Utilizing Artificial Neural Networks for Wind Speed Estimation A Case Study of Dernah City, Libya

Abstract

Electric power is universally acknowledged as a crucial factor in enhancing living standards. As a result, safe electrical energy consumption is crucial to efficient national energy management. To do this, meticulous assessments of the electricity demand are required. Finding viable sites for turbine placement through feasibility studies and measuring local wind speeds are essential steps before establishing the plant wind power. Estimation of wind speed and simulations can be used to conduct these evaluations.. This study uses an artificial neural network (ANN) with the Levenberg-Marquardt (LM) learning algorithm to estimate wind speed for the Libyan city of Dernah. One-year data from the Libya Meteorology Center has been utilized to train, test, and validate the ANN to to predict hourly wind speed . The structure of the ANN was evaluated with neuron counts of 10, 20, 30, 40, and 50, allowing us to determine the optimal number of neurons for accurate predictions. The estimation analysis was performed using results obtained from the Levenberg-Marquardt method (LMA), along with the mean square error (MSE) and the coefficient of determination (R²). The results show that the Levenberg-Marquardt method with 10 neurons performs the best, with values of 0.99661 for R2 and 0.000250 for MSE. These findings confirm that wind speeds can be calculated within reasonable bounds since they show that the estimates of wind speeds based on the scant meteorological data available nearly match the measured values.