发表机构
Tsinghua University（清华大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对高频网络物理系统遥测的多步预测困境，提出物理基础的动态渐近分解框架，通过渐近理论指导的权威转移与正则化，在30天工业数据上实现MAE 0.1764，较LSTM误差降低78.3%，且支持边缘实时部署。
AI 中文摘要
高频工业网络物理系统（CPS）以低于10秒的间隔持续传输传感器遥测数据。预测性维护和运行控制需要跨小时级别的多水平预测。然而，在扩展预测范围（在dt=5秒时H=2160步）上部署机器学习模型遇到了双时间尺度预测困境：短时动力学动量与长时日周期热力学平衡之间的权衡。自回归滚动预测遭受复合误差漂移，而直接多输出投影器则表现出方差爆炸和噪声外推。在本工作中，我们证明了自回归互信息在扩展提前期上指数衰减至零，并且最优最小方差估计器渐近收敛于周期日基线（引理1和定理1）。在这些界限的指导下，我们提出了动态渐近分解（DAD）框架。该架构将连续S形权威转移调度α(h)与单调L2正则化缩放λ(h)耦合，将预测权威从动力学自回归转移到日周期平衡，同时强制物理非负性。在30天工业遥测流（N=518400步）上跨16个前向验证测试窗口进行评估，DAD实现了总体平均绝对误差（MAE）为0.1764，相比深度LSTM（0.8143）误差减少78.3%，相比PatchTST（0.4431）减少60.2%，相比DLinear（0.2227）减少20.8%。计算分析表明，在单个CPU核心上边缘执行时间为0.72毫秒，计算量为0.203 MFLOPs，参数为72k（比Transformer基线快3400倍以上），并通过自动化双阶段CI/CD质量保证框架和物理故障安全架构进行了验证。
英文摘要
High-frequency industrial cyber-physical systems (CPS) stream continuous sensor telemetry at sub-10-second intervals. Multi-horizon forecasting over hours is required for predictive maintenance and operational control. However, deploying machine learning models over extended horizons (H = 2,160 steps at dt = 5 s) encounters the Dual-Timescale Forecasting Dilemma: the trade-off between short-horizon kinetic momentum and long-horizon diurnal thermodynamic equilibrium. Autoregressive rollouts suffer from compounding error drift, while direct multi-output projectors exhibit variance explosion and noise extrapolation. In this work, we prove that autoregressive mutual information decays exponentially toward zero over extended lead times, and that the optimal minimum-variance estimator converges asymptotically to the periodic diurnal baseline (Lemma 1 and Theorem 1). Guided by these bounds, we propose the Dynamic Asymptotic Decomposition (DAD) framework. The architecture couples a continuous sigmoidal authority transition schedule \(α(h)\) with monotonic L2 regularization scaling \(λ(h)\), transferring prediction authority from kinetic autoregression to diurnal equilibrium while enforcing physical non-negativity. Evaluated across a 30-day industrial telemetry stream (N = 518,400 steps) over 16 walk-forward test windows, DAD achieves an overall Mean Absolute Error (MAE) of 0.1764, representing a 78.3% error reduction over Deep LSTM (0.8143), 60.2% over PatchTST (0.4431), and 20.8% over DLinear (0.2227). Computational profiling demonstrates edge execution in 0.72 ms with 0.203 MFLOPs and 72k parameters on a single CPU core (over 3400x faster than Transformer baselines), validated by an automated dual-stage CI/CD quality assurance harness and physical fail-safe architecture.
Comments33 pages, 16 figures, 8 tables, 17 equations
