Parameter estimation and performance comparison ın photovoltaic models using metaheuristic optimization algorithms

Solar energy has an important place among renewable energy sources due to its sustainability and wide application potential. For the most efficient utilisation of this energy source, accurate and reliable estimation of the parameters affecting the performance of photovoltaic (PV) systems plays a critical role. PV parameter estimation enables precise determination of the electrical characteristics of solar panels, improving system efficiency and optimising power generation processes. In this regard, electrical models used in PV systems have a vital importance in terms of accuracy, system design and performance improvement. In this thesis, new and efficient meta-heuristic optimisation algorithms are proposed for the estimation of PV cell parameters. The proposed approaches are the weighted mean of vectors (INFO), quadratic interpolation optimisation (QIO) and frilled lizard optimisation (FLO) algorithms. Furthermore, these algorithms are integrated with the Newton-Raphson (N-R) analytical method to determine the nonlinear current-voltage characteristics of PV systems more precisely, thus improving their optimisation performance. In this thesis, the effectiveness and accuracy of the proposed algorithms are evaluated by testing them on different data sets. Using R.T.C France, Photowatt-PWP201, STM6-40/36 and STP6-120/36 data sets, a wide range of analyses were performed with single diode, double diode and three diode electrical circuit models. The effectiveness of the respective models was compared in detail using various error metrics such as individual absolute error, relative error, mean absolute error, mean bias error, normalised RMSE, normalised MBE, normalised MAE, coefficient of determination and t-statistic. In addition, the performance of the algorithms is analysed in depth by evaluating the obtained current-voltage and power-voltage characteristic slopes, convergence slopes, maximum power points and statistical metrics. The results show that INFO, QIO and FLO algorithms offer significant advantages compared to the meta-heuristic optimisation methods proposed in the literature. These algorithms provide high precision, fast convergence and stable results in optimising PV system performance, strengthening their potential for industrial and academic use. Especially, it is shown that by optimised parameter estimation, the energy conversion efficiency of PV systems can be improved and used more effectively in real world applications.