{"id":16935,"date":"2022-08-18T07:10:00","date_gmt":"2022-08-18T07:10:00","guid":{"rendered":"https:\/\/zbt.de\/?post_type=project&p=16935"},"modified":"2026-08-21T07:10:54","modified_gmt":"2026-08-21T07:10:54","slug":"past-el-predictive-ast-protocols-for-alkaline-electrolysis","status":"publish","type":"project","link":"https:\/\/zbt.de\/en\/projects\/past-el-predictive-ast-protocols-for-alkaline-electrolysis\/","title":{"rendered":"pAST-EL - Predictive AST Protocols for Alkaline Electrolysis"},"content":{"rendered":"
\n
\n \"\"\n
\n <\/div>\n <\/figure>\n
\n

pAST-EL – Predictive AST Protocols for Alkaline Electrolysis<\/h1>\n

Data-Driven Models for Improved Service Life Prediction<\/h4>\n

Reliably predicting the service life of electrolysers under real-world, dynamic operating conditions is a key challenge for the economical production of green hydrogen. While traditional accelerated stress tests (ASTs) provide degradation data quickly, they are of limited use for predicting long-term degradation under varying real-world operating conditions. <\/p>\n<\/div>\n <\/div>\n<\/section>\n

\n

Using Machine Learning to Develop Reliable Service-Life Models<\/h2>\n

Reliably predicting the service life of electrolysers under real-world, dynamic operating conditions is a key challenge for the economically viable production of green hydrogen. While traditional accelerated stress tests (ASTs) provide degradation data quickly, they are only of limited use for predicting long-term degradation under varying real-world operating conditions. <\/p>\n

As part of the pAST-EL<\/strong> project, ZBT and DLR are therefore developing a new type of predictive accelerated stress test (pAST)<\/strong>. The test protocols are specifically designed to generate a particularly meaningful data set for service life prediction models within the shortest possible test times. <\/p>\n

The approach is being experimentally validated using alkaline electrolysis (AEL)<\/strong>. To this end, pAST-based cell tests and long-term measurements are being conducted under realistic dynamic operating conditions. Machine-learning-based lifespan models<\/strong> are developed and validated using the measurement data. Building on this, critical operating conditions will be identified and lifespan-optimised operating strategies<\/strong> derived. Subsequently, their impact on degradation, hydrogen production costs, and greenhouse gas intensity will be investigated. <\/p>\n

The methodology developed is intended to be applicable across various technologies. Protocol development, lifetime prediction, and corresponding software tools are therefore provided as open, user-customisable methods. <\/p>\n<\/div>\n<\/section>\n

\n
\n

Project information<\/h2>\n