Fuel Cell Anode Humidity Prediction Without Hardware Sensors

The patent creates a virtual copy of the humidity sensor function through software modeling rather than using physical hardware sensors. The mixed model (combining physical models and AI models) replicates humidity measurement capabilities by processing data from existing sensors (temperature, pressure, current) to predict humidity levels, thereby avoiding the reliability issues of physical humidity sensors while maintaining measurement functionality.

The patent replaces the mechanical/physical humidity sensing system with a computational/software-based system. Instead of using physical sensors that degrade over time, the system uses algorithms (physical models + AI models) running on existing vehicle controllers to calculate and predict humidity, substituting a reliable software solution for an unreliable hardware component.

The patent merges two different modeling approaches into a hybrid system: physical models (which provide interpretability and work well under known conditions) and AI models (which adapt to varying operating conditions). This combination allows the system to maintain the advantages of both approaches - the physical understanding from traditional models and the adaptive accuracy of machine learning - to achieve high prediction accuracy across diverse vehicle operating conditions.

The patent creates a composite modeling approach by integrating heterogeneous modeling methods (physical equations and AI algorithms) into a unified prediction system. This composite model structure allows different types of information and calculation methods to work together, leveraging the strengths of each approach to achieve superior overall performance compared to using either model type alone.

The patent creates a virtual humidity measurement capability through software modeling, copying the function of a physical humidity sensor without adding the corresponding hardware. The mixed model processes data from existing sensors to generate virtual humidity measurements, eliminating the need for additional physical sensors and their associated complexity and cost.

The system uses its existing sensors (temperature, pressure, current) and computational resources to provide humidity measurement functionality on its own, without requiring external or additional hardware. The controller leverages data already being collected for other purposes and applies modeling algorithms to derive humidity information, making the system self-sufficient.