Author: Pu He (Xi'an Jiaotong University) - The operating temperature of proton exchange membrane fuel cells (PEMFCs) critically impacts their efficiency, durability, and safety, necessitating robust thermal management strategies. This work explores tailored thermal control approaches for air-cooled PEMFCs, water-cooled PEMFCs, and combined heat-and-power (CHP) PEMFC systems.
For air-cooled PEMFCs, a single-input, single-output (SISO) configuration enables the use of PID controllers. However, traditional PID controllers face challenges due to fixed coefficients. To address this, we introduce an adaptive RBF-BP-PID controller that integrates Radial Basis Function and Back Propagation neural networks. This hybrid approach dynamically adjusts coefficients and provides real-time temperature identification, achieving improved control performance.
In water-cooled PEMFCs, a dual-input, dual-output (DIDO) structure introduces coupling effects between the inlet coolant temperature and the temperature gradient across the stack. Conventional feedforward decoupling based on system identification is first applied, followed by a novel multi-objective decoupling strategy using NSGA-III. This approach bypasses the need for explicit system identification and outperforms traditional feedforward methods.
For the more complex DIDO PEMFC-CHP system, we adopt a model-based Model Predictive Control (MPC) framework. A nonlinear state-space model of the thermal management system is linearized at equilibrium points to guide MPC design. Our results demonstrate that MPC provides significantly higher precision and stability than PID, with the Integral Absolute Error (IAE) reduced by nearly an order of magnitude.