SOARES, R. A.; http://lattes.cnpq.br/9101675888242470; SOARES, Rivanildo Alves.
Résumé:
Photovoltaic solar energy generation systems emerge as a promising solution to meet the growing demand for global electricity. However, their implementation faces significant challenges due to the complexity in the dynamic modeling necessary to estimate the energy generation of these systems. In this context, a methodology is presented to enhance the process of estimating energy generated by grid-connected photovoltaic systems, considering the effects of weather conditions and seasonal climate variations at installation sites, as well as the physical and electrical characteristics of these generation systems. The performance evaluation of the proposed methodology (PM) was conducted by comparing it with real data collected in the laboratory, with a methodology traditionally used in the solar energy market and literature, and with data from real photovoltaic systems installed in different cities of Paraíba. In the tests performed with laboratory data, the PM showed good agreement, especially in temperature metrics, which presented the best results with lower errors and high values of the correlation factor (CF) and coefficient of determination (R²). Other variables, such as current and power, showed similar metrics, with average errors around 18% and CF and R² values around 0.85 and 0.73, respectively, indicating satisfactory adjustment. In the application to real systems, it was observed that energy losses due to module degradation represent a considerable source of inefficiency, ranging between 3.57% and 4.48% per year. Additionally, the power limit of the inverter and its losses due to inefficiency are the main contributors to total losses in alternating current (AC), estimated between 4.15% and 4.48%. When evaluating energy generation estimates by real systems, the PM stood out with more balanced results compared to the traditional methodology (TM), showing low error values, a Normalized Mean Bias Error (NMBE) of 5.8%, a good correlation (average R² of 0.90), and a lower average Normalized Root Mean Square Error (NRMSE) of 20%. These indicators highlight the efficiency of the PM in providing a balanced approach between underestimation and overestimation of energy generation from grid-connected photovoltaic systems.