目前台灣能源依賴進口的比例高達98%，因此提升能源的自主性和多元化變得極為重要。多年來，政府積極推動「開源節流」的政策，其中虛擬電廠(Virtual Power Plant, VPP)就是其中一項核心技術，虛擬電廠通過整合分散式能源來改善用電供應與需求。而智慧電表系統作為VPP的核心基礎設施，不僅能夠即時收集和傳輸用戶用電數據，還可以為用戶和用電管理者提供全面的能源監控和利於分析的數據庫。
此外，隨著智慧電表普及，大量用電數據反映家庭用電動態變化，為預測用電狀況與異常檢測提供了基礎。然而，要如何從如此龐大的數據中提前識別過載、短路等潛在的電氣危險，目前仍是個挑戰。
因此，本研究提出了一種解決方案，結合大型語言模型(Large Language Models, LLM)技術，利用其數據處理和模式辨識能力，來預測即將發生的電氣異常事故。研究中使用智慧電表的歷史用電數據進行模型訓練，將數值數據編碼為文本形式，並設計異常檢測系統，用於分析即時用電使用數據中的潛在危險模式。實驗結果顯示，本系統能夠有效率的預測電氣異常事故。
此研究對於能源管理和電氣安全領域具有重要貢獻，提供了一種可行的早期預警機制。透過結合智慧電表數據與LLM，提升了電網安全性，降低事故發生的風險；家庭用戶則可及時採取預防措施，減少財產損失與人身危害。此外，此方法亦可作為未來智慧電網技術的重要組成部分，推動能源安全與永續發展目標的實現。

Taiwan currently relies on imports for 98% of its energy, highlighting the critical need to enhance energy autonomy. Over the years, the government has promoted the "open source and conservation" policy, with the Virtual Power Plant (VPP) emerging as a core technology. VPPs improve the balance between electricity supply and demand by integrating distributed energy resources. As the foundational infrastructure of VPPs, smart meter systems enable real-time collection and transmission of users’ electricity consumption data but also provide energy monitoring and a rich data repository for both users and energy managers.
With the widespread adoption of smart meters, large-scale electricity consumption data reflects patterns of household usage and serves as a basis for consumption forecasting and anomaly detection. However, identifying potential electrical hazards such as overloads and short circuits from massive datasets remains a challenge. This study proposes a solution integrating Large Language Models (LLMs), leveraging their capabilities in data processing and pattern recognition to predict electrical anomalies. The model is trained on historical smart meter data, with numerical information encoded into textual form. An anomaly detection system is designed to identify hazard patterns from real-time consumption data.
This research contributes significantly to the fields of energy management and electrical safety by offering a viable early warning mechanism. Through the integration of smart meter data and LLMs, grid safety is enhanced, the risk of accidents is reduced, and household users are empowered to take timely preventive actions to mitigate property loss and personal harm. The proposed approach has the potential to become a key component of future smart grid technologies, promoting energy security and the realization of sustainable development goals.

[1] 台灣電力公司(2024年8月)。《台電月刊》第740期。檢索日期：2025年2月20日，取自 https://service.taipower.com.tw/tpcjournal/article/7174
[2] 國家發展委員會, (2022). 臺灣2050淨零排放路徑及策略總說明 檢索日期2022年12 月20日, 取自https://www.ndc.gov.tw/Content_List.aspx?n=DEE68AAD8B38BD76。
[3] Baduge, S. K., Thilakarathna, S., Perera, J. S., Arashpour, M., Sharafi, P., Teodosio, B., Shringi, A., & Mendis, P. (2022). Artificial intelligence and smart vision for building and construction 4.0: Machine and deep learning methods and applications. Automation in Construction, 141, 104440.
[4] Bakumenko, A., Hlaváčková-Schindler, K., Plant, C., & Hubig, N. C. (2024). Advancing Anomaly Detection: Non-Semantic Financial Data Encoding with LLMs. arXiv preprint arXiv:2406.03614.
[5] Bhuiyan, E. A., Hossain, M. Z., Muyeen, S. M., Fahim, S. R., Sarker, S. K., & Das, S. K. (2021). Towards next generation virtual power plant: Technology review and frameworks. Renewable and Sustainable Energy Reviews, 150, 111358. https://doi.org/https://doi.org/10.1016/j.rser.2021.111358
[6] Braei, M., & Wagner, S. (2020). Anomaly detection in univariate time-series: A survey on the state-of-the-art. arXiv 2020. arXiv preprint arXiv:2004.00433.
[7] Chang, C.H., Chuang, M.L., Tan, J.C., Hiseh, C.C., and Chou, C.C. (2022, Nov). Indoor Safety Monitoring for Falls or Restricted Areas Using Wi-Fi Channel State Information and Deep Learning Methods in Mega Building Construction Projects. Sustainability, 14(22), 15034.
