithms.