An intelligent environmental-sensorless model for real-time optimization in photovoltaic-energy systems: Fast computation, long-term performance, and experimental validation
Ambe Harrison, Hassan M. Hussein Farh, Abdullrahman A. Al-Shamma’a, Saad Mekhilef
Abstract
Accurate determination of the Maximum Power Point (MPP) is critical for optimizing photovoltaic (PV) systems, especially in resource-constrained settings where the use of climatic sensors is impractical or cost-prohibitive. This study introduces a novel sensorless method for MPP voltage estimation based solely on low-cost electrical measurements—current, voltage, and derived resistance—through the predictive relation MPP = f (I, V, R). This formulation eliminates the need for irradiance and temperature input, which are commonly required in conventional MPP models. To realize this concept, an IVR-Neural Network architecture was developed and trained using data from three representative PV technologies: Monocrystalline, Polycrystalline, and Thin Film. The model was validated across 90 randomly generated irradiance–temperature pairs, demonstrating strong generalization across diverse operating conditions. Experimental results revealed high prediction accuracy, with a Mean Absolute Percentage Error (MAPE) between 2.89 % and 3.27 %, and correlation coefficients R exceeding 0.95 for all module types. Mean Absolute Error (MAE) remained below 0.8 V, confirming the model's precision. Furthermore, the long-term reliability of the method was assessed by simulating PV aging over 10- and 20-year periods. Despite expected optical and electrical degradation, the estimated MPP voltage exhibited minimal variation—less than 0.1 V—highlighting the model's robustness against system aging. Thanks to its lightweight structure and offline training process, the proposed IVRNN model ensures real-time deploy ability with negligible computational cost. Overall, the method provides a fast, accurate, and scalable solution for MPP determination in sensor-constrained environments. It holds significant promise for integration into MPPT controllers, energy yield forecasting tools, and low-cost PV deployment in developing regions or standalone applications.