SmartPV-AIoT: an AIoT-integrated framework for fault diagnosis and remote monitoring in photovoltaic systems

• Examines how faults in photovoltaic systems reduce energy efficiency and increase fire hazards. • Proposes an AIoT-based framework integrating edge computing for remote monitoring and fault diagnosis, enhancing reliability and real-time performance. • Attains 96.75% diagnostic accuracy with a stacked Res-BiLSTM classifier, effectively detecting faults under challenging conditions. • Enhances real-time fault detection while minimizing latency, computational overhead, and costs. • Provides real-time monitoring via a mobile application, improving user convenience and operational efficiency. This paper presents SmartPV-AIoT, a cost-effective and edge-computing-based framework for real-time fault detection and environmental monitoring in photovoltaic (PV) systems. Addressing the challenges faced in remote or intermittently connected environments, the proposed system integrates electrical and environmental sensor data using a Raspberry Pi 4 and NodeMCU microcontroller. A lightweight Res-BiLSTM model, combining residual learning and bidirectional long short-term memory (BiLSTM) layers, is deployed at the edge to perform local fault classification by analyzing seven key parameters derived from I to V curve measurements and ambient conditions. Additionally, the framework employs multi-sensor data fusion by integrating voltage, current, irradiance, and temperature measurements to support comprehensive fault detection and predictive maintenance. The framework was validated on a 4.56 kW PV installation using a dataset of 2400 labeled samples representing normal operation, open-circuit, short-circuit, and partial shading conditions. The classifier achieved a fault detection accuracy of 96.67 %, with a maximum inference latency of 290 ms and low power overhead, making it suitable for real-time, on-site diagnostics. The system supports offline operation with automatic data synchronization upon connectivity restoration and includes mobile-based monitoring for remote accessibility. The proposed framework is designed for modular deployment and scales linearly with array size. Overall, SmartPV-AIoT offers a scalable, resilient, and practical solution for intelligent PV monitoring, particularly in resource-constrained or off-grid environments.