Investigating and predicting the dielectric performance of non-edible Natural Ester using LSTM-based deep learning model

• Published in: 2023 International Conference on Power, Instrumentation, Energy and Control (PIECON) pp. 1 - 6
• Main Authors: Raj, Raymon Antony, Murugesan, Srinivasan, Sarathi, R., D, Sarathkumar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10.02.2023
• Subjects: Breakdown voltage; Deep learning; deep learning studies; dielectric breakdown voltage; LSTM; Mathematical models; natural ester; Oils; Predictive models; Solid modeling; survival; Voltage measurement
• DOI: 10.1109/PIECON56912.2023.10085810

Long short-term memory (LSTM) network deep learning research have become an effective method for foreseeing responses with respect to time. A neural network model called LSTM uses a time series or a data sequence to predict future occurrences. The breakdown voltage of non-edible natural ester compounds like pongamia oil (PO) and its modified counterpart, MPO, is predicted by this machine learning model. The 500 observations of the 30-day measurement of the Dielectric Breakdown Voltage for the PO and MPO are recorded according to IEC 60156 standards. These observations serve as the input for the LSTM-based deep learning model created in MATLAB. When compared to the other prediction model, the LSTM model predicts outcomes quite well. The RMSE and loss of the LSTM model show less divergence from the forecast of MPO's Dielectric Breakdown Voltage than do those of other prediction techniques. MPO limitations demonstrate its longer lifespan. As a result, MPO continues to have better dielectric characteristics.