This paper deals with the enhancement of photovoltaic water pumping systems through machine learning–based modeling and optimization. First, parameter identification of the target PV module KC200GT is carried out using three algorithms: Newton–Raphson, fsolve, and particle swarm optimization (PSO). Among these methods, PSO shows the best agreement with the manufacturer’s data, achieving best performance (R2 = 0.9901 and RMSE = 0.0358). In addition, manufacturer data of the centrifugal pump, generally given for a unique nominal speed, are used to develop a model capable of predicting the pump flow rate, total head, and load torque at different rotational speeds. Furthermore, four machine-learning models – artificial neural networks (ANN), adaptive neuro-fuzzy inference system (ANFIS), support vector machine (SVM), and Gaussian process regression (GPR) – are evaluated for predicting the maximum power point voltage. Comparative analysis indicates that ANFIS achieves the highest prediction accuracy, followed by GPR, ANN, and SVM, outperforming conventional algorithms. Consequently, ANFIS combined with a PI regulator is adopted as the control strategy for tracking the maximum power point. Finally, the dynamic behavior of the photovoltaic water pumping system is examined under variations in irradiance and temperature. The results demonstrate that the proposed control strategy effectively maintains the PV voltage close to its optimal value and responds rapidly to changing climatic conditions, confirming its suitability for PV water pumping applications.
