Scholarlink

<p>Turkey’s growing electricity demand necessitates improved combustion efficiency in power plants, making accurate prediction of NOx and CO emissions critical for environmental sustainability. Previous approaches have been hampered by high error rates, model overfitting, and poor scalability, limiting their effectiveness. Therefore, proposed study developed a robust predictive model for NOx and CO emissions from gas turbine addressing a critical need for enhanced combustion efficiency in power plants. The model employs a Modified Convolutional Neural Network (MCNN) integrated with Bidirectional Long Short-Term Memory (Bi-LSTM) architecture, enhanced with an extrinsic attention mechanism that dynamically weights input features based on learned context vectors. The model demonstrated exceptional performance with R² values of 0.881 for NOx and 0.801 for CO, significantly outperforming existing approaches. Both mean absolute error and root mean square error were substantially reduced compared to previous models, with improvements of approximately 22% and 18% respectively. The extrinsic attention mechanism effectively identified critical emission factors, particularly highlighting the importance of combustion temperature and fuel-air ratio as primary predictors. In real-world applications, the findings can inform power plant operators and policymakers in implementing better emission control strategies, optimizing combustion processes, and complying with environmental regulations. These findings provide power plant operators with a reliable tool for emission prediction and control, can enhance operational efficiency, reduce emissions, and contribute to sustainable energy practices.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Emission reduction in gas turbine: improving CO and NOx emission prediction using modified CNN-Bi-LSTM extrinsic attention regressor

  • 期刊论文

  • 发表日期：

    2026年1月17日

查看全文
  • Atanu Roy,
  • Sabyasachi Pramanik,
  • Kalyan Mitra,
  • Manashi Chakraborty

摘要

Turkey’s growing electricity demand necessitates improved combustion efficiency in power plants, making accurate prediction of NOx and CO emissions critical for environmental sustainability. Previous approaches have been hampered by high error rates, model overfitting, and poor scalability, limiting their effectiveness. Therefore, proposed study developed a robust predictive model for NOx and CO emissions from gas turbine addressing a critical need for enhanced combustion efficiency in power plants. The model employs a Modified Convolutional Neural Network (MCNN) integrated with Bidirectional Long Short-Term Memory (Bi-LSTM) architecture, enhanced with an extrinsic attention mechanism that dynamically weights input features based on learned context vectors. The model demonstrated exceptional performance with R² values of 0.881 for NOx and 0.801 for CO, significantly outperforming existing approaches. Both mean absolute error and root mean square error were substantially reduced compared to previous models, with improvements of approximately 22% and 18% respectively. The extrinsic attention mechanism effectively identified critical emission factors, particularly highlighting the importance of combustion temperature and fuel-air ratio as primary predictors. In real-world applications, the findings can inform power plant operators and policymakers in implementing better emission control strategies, optimizing combustion processes, and complying with environmental regulations. These findings provide power plant operators with a reliable tool for emission prediction and control, can enhance operational efficiency, reduce emissions, and contribute to sustainable energy practices.

Graphical abstract