Long Short Term Memory — Autoencoder Approach for Anomaly Detection in Rooftop Solar PV System

This paper focuses on the application of Long Short-Term Memory Auto Encoder (LSTM-AE) networks in the maintenance of Solar Photovoltaic system for anomaly detection. LSTM-AE networks are particularly effective in handling variable sequence lengths and reconstructing the error, making them suitable for modelling the operational behavior of rooftop solar PV system. By integrating LSTM-AE networks with AI technologies, the research aims to enhance the operational efficiency and reliability of rooftop solar. This integration enables accurate anomaly detection, trend analysis, and proactive maintenance strategies, ultimately improving the overall performance of solar photovoltaic (PV) systems. The implementation of the model on rooftop solar resulted in Root Mean Squared Error scores of 0.00137. The algorithm identifies patterns of system deviations from normal operational behavior and generates predictive maintenance alerts as a decision support tool for operators, aiming to prevent potential failures.