Electrical Equipment Status Monitoring Based on Bidirectional Long Short-Term Memory with Adaptive Moment Estimation

In recent days, electrical equipment status monitoring is vital for ensuring the reliability, security, and resourceful function of industrial plants, power systems, and manufacturing amenities. Previous researchers suggested several traditional methods but failed to detect early signs of failures due to the complexity and diversity of electrical systems. Therefore, this research suggested a hybrid model using Bidirectional Long Short-Term Memory (Bi-LSTM) with Adaptive Moment Estimation (Adam) optimization model namely called as BiLSTM-Adam. Initially, the data is collected from various sensors. Then, a noise reduction model namely spectral gating is employed as it assists in mitigating the irrelevant data from the collected dataset. Further, optimal features are extracted using Empirical Mode Decomposition (EMD) technique, condition indications such as Total Harmonic Distortion (THD) and Root Mean Square (RMS). Finally, electrical equipment status monitoring is accomplished using Bi-LSTM with Adam optimizer as it has ability to handle variable-length sequences and flexibility across applications. This integration improves the ability to robustly monitor and diagnose the operational status of electrical equipment. The proposed Bi-LSTM-Adam model obtained greater results with an accuracy of 99.98% when compared to existing you only look once version 4 (YOLO v4) model.