Deriving and Validating an Irradiance-Dependent Detection-Limit Floor for a Low-Cost Fixed-Load Photovoltaic Fault Monitor

Although low-cost photovoltaic (PV) monitoring systems are able to indicate underperformance, they can fall short of detecting a fault due to sensor noise and model uncertainty. This paper presents an irradiance-dependent closed-form-like detection limit, and ࢔࢏࢓ࡼࢤ)ࡳ(, for reduced fixed load PV fault monitor using power residual thresholding. We derive the limit by cascading the measurement error of module voltage, load current, irradiance and cell temperature through a PV module model using a single PV cell. We validate the noise-propagation model using sixteen days of outdoor measurements on a 10 W crystalline-silicon PV module equipped with an ESP32-based logger. The power-residual scatter in the prediction is in good agreement with the scatter observed under steady irradiance within an order of unity, whereas the residuals are much larger during cloud transition periods. The detection limit rises from an estimated 19 mW under low irradiance to an estimated 100 mW under high irradiance, or from about 14% of operating power to about 1.9% of operating power. The voltage-measurement channel is the main contributor to residual variance, meaning the biggest hardware upgrade to make is an improvement to the voltage signal acquisition. The results give a practical guideline for calculating the minimum power loss that can be detected by the use of a low-cost fixed-load PV monitor. Based on the healthy-module data, this study verifies the sensor-noise floor, and we propose controlled fault-injection experiments as the next step in sensor-noise floor validation.