Modeling of wind speed and relative humidity for Malaysia using ANNs: Approach to estimate dust deposition on PV systems

This paper presents a wind speed and relative humidity predictions using feedforward artificial neural network (FFNN). Wind speed and relative humidity values obtained from weather records for Malaysia are used in training the FFNNs for estimating dust deposition on photovoltaic (PV) systems. Three statistical parameters, namely, mean absolute percentage error, MAPE, mean bias error, MBE, and root mean square error, RMSE are used to evaluate the neural networks. Based on results, the proposed neural network gives accurate prediction of hourly wind speed with MAPE, RMSE and MBE values of 43%, 0.56 and -0.35, respectively. Meanwhile, the MAPE values for predicting daily and monthly wind speed are 13.04% and 4.8%, respectively. On the other hand, the MAPE, RMSE and MBE values in predicting hourly relative humidity are 5.08%, 5.8 and -0.041, respectively. While the MAPE values for the daily and monthly predicted values are 2.66% and 0.57%.