MACHINE LEARNING IN PROTON EXCHANGE MEMBRANE WATER ELECTROLYSIS – PART I - A KNOWLEDGE-INTEGRATED FRAMEWORK

Aug 4, 2026

Upload by Violet

A knowledge-integrated machine learning framework is proposed to advance proton exchange membrane water electrolysis (PEMWE) development for green hydrogen production. The study systematically combines data-driven models with domain-specific insights to address limitations in optimizing performance. It analyzes uncertainties from data acquisition conditions, model mechanisms, and domain expertise, and uses complementary information to guide decomposition of knowledge. The resulting hierarchical “Ladder of Knowledge-integrated Machine Learning” supports knowledge extraction for data augmentation, modeling, and knowledge discovery. Three case studies for cell degradation analysis evaluate interpolation, extrapolation, and information representation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages21

Reading Time53 min

AccessFree to download

Abstract

Knowledge-integrated Machine Learning in PEMWEs

1 Introduction

Ladder of Knowledge-integrated Machine Learning: three-level integration

PEMWEs challenges and motivation for knowledge integration

arXiv :2404 .03660v1 [ cs .LG] 24 Jan 2024 MACHINE LEARNING IN PROTON EXCHANGE MEMBRANE WATER ELECTROLYSIS – PART I: A KNOWLEDGE-INTEGRATED FRAMEWORK Xia Chen Sustainable Building Systems Leibniz University Hannover Hannover, Germany [xia.chen@iek.uni-hannover.de](xia.chen@iek.uni-hannover.de) Alexander Rex Institute of Electric Power Systems Leibniz University Hannover Hannover, Germany [rex@ifes.uni-hannover.de](rex@ifes.uni-hannover.de) Janis Woelke Institute of Electric Power Systems Leibniz University Hannover Hannover, Germany [woelke@ifes.uni-hannover.de](woelke@ifes.uni-hannover.de) Christoph Eckert Institute of Electric Power Systems Leibniz University Hannover Hannover, Germany [eckert@ifes.uni-hannover.de](eckert@ifes.uni-hannover.de) Boris Bensmann Institute of Electric Power Systems Leibniz University Hannover Hannover, Germany [boris.bensmann@ifes.uni-hannover.de](boris.bensmann@ifes.uni-hannover.de) Richard Hanke-Rauschenbach Institute of Electric Power Systems Leibniz University Hannover Hannover, Germany [rhr@ifes.uni-hannover.de](rhr@ifes.uni-hannover.de) Philipp Geyer Sustainable Building Systems Leibniz University Hannover Hannover, Germany [philipp.geyer@iek.uni-hannover.de](philipp.geyer@iek.uni-hannover.de) ABSTRACT In this study, we propose to adopt a novel framework, Knowledge-integrated Machine Learning, for advancing Proton Exchange Membrane Water Electrolysis (PEMWE) development. Given the significance of PEMWE in green hydrogen production and the inherent challenges in optimizing its performance, our framework aims to meld data-driven models with domain-specific insights systematically to address the domain challenges. We first identify the uncertainties originating from data acquisition conditions, data-driven model mechanisms, and domain expertise, highlighting their complementary characteristics in carrying information from different perspectives. Building upon this foundation, we showcase how to adeptly decompose knowledge and extract unique information to contribute to the data augmentation, modeling process, and knowledge discovery. We demonstrate a hierarchical three-level framework, termed the "Ladder of Knowledge-integrated Machine Learning,"in the PEMWE context, applying it to three case studies within a context of cell degradation analysis to affirm its efficacy in interpolation, extrapolation, and information representation. This research lays the groundwork for more knowledge-informed enhancements in ML applications in engineering. Keywords Proton Exchange Membrane Water Electrolysis · Degradation Analysis · Machine Learning · Knowledge Engineering Knowledge-integrated Machine Learning in PEMWEs 1 Introduction The integration of Machine Learning (ML) with domain-specific knowledge is a pivotal advancement in predictive modeling [1, 2] . This combination has brought a new level of precision and insight to fields within engineering and environmental sciences [3, 4] . While the synergy has notably improved accuracy and decision-making processes [5, 6], the challenge of seamlessly blending domain knowledge with ML algorithms continues to evolve. To bridge this gap, the Ladder of Knowledge-integrated Machine Learning has been introduced [7] . This framework aims to optimize the utilization of domain-specific insights, offering a comprehensive approach to integrating prior knowledge information into ML applications. Inspired by the long debate between holistic and reductionist approaches in ML [8], the framework aims firstly to synergize multidisciplinary domain knowledge with data-driven processes in two principal dimensions: firstly, by identifying and understanding the complementary nature of uncertainties in data, knowledge-based methodologies, and data-driven methods; secondly, by exploring knowledge decomposition from various perspectives and aligning these insights with our paradigm. Finally, building upon the previous two foundati

FAQs

What problem does the Knowledge-integrated Machine Learning framework target in PEMWE development?

It targets the difficulty of optimizing PEMWE performance by systematically combining data-driven models with domain-specific knowledge to overcome challenges in modeling and limited interpretability.

Where do the uncertainties come from in the proposed framework?

Uncertainties originate from data acquisition conditions, data-driven model mechanisms, and domain expertise, and are treated as complementary sources of information.

How is the “Ladder of Knowledge-integrated Machine Learning” structured?

It is a hierarchical three-level framework that progressively integrates domain expertise to improve data augmentation/feature engineering, enable extrapolation/generalization, and provide logic for predictions.

Aug 4, 2026

Upload by Violet