Performance Monitoring of Proton Exchange Membrane Water Electrolyzer by Transformers-Based Machine Learning Model
トランスフォーマーベース機械学習モデルによるPEM水電解装置の性能監視 (AI 翻訳)
Bingqing Chen, I. Batalov, Qiu Chen, Weiqi Ji, Lei Cheng
🤖 gxceed AI 要約
日本語
本論文は、PEM水電解槽の運転中に機械学習を用いて仮想的な性能評価（分極曲線）を可能にするフレームワークを提案。エンコーダー・デコーダートランスフォーマーアーキテクチャにより、通常運転データから電気化学的特性を再構築し、状態健全性指標を提供する。グリーン水素の普及に向けたリアルタイムモニタリング技術として貢献する。
English
This paper proposes a machine learning framework using an encoder-decoder transformer to virtually characterize PEM electrolyzers during operation. The model reconstructs polarization curves from operational data, enabling real-time health monitoring without interrupting normal operation. It supports green hydrogen deployment by providing data-driven state-of-health indicators.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本は水素基本戦略でグリーン水素の普及を掲げており、PEM電解槽の効率的運用は重要な課題。本手法はコスト低減や信頼性向上に直結し、日本の水素サプライチェーン構築に資する。
In the global GX context
Globally, green hydrogen is critical for decarbonizing hard-to-abate sectors. This ML-driven monitoring approach reduces operational costs and improves system reliability, aligning with the goals of the global hydrogen economy and supporting large-scale electrolyzer deployment.
👥 読者別の含意
🔬研究者:Novel application of transformer architecture to electrochemical system monitoring, offering a data-driven method for virtual characterization.
🏢実務担当者:Enables real-time health monitoring of PEM electrolyzers without downtime, reducing maintenance costs and improving operational efficiency.
🏛政策担当者:Supports the scalability of green hydrogen production by lowering operational barriers, relevant for hydrogen strategy and subsidy design.
📄 Abstract（原文）
Hydrogen is increasingly recognized as a critical energy carrier of global decarbonization efforts, in sectors where direct electrification is challenging, such as long-haul transportation and high-temperature industrial processes. Among the technological route to produce low-carbon hydrogen, hydrogen production via water electrolysis powered by renewable electricity—often referred to as "green hydrogen"—is emerging as a key technology for sustainable hydrogen generation. Proton Exchange membrane (PEM) water electrolyzer is of particular interest as it offers high power density, excellent flexibility and minimal space requirements. Real-time PEM electrolyzer system health monitoring is essential for the deployment. In lab-scale device, performance degradation is typically assessed through well-controlled, pre-defined electrochemical testing protocols by periodic pauses of normal operation, such as polarization curve measurement with electrochemical impedance spectroscopy (EIS). However, full-scale stack deployments (in MW or GW) often operate under dynamic conditions due to their coupling with intermittent renewable electricity, where such controlled pauses are not realistic. This limits system operators’ ability to access the system’s state-of-health (SoH) in real-time operation. To bridge this gap, machine learning (ML) models offer a promising solution: by learning from operational data, the model enables virtual testing and performance monitoring without interrupting operation. In this work, we propose a ML framework with an encoder-decoder transformer architecture to enable virtual characterization of PEM electrolyzers during operation. Our model is trained to reconstruct electrochemical characterization tests, i.e., polarization curves, conditioned on operational data collected during normal operation. Additionally, the encoder–decoder also learns latent representations that capture meaningful insights relevant to the system internal state, which may serve as a data-driven SoH indicator.
🔗 Provenance — このレコードを発見したソース
- semanticscholar https://doi.org/10.48550/arxiv.2605.19107first seen 2026-07-20 05:25:01
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。