A novel hybrid Teaching-Learning-Based optimization and Reduced-Form approach for efficient parameter extraction and accurate prediction of photovoltaic module performance

Accurate modeling of solar cells and precise identification of their equivalent model parameters are crucial for predicting maximum power output. This paper presents a novel hybrid method for extracting and predicting the five parameters of a photovoltaic module model. The proposed approach integrates an analytical technique, based on the least square method, to reduce the number of parameters to two with the teaching–learning-based optimization algorithm. This method effectively addresses the complexity of photovoltaic parameter extraction, which generates a high-dimensional research space, enhancing efficiency and accuracy. The proposed method was validated using real data from various module technologies yielding lower root mean square error values: 0.002133 A for the Photowatt-PWP201 module, 0.001721 A for the STP6-120/36 module, 0.000775 A for the RTC cell, 0.013969 A for the STM6-40/36 and 0.0152 A for the monocrystalline module measured at the Laboratory of Photovoltaic Systems, National Institute of Research and Development in Electrical Engineering in Bucharest, Romania. Furthermore, the results demonstrate that this method achieves lower normalized error of maximum power below 2.96% on the reference day for two module technologies, monocrystalline and polycrystalline, from the national renewable energy laboratory. The correlation between measured and predicted maximum power values over one year at two locations was consistently high, with determination coefficients close to 0.9987.