[8] Chang, C.H., Tan, J.C., Wang, R.G., Wu, P.Y., and Chou, C.C. (2020, Sep). Sound-based indoor positioning for rescue using deep learning, building information models and virtual reality. Journal of the Chinese Institute of Civil and Hydraulic Engineering, 32(5), 383-392.
[9] GATTAL, R. (2024). LLM Based Approach for Anomaly Detection in Smart Grids Université de Echahid Cheikh Larbi Tébessi–Tébessa-].
[10] Gruver, N., Finzi, M., Qiu, S., & Wilson, A. G. (2023). Large language models are zero-shot time series forecasters. Advances in neural information processing systems, 36, 19622-19635.
[11] Hsieh, C.C., Liu, C.Y., Wu, P.Y., Jeng, A.P., Wang, R.G., and Chou, C.C. (2019, Jun). Building information modeling services reuse for facility management for semiconductor fabrication plants. Automation in Construction, 102, 270-287.
[12] Hwang, J., Suh, D., & Otto, M.-O. (2020). Forecasting Electricity Consumption in Commercial Buildings Using a Machine Learning Approach. Energies, 13(22), 5885. https://www.mdpi.com/1996-1073/13/22/5885
[13] Jeng, A.P., Wang, R.G., Wu, P.Y., Tai, J.K., Tan, J.C., and Chou, C.C. (2019, Apr). Deep learning-based smart meter data analytics for early warning of possible electrical fires. Journal of the Chinese Institute of Civil and Hydraulic Engineering, 31(2), 193-204.
[14] Jin, M., Wang, S., Ma, L., Chu, Z., Zhang, J. Y., Shi, X., Chen, P.-Y., Liang, Y., Li, Y.-F., & Pan, S. (2023). Time-llm: Time series forecasting by reprogramming large language models. arXiv preprint arXiv:2310.01728.
[15] Jin, M., Wen, Q., Liang, Y., Zhang, C., Xue, S., Wang, X., Zhang, J., Wang, Y., Chen, H., & Li, X. (2023). Large models for time series and spatio-temporal data: A survey and outlook. arXiv preprint arXiv:2310.10196.
[16] Keskes, M. I. (2025). Generative Adversarial Networks for Synthetic Data Generation in Deep Learning Applications.
[17] Li, B., & Ghiasi, M. (2021). A new strategy for economic virtual power plant utilization in electricity market considering energy storage effects and ancillary services. Journal of Electrical Engineering & Technology, 16(6), 2863-2874.
[18] Liu, C., Yang, R. J., Yu, X., Sun, C., Wong, P. S. P., & Zhao, H. (2021). Virtual power plants for a sustainable urban future. Sustainable Cities and Society, 65, 102640. https://doi.org/https://doi.org/10.1016/j.scs.2020.102640
[19] Liu, C.Y., Wu, P.Y., Tan, J.C., Chuang, M.L., Wang, R.G., and Chou, C.C. (2021, Dec). An automatic fire rescue deployment method using building information modeling and ontology: Taking long-term care institutes as an example. Journal of the Chinese Institute of Civil and Hydraulic Engineering, 33(8), 641-651.
[20] Liu, M., Zhang, L., Chen, J., Chen, W.-A., Yang, Z., Lo, L. J., Wen, J., & O’Neill, Z. (2025). Large language models for building energy applications: Opportunities and challenges. Building Simulation,
[21] Liu, X., Ding, Y., Tang, H., & Xiao, F. (2021). A data mining-based framework for the identification of daily electricity usage patterns and anomaly detection in building electricity consumption data. Energy and Buildings, 231, 110601. https://doi.org/https://doi.org/10.1016/j.enbuild.2020.110601
[22] Liu, X., Golab, L., Golab, W., Ilyas, I. F., & Jin, S. (2017). Smart Meter Data Analytics. ACM Transactions on Database Systems, 42(1), 1-39. https://doi.org/10.1145/3004295
[23] Liu, Y., Li, Z., Pan, S., Gong, C., Zhou, C., & Karypis, G. (2021). Anomaly detection on attributed networks via contrastive self-supervised learning. IEEE transactions on neural networks and learning systems, 33(6), 2378-2392.
[24] Mahjoub, S., Chrifi-Alaoui, L., Marhic, B., & Delahoche, L. (2022). Predicting energy consumption using LSTM, multi-layer GRU and drop-GRU neural networks. Sensors, 22(11), 4062.
