Innovative Small-Scale PV Modeling: Optimizing the Single-Diode Approach for Sub-Watt Systems

Internet of Things (IoT) energy-harvesting-nodes operating in the ultra-low-power (ULP), sub-watt (<1 W) domain rely on small-form-factor photovoltaics (PV) and typically operate under low irradiance, making accurate behavioral PV modeling essential for reliable energy availability prediction. We propose a robust parameter extraction method for the single-diode model (SDM) for sub-watt systems, reducing behavioral variability under parameter uncertainty in ULP PV systems. This method employs the Dogleg trust-region algorithm with calibrated diode ideality factor (n). Device-calibrated values of n = 1.1 provide a good approximation of the current-voltage (I-V) curve, while n = 1.7 improves agreement under load. Under standard test conditions (STC) (1000 W/m², 25°C), the model reproduces system current-voltage (I-V) and power-voltage (P-V) characteristics, yielding root mean square error (RMSE) values of 0.0169 A and 0.017 W, and mean absolute percentage error (MAPE) values of 7.49% and 4.18%. The method is assessed under relevant environmental conditions and predicts power under varying irradiance with an RMSE of 0.0192 W (MAPE 11.73%). Comparison with Photovoltaic Geographical Information System (PVGIS) shows that linear irradiance scaling can misrepresent low-light behavior. In contrast, the proposed SDM captures nonlinearities and environmental effects, yielding closer agreement under ULP-relevant operating conditions. This improved behavioral fidelity enables more reliable energy forecasting, supporting robust system design and batteryless IoT deployments.