Blind PV temperature model calibration

Abstract

The determination of module temperature in a photovoltaic (PV) system is a crucial factor in PV modelling and the assessment of system health status. However, the scarcity of on-site temperature measurements poses a challenge, and existing PV temperature models encounter difficulties in accurately estimating temperatures in systems characterized by unique structural or locational attributes. This paper introduces a novel approach that enables the calibration of PV temperature models without relying on direct temperature measurements. Referred to as blind calibration, this method eliminates the requirement for temperature measurements, thus offering a promising solution to the aforementioned challenges. The method is validated using three datasets, demonstrating accurate PV temperature ( T PV ) estimation with mean absolute errors below 2 °C. The findings highlight the suitability of the proposed approach for various PV system types, while acknowledging limitations regarding certain system configurations.