Solar radiation estimation in different climates with meteorological variables using Bayesian model averaging and new soft computing models

Abstract

Solar radiation (SR) is considered as a critical factor in determining energy management. In this research, the potential of the Bayesian averaging model (BMA) was investigated for estimating monthly SR. The inputs were monthly average temperature, wind speed, relative humidity, and sunshine duration. The BMA model was employed to estimate SR by extracting information from multiple adaptive neuro-fuzzy systems (ANFIS) and multi-layer perceptron (MLP) models. In this study, Archimedes optimization algorithm (AOA), particle swarm optimization (PSO), genetic algorithm (GA), and bat algorithm (BA) were used to tune the parameters of the ANIFS and MLP. In addition, a multitude of error indices such as root mean square error (RMSE), and Nash Sutcliff efficiency (NSE), and several graphical tools were used to investigate the accuracy of the models. The results showed the better performance of the BMA model than other models for estimating solar radiation. For example, BMA with RMSE of 6.78, MAE of 5.25, and NSE of 0.96 had the best accuracy in the training stage of the Tabriz station. On the other hand, in the testing level of Tehran station, BMA (RMSE=7.89 MJ/ m2, MAE=6.89 MJ/ m2, NSE=0.95) gave the best accuracy, and the MLP model (RMSE= 14.12 MJ/ m2, MAE=12.23 MJ/ m2, and NSE=0.77) gave the worst performance, respectively.

Keywords

• Solar radiation
• Energy management
• Soft computing models
• Archimedes optimization algorithm