Implementing the Potential of Automatic Detection of Power Quality Disruption using Deep Learning Techniques and its Approaches

A variety of energy sources are being included in the existing power grid to meet the increasing electricity demand. However, this increases the possibility of Power Quality (PQ)deterioration during processes like production, and trans- mission. There has been a significant uptick in research efforts devoted to PQ in recent years. Harmonics in energy and spikes or drops in voltage must be investigated to ensure PQ. All these abnormalities are referred to as power quality disturbances (PQD). To reduce the danger, it is necessary to detect and classify the PQD. The standard method separates this task into several distinct phases, including signal processing, feature extraction, selection, and feature-based categorization. Unfortunately, there are flaws in the signal processing approach, and selecting features manually is time-consuming and inaccurate. For this reason, conventional models are less effective at classifying PQD and are more susceptible to noise. In this research, we present the Deep Learning (DL) framework as an alternative to the existing approaches. A wide variety of signal data with various PQDs are produced using MATLAB. Using DL models such as Long Short-Term Memory (LSTM), Autoencoder (AE), and a hybrid of Long Short-Term Memory with Autoencoder, the PQD are autonomously recognized and classified by the type. Finally, the comparative findings of all three DL models are presented. With an accuracy of 95.3%, the proposed hybrid model (LSTM- Autoencoder) is discovered to be an effective PQD classifier. Keywords— x