在能源轉型壓力與電力需求並存的臺灣，火力發電是否加劇在地空氣污染長期備受爭議。本研究聚焦於首座全面改用天然氣複循環機組的通霄發電廠，透過結合地理人工智慧(Geo-AI)與集成機器學習技術，評估其對新竹-苗栗空品區PM2.5與NO2濃度之貢獻。
研究首先整合逐時發電量、NOx排放、氣象、土地利用、衛星遙測等六大類資料，建置涵蓋約24萬筆紀錄、逾500個變數的資料庫；並依前1–12小時延遲效應衍生12組資料集，挑選最佳滯後期作為建模基礎。在模型面向，採XGBR、LightGBM、GBR、CBR與Random Forest五種演算法，僅保留訓練集R2≧0.7之模型輸出，再以加權集成方式產出50 m × 50 m、逐時濃度推估圖。模型效能經外部驗證、十折交叉、時間/空間分層及極端值測試多重檢驗，對PM2.5與NO2分別取得模型解釋力為0.87與0.84，RMSE分別約3 µg/m3與1.8 ppb，且在夏季低濃度與前10 %極端高濃度樣本下仍維持R2≧0.74的穩定表現。
SHAP分析顯示，通霄發電廠排放對區域NO2的平均貢獻約11.8 %，遠高於對PM2.5的2.5 %，凸顯天然氣機組氮氧化物排放仍需優先管控。兩起代表性高濃度事件的時空推估亦呈現「沿海至市區」的帶狀傳輸特徵，模型與觀測高度吻合，驗證其解析熱點與機制之能力。
儘管本研究建立了細尺度、逐時的空氣品質推估模型，仍存在三項限制：其一，所使用之電廠排放資料僅涵蓋排放前源頭，未納入排放後在大氣中的轉化與擴散過程，導致電廠貢獻的估計偏向保守；其二，竹苗區監測站分布不均，山區資料不足，使模型於特定區位的準確性有限；其三，資料時段僅涵蓋 2022至2023 年，難以反映長期趨勢與極端事件影響。未來建議擴充資料至十年以上並涵蓋全臺範圍，納入更多衛星與感測器資料，強化模型在不同地形與時間尺度下的適用性與預測能力。
綜合以上，本研究開發之細尺度、逐時的暴露評估模型，首度量化了天然氣發電廠對在地空氣品質的即時影響，可為機組調度、精準減排與健康風險管理提供科學依據；亦為後續建置全臺即時預警與決策支援系統奠定技術基礎。

This study integrates Geospatial Artificial Intelligence (Geo-AI) with Machine Learning to quantify how Taiwan Power Company (TPC) operations affect PM2.5 and NO2 across the Hsinchu–Miaoli Air-Quality Zone. We merged six data domains, including Environmental Protection Administration (EPA) monitoring records, meteorological factors, satellite data, land-use database, and power-plant generation and emission data, yielding more than 500 predictors. After evaluating Lag effect, we selected a one-hour-lagged dataset for model training. Five algorithms: Extreme Gradient Boosting, Light Gradient Boosting, Gradient Boosting, CatBoost, and Random Forest, were first trained separately; predictions from models with training R2 ≥ 0.70 were then performance-weighted to build an Ensemble Model. The resulting models achieved R2 = 0.87 for PM2.5 and R2 = 0.84 for NO2, outperforming any single algorithm and maintaining R2 > 0.70 under external, cross-fold, temporal, and spatial validations. SHapley Additive exPlanations (SHAP) indicate that emissions from the Tongxiao Power Plant contribute on average 11.8 % of NO2 but only 2.5 % of PM2.5, underscoring the continuing importance of nitrogen-oxide control even after the plant’s transition to natural gas. Our findings confirm the high potential of Geo-AI fused with power-generation data for fine-scale air-quality forecasting and provide a science-based foundation for air-quality management and energy-dispatch policies.

Achakulwisut, P., Brauer, M., Hystad, P., & Anenberg, S. C. (2019). Global, national, and urban burdens of paediatric asthma incidence attributable to ambient NO2 pollution: estimates from global datasets. The Lancet Planetary Health, 3(4), e166-e178.
Abdillah, S. F. I., Yang, T.-W., You, S.-J., & Wang, Y.-F. (2025). Assessing trade-off between carbon and flue gas emissions in power plants: Influences of air pollution control devices, fuel types, and combustion reactors. Journal of Environmental Chemical Engineering, 13(2), 115707.
Asri, A. K., Chang, H.-T., Yu, C.-P., Liu, W.-Y., Chern, Y.-R., Xie, R.-H., Lung, S.-C. C., Chi, K. H., Chen, Y.-C., Cheng, S.-S., Adamkiewicz, G., Spengler, J. D., & Wu, C.-D. (2025). Spatiotemporal estimation of ambient forest phytoncides: Unveiling patterns through geospatial-based machine learning approach. Ecological Indicators, 175, 113526.
