Robust and efficient hybrid autoencoder-ADAM (HAA) algorithm for analysing anomalies in Indian electricity consumption data

Anomaly detection in electricity-consumption data plays a crucial role in ensuring the reliability and stability of modern smart-grid systems. In this study, we propose the Hybrid Autoencoder-ADAM (HAA) algorithm, specifically designed for anomaly detection in Indian electricity consumption data from 2014 to 2023, considering distinct seasonal patterns. The HAA algorithm combines autoencoders with adaptive optimisation (ADAM) to effectively capture and reconstruct normal consumption patterns. Comparative analysis show that the HAA algorithm outperforms Long Short-Term Memory (LSTM) and XGBoost in accuracy and robustness for anomaly detection. It demonstrates adaptability across different seasons, regions and periods, offering valuable insights for advancing smart grid analytics and energy conservation strategies. Future research includes hyper-parameter optimisation and exploring ensemble methods to enhance its real-world applicability in operational smart-grid scenarios. The HAA algorithm presents a promising approach for large-scale smart grid anomaly detection, emphasising its efficiency and effectiveness in improving energy management and resource optimisation.