首页 > AI前沿 > Toward Out-of-Distribution Generalization in Neural AC Power Flow

Toward Out-of-Distribution Generalization in Neural AC Power Flow

arXiv机器学习 2026-11-18 10:41 16 阅读 查看原文
用户需要将给定的英文纯文本新闻正文排版成语义化的HTML片段。 规则: 1. 只输出HTML片段本身,不要Markdown围栏,不要解释。 2. 每个自然段落用`

`包裹,长文用`

`/`

`划分小节(原文较短,可能不需要或者按需划分,但原文只有一段,所以直接用`

`,或者可以根据内容分句但保持原意,不过规则说“保留原文所有信息,不得增删改内容”,所以最好直接把整个正文作为一个或几个`

`,或者按逻辑划分。原文是一段,直接放一个`

`即可,或者为了体现排版,可以把发现列表部分用`

  • `,但原文是 "(1) ...; (2) ...; (3) ...; and (4) ..."。我可以把它转成列表,或者保持在段落中。规则3说“列表内容用`
    • `”,原文的(1)到(4)算是列表内容,可以转换,或者保持原样。为了严格遵守“不得增删改内容”,把句子中的(1)等改成`
    • `可能会改变文本结构,但语义上是列表。保险起见,保持段落结构,或者将发现部分拆为`
        `。不过“不得增删改内容”通常指文字本身。我可以把整段放在`

        `里,或者把前半部分放`

        `,发现部分放`

          `。 让我们仔细看原文: "Growing uncertainty in power system operations from renewables integration, extreme weather, and data center loads challenges the reliability of traditional power flow solvers at scale. Machine learning-based proxies are a potentially promising alternative, but the prohibitive cost of generating training data for large grids makes out-of-distribution generalization a key capability for deployment; currently, models trained on small grids lack that capability. In this work, we systematically evaluate how physics-informed losses, multi-grid training, and perturbation-based data augmentation improve the zero-shot transfer of three graph neural network-based power flow models from smaller systems to a larger, unseen 500-bus system. We find that (1) physics-informed losses improve transfer relative to purely supervised objectives; (2) multi-grid training on smaller systems matches the best single grid but does not close the gap to target-grid training; (3) perturbation-based augmentation on edge attributes improves transfer by up to 54.6%; and (4) random line/generator dropout does not consistently improve performance." 可以排版为:

          Growing uncertainty in power system operations from renewables integration, extreme weather, and data center loads challenges the reliability of traditional power flow solvers at scale. Machine learning-based proxies are a potentially promising alternative, but the prohibitive cost of generating training data for large grids makes out-of-distribution generalization a key capability for deployment; currently, models trained on small grids lack that capability. In this work, we systematically evaluate how physics-informed losses, multi-grid training, and perturbation-based data augmentation improve the zero-shot transfer of three graph neural network-based power flow models from smaller systems to a larger, unseen 500-bus system.

          We find that:

          • (1) physics-informed losses improve transfer relative to purely supervised objectives;
          • (2) multi-grid training on smaller systems matches the best single grid but does not close the gap to target-grid training;
          • (3) perturbation-based augmentation on edge attributes improves transfer by up to 54.6%; and
          • (4) random line/generator dropout does not consistently improve performance.
          等等,规则说“不得增删改内容”。把 "We find that (1)..." 拆开,算不算改内容?如果直接放在一个`

          `里可能更稳妥,避免被判定“增删改”。 或者:

          Growing uncertainty ... 500-bus system. We find that (1) physics-informed losses improve transfer relative to purely supervised objectives; (2) multi-grid training on smaller systems matches the best single grid but does not close the gap to target-grid training; (3) perturbation-based augmentation on edge attributes improves transfer by up to 54.6%; and (4) random line/generator dropout does not consistently improve performance.

          这样完全保留了原文的文本内容,没有增加或删减,只是加了`

          `和``。 或者使用列表,但保留原来的括号数字:

          Growing uncertainty in power system operations from renewables integration, extreme weather, and data center loads challenges the reliability of traditional power flow solvers at scale. Machine learning-based proxies are a potentially promising alternative, but the prohibitive cost of generating training data for large grids makes out-of-distribution generalization a key capability for deployment; currently, models trained on small grids lack that capability. In this work, we systematically evaluate how physics-informed losses, multi-grid training, and perturbation-based data augmentation improve the zero-shot transfer of three graph neural network-based power flow models from smaller systems to a larger, unseen 500-bus system. We find that

          • (1) physics-informed losses improve transfer relative to purely supervised objectives;
          • (2) multi-grid training on smaller systems matches the best single grid but does not close the gap to target-grid training;
          • (3) perturbation-based augmentation on edge attributes improves transfer by up to 54.6%; and
          • (4) random line/generator dropout does not consistently improve performance.
          这种排版更符合“列表内容用`
          • `”的指导,因为原文的(1)到(4)明显是列表形式。而且没有增删改文字内容(除了标点微调,比如把 "We find that" 后面的空格去掉,或者保留)。 原文: "We find that (1) ...; (2) ...; (3) ...; and (4) ..." 如果拆成列表:

            Growing uncertainty in power system operations from renewables integration, extreme weather, and data center loads challenges the reliability of traditional power flow solvers at scale. Machine learning-based proxies are a potentially promising alternative, but the prohibitive cost of generating training data for large grids makes out-of-distribution generalization a key capability for deployment; currently, models trained on small grids lack that capability. In this work, we systematically evaluate how physics-informed losses, multi-grid training, and perturbation-based data augmentation improve the zero-shot transfer of three graph neural network-based power flow models from smaller systems to a larger, unseen 500-bus system. We find that:

            • (1) physics-informed losses improve transfer relative to purely supervised objectives;
            • (2) multi-grid training on smaller systems matches the best single grid but does not close the gap to target-grid training;
            • (3) perturbation-based augmentation on edge attributes improves transfer by up to 54.6%; and
            • (4) random line/generator dropout does not consistently improve performance.
            这里加了一个冒号,算“增”吗?为了严格不增删改,可以直接:

            Growing uncertainty in power system operations from renewables integration, extreme weather, and data center loads challenges the reliability of traditional power flow solvers at scale. Machine learning-based proxies are a potentially promising alternative, but the prohibitive cost of generating training data for large grids makes out-of-distribution generalization a key capability for deployment; currently, models trained on small grids lack that capability. In this work, we systematically evaluate how physics-informed losses, multi-grid training, and perturbation-based data augmentation improve the zero-shot transfer of three graph neural network-based power flow models from smaller systems to a larger, unseen 500-bus system. We find that

            • (1) physics-informed losses improve transfer relative to purely supervised objectives;
            • (2) multi-grid training on smaller systems matches the best single grid but does not close the gap to target-grid training;
            • (3) perturbation-based augmentation on edge attributes improves transfer by up to 54.6%; and
            • (4) random line/generator dropout does not consistently improve performance