Babaan, J., Hsu, F.-T., Wong, P.-Y., Chen, P.-C., Guo, Y.-L., Lung, S.-C. C., Chen, Y.-C., & Wu, C.-D. (2023). A Geo-AI-based ensemble mixed spatial prediction model with fine spatial-temporal resolution for estimating daytime/nighttime/daily average ozone concentrations variations in Taiwan. Journal of Hazardous Materials, 446, 130749.
Babaan, J., Wong, P.-Y., Chen, P.-C., Chen, H.-L., Lung, S.-C. C., Chen, Y.-C., & Wu, C.-D. (2024). Geospatial artificial intelligence for estimating daytime and nighttime nitrogen dioxide concentration variations in Taiwan: A spatial prediction model. Journal of Environmental Management, 360, 121198.
Chen, G., Guo, Y., Yue, X., Tong, S., Gasparrini, A., Bell, M. L., Armstrong, B., Schwartz, J., Jaakkola, J. J. K., Zanobetti, A., Lavigne, E., Nascimento Saldiva, P. H., Kan, H., Royé, D., Milojevic, A., Overcenco, A., Urban, A., Schneider, A., Entezari, A., . . . Li, S. (2021). Mortality risk attributable to wildfire-related PM2.5 pollution: a global time series study in 749 locations. The Lancet Planetary Health, 5(9), e579-e587.
Chio, C. P., Lo, W. C., Tsuang, B. J., Hu, C. C., Ku, K. C., Wang, Y. S., Chen, Y. J., Lin, H. H., & Chan, C. C. (2022). County-Wide Mortality Assessments Attributable to PM2.5 Emissions from Coal Consumption in Taiwan. Int J Environ Res Public Health, 19(3).
Di, Q., Amini, H., Shi, L., Kloog, I., Silvern, R., Kelly, J., Sabath, M. B., Choirat, C., Koutrakis, P., Lyapustin, A., Wang, Y., Mickley, L. J., & Schwartz, J. (2020). Assessing NO2 Concentration and Model Uncertainty with High Spatiotemporal Resolution across the Contiguous United States Using Ensemble Model Averaging. Environmental Science & Technology, 54(3), 1372-1384.
Guo, Y., Gasparrini, A., Armstrong, B. G., Tawatsupa, B., Tobias, A., Lavigne, E., Coelho, M., Pan, X., Kim, H., Hashizume, M., Honda, Y., Guo, Y. L., Wu, C. F., Zanobetti, A., Schwartz, J. D., Bell, M. L., Scortichini, M., Michelozzi, P., Punnasiri, K., . . . Tong, S. (2017). Heat Wave and Mortality: A Multicountry, Multicommunity Study. Environ Health Perspect, 125(8), 087006.
Guo, Q., He, Z., & Wang, Z. (2023). Predicting of Daily PM2.5 Concentration Employing Wavelet Artificial Neural Networks Based on Meteorological Elements in Shanghai, China. Toxics, 11(1).
He, C., Kim, H., Hashizume, M., Lee, W., Honda, Y., Kim, S. E., Kinney, P. L., Schneider, A., Zhang, Y., Zhu, Y., Zhou, L., Chen, R., & Kan, H. (2022). The effects of night-time warming on mortality burden under future climate change scenarios: a modelling study. The Lancet Planetary Health, 6(8), e648-e657.
He, S., Dong, H., Zhang, Z., & Yuan, Y. (2022). An Ensemble Model-Based Estimation of Nitrogen Dioxide in a Southeastern Coastal Region of China. Remote Sensing, 14(12), 2807.
Hsu, C.-Y., Xie, H.-X., Wong, P.-Y., Chen, Y.-C., Chen, P.-C., & Wu, C.-D. (2022). A mixed spatial prediction model in estimating spatiotemporal variations in benzene concentrations in Taiwan. Chemosphere, 301, 134758.
Huang, C., Sun, K., Hu, J., Xue, T., Xu, H., & Wang, M. (2022). Estimating 2013-2019 NO2 exposure with high spatiotemporal resolution in China using an ensemble model. Environ Pollut, 292(Pt A), 118285.
Hsu, C.-Y., Wong, P.-Y., Chern, Y.-R., Candice Lung, S.-C., & Wu, C.-D. (2023). Evaluating long-term and high spatiotemporal resolution of wet-bulb globe temperature through land-use based machine learning model. Journal of Exposure Science & Environmental Epidemiology, 34, 941-951.
Hsu, C.-Y., Lee, R.-Q., Wong, P.-Y., Candice Lung, S.-C., Chen, Y.-C., Chen, P.-C., Adamkiewicz, G., & Wu, C.-D. (2024). Estimating morning and evening commute period O3 concentration in Taiwan using a fine spatial-temporal resolution ensemble mixed spatial model with Geo-AI technology. Journal of Environmental Management, 351, 119725.
Khreis, H., Kelly, C., Tate, J., Parslow, R., Lucas, K., & Nieuwenhuijsen, M. (2017). Exposure to traffic-related air pollution and risk of development of childhood asthma: A systematic review and meta-analysis. Environ Int, 100, 1-31.
