The paper titled ‘Data‑Driven Modeling for the Prediction of Stack Gas Concentration in a Coal‑Fired Power Plant in Türkiye’ has been published by Prof. Didem Saloğlu Dertli in the Water Air and Soil Pollution.

Deep learning and machine learning methods were utilized in Türkiye to forecast stack gas concentrations in coal-fired power plants using real-time data from Continuous Emission Monitoring Systems (CEMS). Models such as Long Short-Term Memory (LSTM), Multi-Layer Perceptron (MLP), Light Gradient Boosted Machine (LightGBM), and Stochastic Gradient Descent (SGD) were evaluated based on their accuracy in predicting key emissions, including CO, SO₂, NOₓ, O₂, and dust levels.

The findings revealed LSTM as the most effective model for predicting NOₓ and SO₂ (R² = 0.87 and 0.85), while LightGBM performed best for O₂ (R² = 0.85). Both models demonstrated strong results for dust concentrations (R² = 0.78). This research highlights the potential of AI-driven models to optimize emission management, offering critical tools for reducing environmental impact and enhancing operational efficiency in energy production.