Publication: Enhancing Power Plants Safety by Accurately Predicting CO and NOx Leakages from Gas Turbines Using FFNN and LSTM Neural Networks
Date
2023
Authors
Yousif S.T.
Alnaimi F.
Bazi A.A.
Thiruchelvam S.
Journal Title
Journal ISSN
Volume Title
Publisher
Cal-Tek srl
Abstract
Gas power plants are fast-establishing power plants capable of producing reliable energy in high watts volumes. One of its significant features is its dependency on natural air as raw material to run the gas turbine. Air passes through several stages that involve heating the air to increase its pressure before being used in electric power generation. Leakage in gas power stations is considered a vital indication of irregular processes of those stages. Any fault existing in the meanwhile operations can result in lousy production performance. Considering the human and economic losses of gas leakage, it has become a challenge to prevent the same. One of the essential approaches to managing gas leakage reduction is an accurate prediction. This paper proposes an automatic prevention approach relying on deep learning technology for predicting gas leakage status. Furthermore, a novel dataset was supplied by a natural gas power plant to predict CO and NOx emissions. The dataset is used to train the deep learning models using Long-short Term Memory and Feed-Forward Neural Networks. The optimum accuracy obtained is over 92% for CO and over 58% for NOx while using the LSTM model as a predictor. � 2023 The Authors.
Description
Keywords
Deep Learning , FFNN , Gas Leakage , LSTM , Feedforward neural networks , Forecasting , Gas emissions , Gas plants , Gas turbines , Long short-term memory , Losses , Nitrogen oxides , Deep learning , FFNN , Gas leakages , Gas power plants , LSTM , Natural air , Neural-networks , Plant safety , Power station , Reliable energy , Power generation