Development of a real-time power quality monitoring system using artificial intelligence

The increasing reliance on electrical power systems demands continuous monitoring of power quality (PQ) to prevent failures and ensure stability. Traditional methods of PQ monitoring often struggle with slow processing times and inefficiency, leading to the need for more advanced, real-time solutions. The capacity to swiftly and precisely process and analyze vast amounts of power data in addition to identifying anomalies, such as voltage sags, swells, and harmonics, is the main problem. This research proposes the development of a Real-Time PQ Monitoring System using Artificial Intelligence (AI) to address these challenges. The PQ data have been gathered, and the system employs robust preprocessing techniques, including data cleaning and normalization using Z score normalization. Feature extraction using Fast Fourier Transform (FFT), transforming time-domain data into the frequency domain to capture essential frequency components. Refined Ant Colony Optimization-Based Capsule Networks (RACO-CapsNet) is the suggested approach that combines the advantages of deep learning (DL) with optimization techniques to improve the precision, effectiveness, and resilience of PQ observation. Real-time anomaly detection is significantly enhanced by the system, as evidenced by performance metrics such as accuracy (97%), precision (98%), recall (98%), and F1-score (97%). Additionally, it exhibits a reduced detection latency, with disturbances detected in real time. The system's ability to identify and classify multiple PQ events simultaneously is validated through various scenarios. The RACO-CapsNet technique is highly effective for PQ monitoring in modern electrical grids, offering robust, scalable, and accurate real-time PQ management. Furthermore, the incorporation of AI-driven techniques allows for adaptive learning, ensuring that the system continually improves its performance based on real-time data. The combination of feature extraction through FFT and optimization via RACO ensures that even subtle PQ disturbances are accurately identified and categorized, reducing the likelihood of false detections.