<p>Taiwan’s energy landscape, characterized by a reliance on coal, natural gas, nuclear power, renewable energy, and fuel oil, reflects a strategic diversification aligned with global sustainability goals. In light of escalating greenhouse gas (GHG) emissions, which primarily originate from fossil fuel consumption, electricity production has been identified as the principal contributor, accounting for over 50% of annual emissions. To address this critical issue, our study employs state-of-the-art deep learning methodologies, specifically artificial neural networks (ANN) and bidirectional long short-term memory (BiLSTM) networks, to predict GHG emissions based on the composition and evolution of Taiwan’s power generation mix. Anchored in data from 2013 to 2021, this period corresponds with the implementation of Taiwan’s Carbon Footprint Verification initiative, which reflects a growing commitment to environmental accountability. Our comprehensive analysis evaluates the predictive performance of these models using robust statistical metrics, including RMSE, MBE, rRMSE, <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$R^2$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>, and MAPE. Notably, the performance indicators achieved with the ANN model are compelling. The <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$R^2$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> value of 0.9851 reflects strong predictive correlation, the RMSE of 0.0392 indicates high precision, and the MBE of <InlineEquation ID="IEq3"> <EquationSource Format="TEX">$-$</EquationSource> <EquationSource Format="MATHML"><math> <mo>-</mo> </math></EquationSource> </InlineEquation>0.0378 reveals minimal bias. In addition, the exceptionally low rRMSE and MAPE values, recorded at 0.3157% and 0.3062% respectively, further validate the model’s forecasting accuracy. This work not only highlights the increasing trend of GHG emissions but also demonstrates the effectiveness of predictive models as powerful analytical tools within the energy sector. These models support informed policy-making and strategic planning, encouraging cleaner energy production in Taiwan while providing a replicable framework for GHG emission estimation in other geopolitical contexts.</p>

Predictive Modeling to Mitigate Greenhouse Gas Emissions in Taiwan’s Diversified Energy Landscape

期刊论文
发表日期：

2025年10月8日

查看全文

Sze-Teng Liong,
Gen-Bing Liong,
Y. S. Gan

摘要

Taiwan’s energy landscape, characterized by a reliance on coal, natural gas, nuclear power, renewable energy, and fuel oil, reflects a strategic diversification aligned with global sustainability goals. In light of escalating greenhouse gas (GHG) emissions, which primarily originate from fossil fuel consumption, electricity production has been identified as the principal contributor, accounting for over 50% of annual emissions. To address this critical issue, our study employs state-of-the-art deep learning methodologies, specifically artificial neural networks (ANN) and bidirectional long short-term memory (BiLSTM) networks, to predict GHG emissions based on the composition and evolution of Taiwan’s power generation mix. Anchored in data from 2013 to 2021, this period corresponds with the implementation of Taiwan’s Carbon Footprint Verification initiative, which reflects a growing commitment to environmental accountability. Our comprehensive analysis evaluates the predictive performance of these models using robust statistical metrics, including RMSE, MBE, rRMSE, $R^2$ $R^{2}$ , and MAPE. Notably, the performance indicators achieved with the ANN model are compelling. The $R^2$ $R^{2}$ value of 0.9851 reflects strong predictive correlation, the RMSE of 0.0392 indicates high precision, and the MBE of $-$ $-$ 0.0378 reveals minimal bias. In addition, the exceptionally low rRMSE and MAPE values, recorded at 0.3157% and 0.3062% respectively, further validate the model’s forecasting accuracy. This work not only highlights the increasing trend of GHG emissions but also demonstrates the effectiveness of predictive models as powerful analytical tools within the energy sector. These models support informed policy-making and strategic planning, encouraging cleaner energy production in Taiwan while providing a replicable framework for GHG emission estimation in other geopolitical contexts.