Li, L., Lurmann, F., Habre, R., Urman, R., Rappaport, E., Ritz, B., Chen, J. C., Gilliland, F. D., & Wu, J. (2017). Constrained Mixed-Effect Models with Ensemble Learning for Prediction of Nitrogen Oxides Concentrations at High Spatiotemporal Resolution. Environ Sci Technol, 51(17), 9920-9929.
Li, L., Zhang, J., Qiu, W., Wang, J., & Fang, Y. (2017). An Ensemble Spatiotemporal Model for Predicting PM2.5 Concentrations. Int J Environ Res Public Health, 14(5).
Lin, Y.-C., Cheng, F.-Y., Lee, Y.-J., Nguyen, T.-T.-N., Tsai, C.-J., Wen, H.-C., Wu, C.-H., Chang, W.-C., & Huan, C.-C. (2023). Assessments of the Emission Contributions from an Ultra-Supercritical Coal-Fired Power Plant to Ambient PM2.5 in Taiwan. Aerosol and Air Quality Research, 23(12), 230059.
Meng, X., Liu, C., Zhang, L., Wang, W., Stowell, J., Kan, H., & Liu, Y. (2021). Estimating PM2.5 concentrations in Northeastern China with full spatiotemporal coverage, 2005–2016. Remote Sensing of Environment, 253, 112203.
Ma, Z., Dey, S., Christopher, S., Liu, R., Bi, J., Balyan, P., & Liu, Y. (2022). A review of statistical methods used for developing large-scale and long-term PM2.5 models from satellite data. Remote Sensing of Environment, 269, 112827.
Oliveira, M., Slezakova, K., Delerue-Matos, C., Pereira, M. C., & Morais, S. (2019). Children environmental exposure to particulate matter and polycyclic aromatic hydrocarbons and biomonitoring in school environments: A review on indoor and outdoor exposure levels, major sources and health impacts. Environment International, 124, 180-204.
Pozzer, A., Anenberg, S. C., Dey, S., Haines, A., Lelieveld, J., & Chowdhury, S. (2023). Mortality Attributable to Ambient Air Pollution: A Review of Global Estimates. Geohealth, 7(1), e2022GH000711.
Stafoggia, M., Bellander, T., Bucci, S., Davoli, M., de Hoogh, K., de' Donato, F., Gariazzo, C., Lyapustin, A., Michelozzi, P., Renzi, M., Scortichini, M., Shtein, A., Viegi, G., Kloog, I., & Schwartz, J. (2019). Estimation of daily PM10 and PM2.5 concentrations in Italy, 2013–2015, using a spatiotemporal land-use random-forest model. Environment International, 124, 170-179.
Tu, L.-K., Wu, Y.-L., Wang, L.-C., & Chang-Chien, G.-P. (2012). Monitoring and Dispersion Modeling of Polybrominated Diphenyl Ethers (PBDEs) in the Ambient Air of Two Municipal Solid Waste Incinerators and a Coal-fired Power Plant. Aerosol and Air Quality Research, 12(1), 113-122.
Tiotiu, A. I., Novakova, P., Nedeva, D., Chong-Neto, H. J., Novakova, S., Steiropoulos, P., & Kowal, K. (2020). Impact of Air Pollution on Asthma Outcomes. Int J Environ Res Public Health, 17(17).
van Donkelaar, A., Martin, R. V., Li, C., & Burnett, R. T. (2019). Regional Estimates of Chemical Composition of Fine Particulate Matter Using a Combined Geoscience-Statistical Method with Information from Satellites, Models, and Monitors. Environmental Science & Technology, 53(5), 2595-2611.
Vardoulakis, S., Giagloglou, E., Steinle, S., Davis, A., Sleeuwenhoek, A., Galea, K. S., Dixon, K., & Crawford, J. O. (2020). Indoor Exposure to Selected Air Pollutants in the Home Environment: A Systematic Review. Int J Environ Res Public Health, 17(23).
Wang, Y.-F., Chao, H.-R., Wang, L.-C., Chang-Chien, G.-P., & Tsou, T.-C. (2010). Characteristics of Heavy Metals Emitted from a Heavy Oil-Fueled Power Plant in Northern Taiwan. Aerosol and Air Quality Research, 10(2), 111-118.
Wei, J., Li, Z., Pinker, R. T., Wang, J., Sun, L., Xue, W., Li, R., & Cribb, M. (2021). Himawari-8-derived diurnal variations in ground-level PM2.5 pollution across China using the fast space-time Light Gradient Boosting Machine (LightGBM). Atmos. Chem. Phys., 21(10), 7863-7880.
Wu, Y.-S., Tsai, C.-Y., Chang, K.-H., & Chiang, C.-F. (2021). Impact of Air Pollutants Emitted by Taichung Power Plant on Atmospheric PM2.5 in Central Taiwan. Aerosol and Air Quality Research, 21(4), 200358.
Zamani, M., Cao, C., Ni, X., Bashir, B., & Talebiesfandarani, S. (2019). PM2.5 Prediction Based on Random Forest, XGBoost, and Deep Learning Using Multisource Remote Sensing Data. Atmosphere, 10.