Polymer electrolyte fuel cell (PEFC) system efficiency can be decreased by instabilities resulting from the accumulation of water in the cathode as well as by excessive air delivery parasitic loads used to prevent liquid water accumulation. In this work, we present a new instability detection diagnostic tailored for the nonlinear and chaotic dynamics of PEFC operation with multi-phase flow in the gas channels. The instability statistic, the Lyapunov exponent of the reduced-order voltage return map, A, is a measure of the exponential rate of divergence in the dynamic voltage signal measured from the fuel cell. A key advantage of this statistic for embedded control is that it is a self-referencing measure of the system stability for feedback and is not based on an a priori performance threshold. Our experiments demonstrate that the Lyapunov exponent statistic provides a warning typically 100 s in advance of significant power loss. Using this statistic as a control diagnostic, a new control scheme that detects PEFC instability in real time and mitigates it with pressure perturbations was applied experimentally to several fuel cell systems, including one that simulates stack operation. Our control scheme resulted in increased PEFC power, decreased cathode flooding leading to a lower parasitic load for air delivery, and stable PEFC performance. (C) 2014 Elsevier B.V. All rights reserved.