PILoN: Physics-informed logistic nonlinear model for photovoltaic performance modelling

For over 20 years, Sandia National Laboratories, USA, have been working on improving photovoltaic (PV) performance models. Between 1991 and 2003, they proposed the Sandia Array Performance Model (SAPM), which has become an integral component of the System Advisor Model (SAM) distributed by NREL. This study raises concerns about the suitability of SAPM and exposes accuracy loopholes stemming from repetitive translations of field measurements, which distort the inherent temperature–irradiance physics of PV cells. To address these limitations, this paper proposes a Physics-Informed Logistic Nonlinear (PILoN) model for high-performance PV module/array modelling. PILoN replaces SAPM’s complex, multi-step regression procedure with a unified, physics-informed framework that uses surrogate models for open-circuit voltage and short-circuit current, eliminating the need for difficult-to-measure variables such as air mass. It captures the non-separable coupling between irradiance and temperature using bounded logistic functions for the ratios of maximum power point voltage to open-circuit voltage and current to short-circuit current. A Physics-Informed Particle Swarm Optimization embeds physical constraints and monotonicity priors directly into the calibration objective. Extensive validation across several commercial technologies shows that PILoN achieves superior accuracy, especially under low irradiance where SAPM’s linear assumptions fail. For three benchmark modules (SM55, KC200GT and ST40), PILoN attains voltage RMSE values of 0.16, 0.31 and 0.53 V and current RMSE values of 0.0029, 0.0082 and 0.0046 A, corresponding to voltage and current accuracies above 96 % and 99.5 %. On the Sandia SM55 module, the mean current error of SAPM over low-to-high irradiance is about 46 %, whereas PILoN reduces this to 3.3 % while preserving physical consistency across the full operating range, positioning it as a less measurement-intensive and more robust alternative to SAPM for design, optimization and performance prediction of modern PV systems.