Energy, exergy and performance analysis of a 380 kWP roof-top PV plant assisted with data-driven models for energy generation

Authors : Abhijeet Rathore, - Almas, Sivasankari Sundaram

Pages : 1164-1183

View : 9 | Download : 18

Publication Date : 2024-09-10

Article Type : Research

Abstract :Energy and Exergy based performetric analysis integrated with deep learning assisted energy modelling for grid connected solar PV system, tested to non-trained location is proposed. The first objective is to perform an energy and exergy based performetric analysis for a realistic 380 kWp grid connected roof-top PV system whose performance parameter is used for testing the proposed energy prediction models. The second objective is to formulate a simple and an improved energy estimation method applicable for 34 locations in South India, without change in model-coefficients. So, a long-term annual performance analysis of a 380 kWp PV based distributed generator situated at 12.97°N and 77.59°E is performed which estimates the characteristic performance indicators like energy efficiency, exergy efficiency, performance ratio and capacity factor amounting to 8.49%, 1.03%, 37%, and 8.03% respectively. The performance ratio of the plant is less as evident from the least exergy efficiency. The annual average losses in the system like thermal capture loss, array capture loss, system loss and miscellaneous loss amount to 0.46 (h/d), 2.51(h/d), 0.71 (h/d) and 2.97(h/d) respectively. The annual average energy generation of 380 kWp is 732.84 kWh/year. Furthermore, for realizing the second objective, a total of four models are proposed namely linear, exponential, non-linear and deep learning based neural network model resulting in R of 0.933, 0.9071, 0.9386, and 0.9603 respectively is formulated. The proposed models are tested for non-trained locations where the R value justifying the closeness between the actual and the predicted value is as high as 0.8. The proposed models are then compared upon their performances and benchmarked against the reported models.
Keywords : Deep-learning and 380 kWp PV Plant, Energy Efficiency, Exergy Efficiency, Testing for Nontrained Location