| Popis: |
Prediction, separation, and improvement of beam (Hb), diffuse (Hd), and global solar radiation (H) using an efficient Gumbel probabilistic model (GP) and hybridization of GP with auto-regression integrated moving average (ARIMA) is essential in regions of Southern Africa and the Middle East. This is because, most of the inhabitants of these localities are not connected to the national grid due to extreme cost implications, frequent power outages, and in most cases, unavailability of the national power supply; as well as the fact that most government meteorological stations are technologically or financially unable to routinely measure these radiometric parameters in most metropolitan cities, developing cities, and remote villages, where there is a severe need for electricity. Although the prediction of H has many advantages in adapting and deploying clean and affordable energy infrastructure through solar photovoltaic (PV) systems, the separation of H into Hb and Hd will further increase cleaner and more affordable systems like solar PV thermal/concentrators, which require Hb for its use. In this era of climate change, Hb, Hd, and H information are needed to detect and adapt climate mitigation plans in places adversely affected by climate change and global warming externalities. Eight different configurations of input combination elements were assembled to stimulate their hybrid evolutionary ARIMA-GP controlled, swapped model through the instrumentality of generalized solar meteorological datasets needed for modeling. The result revealed that the swapped ARIMA models outperformed the controlled and controlled ARIMA models using Gumbel's numerical approach. The best min beam and diffuse irradiance frocontrolled ARIMA models were probabilistically optimized using the Gumbel (GP) probabilistic model to produce ARIMA-GP. From the evaluations of the error metrics, the ARIMA-GP models outperformed the swapped and controlled ARIMA models, as well as the Gumbel models used to separate HB and Hd from H. The selected best performing models produced the RMSE-induced decrease and increase in R2 of 73.73% and 15.25% respectively for GSRML-GP3 (model H), 71.71%, and 14.19% for DSRML-GP5 (model Hd), and 61.91% and 11.70% for DNIML-GP5 (model Hb). These high values of the percentage improvement of the suitable model obtained for the Hb, Hd, and H modeling suggest that the proposed hybrid evolutionary ARIMA-GP model can be adopted to improve the prediction of solar beam, diffuse and global radiation in any region of the world, as the input parameters incorporate the data sets on geo-climatic conditions needed for global modeling. This study also suggests that it is more realistic to use the generalized functional forms GP and ARIMA-GP to separate Hb and Hd from the parameter H and also to generate their corresponding coefficients in the studied regions than applying empirical and machine learning models consolidated in the literature. The proposed ARIMA-GP models are sufficient as a valid evolutionary hybrid model that will accelerate a holistic understanding of the solar resources available in the selected regions and beyond, as well as spread the application of solar photovoltaic/thermal technologies within these countries. |
