上海華瑞眾信技術有限公司 張正芳，田海濤

摘要：為解決工業場景下能碳數據時序性分析滯后、邊緣端實時決策能力不足、系統部署成本高等問題，本研究提出將長短期記憶網絡（LSTM）技術與能碳融合一體機結合的數字化能碳管理中心建設方案。首先闡述LSTM網絡的門控機制與時序數據處理優勢，通過對比分析其與Transformer等架構在時間復雜度、硬件適配性等維度的差異，驗證其在邊緣計算場景的適用性；其次設計能碳融合一體機的硬件架構與軟件功能模塊，集成數據采集、LSTM模型推理、實時決策等核心能力；最后通過能耗預測、碳排放核算、可再生能源生產預測及設備故障預警四大典型應用場景的實證分析，驗證該方案在預測準確率（準確率＞85%）及部署靈活性方面的技術優勢。研究結果表明， LSTM技術與能碳融合一體機的深度融合，可實現能碳管理的數字化、智能化與邊緣部署落地，為工業企業“雙碳”目標達成提供高效技術支撐。

關鍵詞：LSTM；能碳融合一體機；邊緣計算；數字化能碳管理；時序數據建模

Abstract: To address issues such as lagging time-series analysis of energy and carbon data, insufficient real-time decision-making capabilities at the edge, and high system deployment costs in industrial scenarios, a construction plan for a digital energy and carbon management center is proposed, integrating Long Short-Term Memory (LSTM) technology with an energy and carbon fusion all-in-one machine. Firstly, the gating mechanism of the LSTM network and its advantages in time-series data processing are elaborated. By comparing and analyzing its differences from architectures such as Transformer in terms of time complexity and hardware adaptability, its applicability in edge computing scenarios is verified. Secondly, the hardware architecture and software functional modules of the energy and carbon fusion all-in-one machine are designed, integrating core capabilities such as data collection, LSTM model inference, and real-time decision-making. Finally, through empirical analysis of four typical application scenarios: energy consumption prediction, carbon emission accounting, renewable energy production prediction, and equipment failure warning, the technical advantages of this plan in terms of prediction accuracy (accuracy > 85%) and deployment flexibility are verified. The research results show that the deep integration of LSTM technology with the energy and carbon fusion all-in-one machine can achieve digitalization, intelligence, and edge deployment implementation of energy and carbon management, providing efficient technical support for industrial enterprises to achieve their "dual carbon" goals.

Key words: LSTM; Energy-Carbon Fusion All-in-One Machine; Edge Computing; Digital Energy-Carbon Management; Temporal Data Modeling Translation

在線預覽：LSTM技術在數字化能碳管理中心建設中的應用研究.pdf

摘自《自動化博覽》2026年第一期暨《2026具身智能專刊》