Experimental evaluation of linear model based control strategies for PEMFCs

In this work, we investigate linear model based multivariable control schemes for proton exchange membrane fuel cells (PEMFCs). Much of the literature relies on a mechanistic model to design model predictive controllers; however, this can be a difficult and time-consuming exercise for a PEMFC. An effective approach for developing models for control purposes is to use time series analysis and develop control oriented state space models directly from input-output data. In the present work, we develop an innovation form of state space model from input-output perturbation data obtained from a PEMFC. We then demonstrate the development of infinite horizon unconstrained linear model predictive controllers (LMPC) using these models, and compare their performance to IMC based PI control. We conduct servo and regulatory control studies on an experimental single cell PEMFC system, and demonstrate that the proposed control schemes regulate the power obtained from the fuel cell as desired even in the presence of disturbances.