Physics-Aware Distillation of a Grid Foundation Model for Real-Time Edge Inference
作者: Vlad Alexandru Sănduleac, Shahariya Rasheed Rayamarakkarveettil Puthanpurayil, Noman Khan, Remus Teodorescu
分类: cs.LG
发布日期: 2026-10-15
💡 一句话要点
提出物理感知蒸馏方法以解决边缘推理中的模型体积问题
🎯 匹配领域: 支柱二：RL算法与架构 (RL & Architecture) 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 物理感知 模型蒸馏 边缘推理 电网模型 深度学习 实时推理 可再生能源 功率流分析
📋 核心要点
- 现有的网格基础模型体积庞大，无法满足边缘设备实时推理的需求，尤其是在低惯性操作场景中。
- 本文提出了一种物理感知蒸馏方法，将大型GridFM模型压缩为适合边缘推理的小型模型，同时保持高精度。
- 实验结果显示，蒸馏后的学生模型在主动功率调度上错误率降低了48%，且在反应功率调度上表现优于教师模型。
📝 摘要（中文）
网格基础模型（GridFMs）是基于图的变换器，预训练于掩蔽功率流目标，能够比迭代求解器快几个数量级地评估功率流。然而，其模型规模主要针对数据中心推理，而低惯性操作日益需要在边缘进行推理。本文将一个20.1M参数、12层的GridFM教师模型蒸馏为一个0.62M参数、6层的学生模型，实现了参数减少32.5倍，并在消息传递深度上减半。我们提出的目标结合了掩蔽功率流损失、输出、物理残差、潜在状态和注意力转移。实验表明，蒸馏后学生模型在所有四个预测量上降低了错误率，主动功率调度的错误率降低了48%。
🔬 方法详解
问题定义：本文旨在解决网格基础模型在边缘推理中的体积过大问题，现有方法无法满足低延迟和小批量推理的需求。
核心思路：通过物理感知蒸馏，将大型教师模型转化为小型学生模型，利用多种损失函数结合以保持模型性能。
技术框架：整体架构包括教师模型的训练、蒸馏过程和学生模型的评估，主要模块包括掩蔽功率流损失、物理残差损失等。
关键创新：首次将物理残差项应用于类型化的电网图中，显著提升了模型在边缘设备上的推理效率。
关键设计：学生模型采用6层结构，参数量减少至0.62M，损失函数结合了掩蔽功率流损失与输出匹配，物理残差项在层聚合形式下表现出一定的负面影响。
🖼️ 关键图片
📊 实验亮点
实验结果显示，蒸馏后的学生模型在主动功率调度上错误率降低了48%，并且在反应功率调度上每个蒸馏后的学生模型的准确性均高于其教师模型，展示了显著的性能提升。
🎯 应用场景
该研究的潜在应用领域包括智能电网、可再生能源管理和边缘计算等。通过在边缘设备上实现实时推理，可以提高电力系统的响应速度和灵活性，推动智能电网的进一步发展。
📄 摘要（原文）
Grid foundation models (GridFMs) are graph transformers pretrained on a masked power-flow objective, and evaluate power flow orders of magnitude faster than an iterative solver. Their size, however, targets data-centre inference, whereas low-inertia operation increasingly asks for inference at the edge, next to the measurement, at batch size one and within a single synchrophasor (PMU) reporting frame. We distil a 20.1M-parameter, 12-layer GridFM teacher into a 0.62M-parameter, 6-layer student, a $32.5\times$ parameter reduction at half the message-passing depth, using an objective that combines the masked power-flow loss with output, physics-residual, latent-state and attention transfer. The physics-residual term, which matches the teacher's power-residual profile, has not previously been applied to typed power-grid graphs. We evaluate on a renewable variant of the IEEE 118-bus benchmark grid in which 26 of the conventional (synchronous) generators are replaced by inverter-based renewables with battery flexibility and standardised reactive-power limits (IEEE 1547), giving $97\,529$ physically valid (AC-OPF-feasible) operating scenarios, and we ablate all eight combinations of the transfer terms, reporting accuracy against ground truth and fidelity to the teacher separately. Distillation lowers student error on all four predicted quantities, by up to $48\%$ on active-power dispatch, and every distilled student is more accurate than its teacher on reactive-power dispatch. The gain is largely carried by the output matching term. The physics-residual term, in the layer-aggregated form used here, is consistently harmful, which we attribute to the spatial averaging it applies to a bus-level residual field. The resulting student is $32.5\times$ smaller in parameter count and has half the message-passing depth, making it a candidate for subsequent edge-device benchmarking.