An Enhanced FSO-BPNN Framework for Anomaly Detection and Early Warning in Power System Monitoring

Abstract

The increasing complexity of contemporary power networks necessitates the development of enhanced early warning systems and intelligent monitoring to ensure stability and operational efficiency. Traditional approaches to risk prevention and predictive maintenance often fail due to limitations in identifying real-time abnormalities and adapting to dynamic system characteristics. To address these issues, the present research proposes an improved fish swarm optimization with Backpropagation Neural Network (IFSO-BPNN) for anomaly detection (AD) and fault detection (FD) early warning in power system (PS) monitoring that integrates an IFSO algorithm with a BPNN. The major goal is to increase the accuracy of AD and FD in smart grids by utilizing deep learning (DL) and optimization approaches. The IFSO method integrates adaptive weighting and behavioral dynamics into classic fish swarm optimization, improving overall search capabilities. By tweaking BPNN parameters using IFSO, the model achieves higher convergence rates and improved classification accuracy. The assessment dataset was compiled usingInternet of Things (IoT) sensors and pan/tilt camera-based surveillance systems at Beijing power plants, with preprocessing techniques such as min-max normalization and feature extraction using Independent Component Analysis (ICA) to improve model performance. Resultsfrom experiments show that the IFSO-BPNN model outperforms standard algorithms with an accuracy ofFD99.98% and AD 0.9980. These findings illustrate the system's capacity to detect anomalies quickly and perform preventive maintenance. The proposed method, which combines swarm intelligence with neural networks, helps to construct smarter, more robust power grids capable of meeting future energy demands with lower failure risks.