Fuel Cell Gas Supply Control Using Learning Term Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In fuel cell systems with variable gas supply devices, precise control of gas supply states downstream is challenging, especially during changes or abnormalities, leading to potential operational issues.
Innovation Solution
A fuel cell system incorporating a control section that performs PI control with a learning term, which integrates the I term only under specific conditions, to judge abnormalities in the variable gas supply device by monitoring deviations in pressure values and adjusting the gas supply accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a variable gas supply device is used to adjust gas pressure dynamically, then the adaptability of the fuel cell system improves, but the reliability of gas supply control deteriorates due to potential abnormalities and failures
Solution Approach 1:
The system performs preliminary action by integrating the I term of PI control only during predetermined learning allowable conditions (stable operation states) to build a reliable learning term database before actual abnormality detection occurs. This pre-conditioned integration ensures that the learning term accurately represents normal operation characteristics, enabling reliable abnormality judgment when conditions change.
Solution Approach 2:
The system implements feedback by continuously monitoring the learning term (integrated I term) and comparing it against abnormality thresholds to detect variable gas supply device failures. The control section adjusts the I term integration based on feedback from operation state monitoring, ensuring that learning occurs only when the system is in a stable, predictable state.
2Reliability
If abnormality judgment is performed continuously, then the reliability of detection improves, but false positives increase due to unstable operation states
Solution Approach 1:
The system performs preliminary action by establishing learning allowable conditions that define when the system is in a stable, representative operating state. Only when these conditions are met does the system integrate the I term for abnormality detection purposes, ensuring that the learning term is built from reliable data before any abnormality judgment is made.
Solution Approach 2:
The system applies local quality by treating different operation states differently with respect to I term integration. Stable operation states (satisfying learning allowable conditions) are designated as suitable for learning, while unstable or transitional states are excluded. This localized approach ensures that the learning term reflects only normal operation characteristics, improving abnormality detection precision.
Data Source
AI summary
A fuel cell system of the present invention includes a fuel cell, a supply channel which supplies, to the fuel cell, a fuel gas supplied from a fuel supply source, a variable gas supply device which adjusts a gas state on an upstream side of this supply channel to supply the gas to a downstream side, a control section which performs PI control of a gas supply command amount with respect to the variable gas supply device, and an abnormality judgment section to judge whether or not the variable gas supply device is abnormal. The controller uses, as a part of a correction term of the PI control, a learning term constituted by integrating an I term only in a case where an operation state of the fuel cell satisfies predetermined learning allowable conditions. The abnormality judgment section judges based on this learning term whether or not the variable gas supply device is abnormal.