Assessment of artificial neural network and Weibull distribution methods for wind power density prediction in Ethiopia

Abstract

Abstract Wind power is variable and intermittent over a range of periods, and its output is reliant on meteorological parameters. Thus, the installation, operation, and management of wind farms depend heavily on precise wind power predictions. This paper proposes the investigation of wind power potential in Dera, Dire-Dawa, and Negele-Borena, Ethiopia, using an artificial neural network (ANN) and a Weibull probability distribution function (PDF). The input data (temperature, pressure, relative humidity, wind speed, and height above the surface) that were used for the ANN model of MATLAB R2018b were first reprocessed to assess wind power density (WPD). The findings showed that the ANN was much closer to the real wind speed than the power law [(0.0043 ⩽ MBE ⩽ 0.0312) and (0.2036 ⩽ RMSE < 0.5044)]. The ANN (0.990 ⩽ R 2 ⩽ 0.998 and 0.987 ⩽ COE ⩽ 1.0101) approximates the wind power densities for the three sites more accurately than the Weibull PDF. The Dera, Dire-Dawa, and Negele-Borena sites provide monthly estimates of wind power densities ranging from 20.97–103.81 W m −2 , 36.24–185.22 W m −2 , and 52.77–208.33 W m −2 , respectively. In the future, the ANN can also be used to predict wind speed and capture WPD across Ethiopia.