Analyzing CO₂ Emissions using Machine Learning and Explainable AI: A Case Study of Natural Gas-Fired Powered Plant in Bangladesh

This study analyzes the CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions from a natural gas-fired power plant in Bangladesh, focusing on the close interconnections among gas consumption, electricity generation, and fuel-related emissions. Using historical data from 2015 to 2022, we employ various forecasting models—including Seasonal Autoregressive Integrated Moving Average (SARIMA), AutoRegressive Integrated Moving Average (ARIMA), Prophet models, and machine learning techniques such as Support Vector Regression (SVR)—alongside feature engineering methods to enhance prediction accuracy, model fitness. Our results reveal a strong positive correlation (0.99) between gas consumption, power generation, and CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions, confirming that gas consumption is the primary contributor. The SVR model is well fitted with training R2 value of 0.9820 and test R2 value of 0.9842.Time series forecasting models predict a downward trend in CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions from 2023 to 2026, suggesting potential improvements in fuel efficiency and a shift towards alternative energy sources. Accurate emission estimates are critical for stakeholders involved in policy initiatives that target short, medium, and long-term emission reductions.