Assessing the economic and technical efficacy of grid-connected photovoltaic systems through a combination of mathematical and deep learning models with experimental validation: A case study

Abstract

ABSTRACT The current study presents a techno-economic technique and modeling to evaluate the productivity of a 1.25 kW photovoltaic grid-connected (GCPV) system. The method used four key factor indicators: yield, and capacity factors, also payback period and cost of energy. MATLAB software was used to analyze secondly meteorological data and create a model for the GCPV. Also, the GCPV has been installed, tested, and data collected in Sohar, Oman. The data revealed that the daily average irradiation is 6.23 kWh/m2.day in the studied location. The numerical and experimental results of the assessment indicated that the investment in GCPV technology in this site is very encouraging, with a yield factor of 1710 and 1836 kWh/kWp and a capacity factor of 19.37% and 20.74%, respectively. Also, it is found that 8.25 years is the payback period. Additionally, the numerical and experimental results show 0.063 and 0.047 USD/kWh cost of energy, respectively. Also, four predictive models based on deep learning utilizing Recurrent Neural Network (RNN) and Time Lag Recurrent Network (TLRN) were developed to forecast GCPV current performance with irradiance, and temperature as input data. FRNN-2 and FRNN-3 showed the highest performance for prediction models due to their lower MSE (0.0018–0.0037), indicating better accuracy in comparison to experimental results. This research yields important technical and financial data for anyone looking to invest in GCPV technology in Oman.