Preventive Maintenance for Large PV Farms Using Data-Driven Parameter Estimation and Relative Performance Evaluation

Abstract

Large-scale photovoltaic (PV) farms require regular maintenance to ensure longevity and maximize energy efficiency. Considering the necessity and associated challenges of maintaining large PV farms, this work proposes a real-time health monitoring system based on relative performance computation and environmental parameter estimation. Unlike conventional performance ratio (PR) methods that depend on extensive sensor networks, the proposed system estimates irradiance and temperature using a reference string operating at optimal performance, overcoming the limitations of traditional PR approaches. Since the proposed method relies on a high-performance reference string, the panel degradation factor is included in the individual string's relative performance ratio calculation to ensure long-term health evaluation accuracy. As a result, the proposed health monitoring system is durable and efficient in reporting string-level instantaneous anomalies while optimizing costs and data usage. With optimized data requirements, the presented approach reduces the number of comparisons needed to identify faulty strings, thereby minimizing sensor requirements. Furthermore, for efficient data transfer between strings and the main controller, a Long Range (LoRa)-based communication protocol is proposed, featuring a specially designed packet structure dedicated to PV farm monitoring. The system's effectiveness is experimentally validated on a 70 kW campus PV farm, demonstrating real-time, cost-effective fault detection, and reliable data transmission.