[25] Naval, N., & Yusta, J. M. (2021). Virtual power plant models and electricity markets - A review. Renewable and Sustainable Energy Reviews, 149, 111393. https://doi.org/https://doi.org/10.1016/j.rser.2021.111393
[26] Rawan Elhadad, Y.-F. T. W.-N. T. (2022). Anomaly Prediction in Electricity Consumption Using a Combination of Machine Learning Techniques. International Journal of Technology, 13(6), 291-319. https://doi.org/https://doi.org/10.14716/ijtech.v13i6.5931
[27] Rick, R., & Berton, L. (2022). Energy forecasting model based on CNN-LSTM-AE for many time series with unequal lengths. Engineering Applications of Artificial Intelligence, 113, 104998.
[28] Rouzbahani, H. M., Karimipour, H., & Lei, L. (2021). A review on virtual power plant for energy management. Sustainable Energy Technologies and Assessments, 47, 101370. https://doi.org/https://doi.org/10.1016/j.seta.2021.101370
[29] Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., & Bolton, A. (2017). Mastering the game of go without human knowledge. nature, 550(7676), 354-359.
[30] Tan, J.C., Chuang, M.L., Chang, C.H., Wang, R.G., and Chou, C.C. (2022, Nov). Preliminary exploration of applying agile project management to compound disaster preparation: Taking the case of finding agile consultants as an example. Journal of the Chinese Institute of Civil and Hydraulic Engineering, 34(7), 649-661.
[31] Su, J., Jiang, C., Jin, X., Qiao, Y., Xiao, T., Ma, H., Wei, R., Jing, Z., Xu, J., & Lin, J. (2024). Large language models for forecasting and anomaly detection: A systematic literature review. arXiv preprint arXiv:2402.10350.
[32] Tai, J.K., Liu, C.Y., Chu, C.P., Wang, S.J., Wang, R.G., and Chou, C.C. (2023, Nov). Automated Spatial Configuration for Indoor Refugee Shelters Using Building Information Modeling and Ontology. Journal of the Chinese Institute of Civil and Hydraulic Engineering, 35(7), 669-680.
[33] Wang, L., Yang, N., Huang, X., Yang, L., Majumder, R., & Wei, F. (2023). Improving text embeddings with large language models. arXiv preprint arXiv:2401.00368.
[34] Wang, R.G., Chuang, M.L., Ke, C.Y., Chien, Y.F., Ho, W.J., Chiang, K.C., Hung, Y.C., Tsai, C.H., and Chou, C.C. (2024, Dec). Predicting Imminent Electrical Safety Incidents Using Smart Meter Big Data With Large Language Models. IEEE Access, 12, 184940.
[35] Wang, R.G., Ho, W.J., Chiang, K.C., Hung, Y.C., Tai, J.K., Tan, J.C., Chuang, M.L., Ke, C.Y., Chien, Y.F., Jeng, A.P., and Chou, C.C. (2023, Sep). Analyzing Long-Term and High Instantaneous Power Consumption of Buildings from Smart Meter Big Data with Deep Learning and Knowledge Graph Techniques. Energies, 16, 6893.
[36] Wang, R.G., Wu, P.Y., Liu, C.Y., Tan, J.C., Chuang, M.L., and Chou, C.C. (2022, Jul). Route Planning for Fire Rescue Operations in Long‐term Care Facilities Using Ontology and Building Information Models. Buildings, 12, 1060.
[37] Wang, S.J., Wang, R.G., Ke, C.Y., Chien, Y.F., Chuang, M.L., and Chou, C.C. (2024, Nov). Exploring and Applying Digital Transformation in the Transportation of Prefabricated Construction Components. Journal of the Chinese Institute of Civil and Hydraulic Engineering, 36(5), 451-459.
[38] Wang, Y., Chen, Q., Hong, T., & Kang, C. (2019). Review of Smart Meter Data Analytics: Applications, Methodologies, and Challenges. IEEE Transactions on Smart Grid, 10(3), 3125-3148. https://doi.org/10.1109/tsg.2018.2818167
[39] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.
[40] Viana, M. S., Ramos, D. S., Manassero Junior, G., & Udaeta, M. E. M. (2023). Analysis of the Implementation of Virtual Power Plants and Their Impacts on Electrical Systems. Energies, 16(22), 7682.
[41] Zekić-Sušac, M., Mitrović, S., & Has, A. (2021). Machine learning based system for managing energy efficiency of public sector as an approach towards smart cities. International journal of information management, 58, 102074.
[42] Zhang, J. E., Wu, D., & Boulet, B. (2021). Time series anomaly detection for smart grids: A survey. 2021 IEEE electrical power and energy conference (EPEC),
[43] Zhu, J., Cai, S., Deng, F., Ooi, B. C., & Wu, J. (2024). Do LLMs Understand Visual Anomalies? Uncovering LLM's Capabilities in Zero-shot Anomaly Detection.