Data-driven multidimensional sensing and real-time conductivity control for PEMFC liquid cooling systems

Precise conductivity control in proton exchange membrane fuel cell liquid cooling systems remains a challenge due to monitoring delays, fixed flow rates, and passive control in conventional designs. Herein, this study proposes a data-driven framework for multidimensional sensing and real-time conductivity control. A structurally optimized multifunctional expansion tank, incorporating reinforcing ribs and a kiss-off structure, integrates conductivity sensing and deionization, achieving high mechanical strength and sealing reliability. Specifically, it attains a burst pressure of 395 kPa, with only a 0.8% deviation from simulation, and maintains sealing integrity with a pressure drop of less than 2 kPa after 20 thermo‑pressure coupled cycles. A multi-source data-driven strategy dynamically regulates coolant flow, enhanced by a data-driven predictive model. The experimental results show that the system rapidly reduces the coolant conductivity from 151.18 μS/cm at 80°C and stably maintains it below 5 μS/cm. Furthermore, it extends the purification module’s service life by more than 30% and enhances the system’s response stability. The work provides a new pathway for intelligent conductivity management, enhancing the durability and operational economy of PEMFC thermal management systems.