Deep Neural Network-Based Power Factor and Phase Angle Estimation for Real-Time Monitoring

Accurate and real-time estimation of power factor (PF) and phase angle (PA) is critical for tracking energy efficiency and monitoring grid stability. This study proposes a deep neural network (DNN)-based approach for PF and PA estimation using voltage and current signals in real-time applications. Unlike conventional methods that rely on time-series analysis, the proposed method performs PF estimation using convolutional neural networks (CNNs) for efficient feature extraction by converting 1-D electrical signals into 2-D image representations. The experimental validation demonstrates that the developed model achieves high accuracy across various load conditions, including resistive, capacitive, and inductive loads. The results show that the DNN-based approach provides fast and precise estimations, making it a viable alternative to expensive equipment, such as power analyzers. Results show that the proposed CNN-based model achieved an $R<^>{2}$ value of 0.9683 for PA estimation and 0.8842 for PF estimation, with root mean square error (RMSE) values as low as 6.71 and 0.11, respectively. The system successfully predicted PF and PA with less than 5% error during the experimental test phases. Future works will focus on extending the model's capabilities to handle more complex power quality (PQ) disturbances and integrating it into the proposed energy monitoring system.