Abstract: As the proton exchange membrane fuel cell（PEMFC） generation system is nonlinear and strongly coupled,in this paper, a robust nonlinear observer（RNO） with multi-boundary layer was proposed to estimate the six state variables in the PEMFC air-feed system. Based on the sixth-order nonlinear dynamic model of the PEMFC system,the nonlinear sliding mode algorithms with multiple boundary layers were used to generate three nonlinear error correction terms that were used to adjust the output estimation errors of the system and steer them to zero in finite time in the condition of parameter uncertainties and noises, moreover, the stability of RNO was also proved. Meanwhile, the accuracy, convergence property and robustness performance of the RNO with multi-boundary layer for the six states were analyzed and compared with the traditional sliding mode observer（SMO）.Finally, experiment results show that the proposed RNO is not affected by the parameter variations, measurement noises and external disturbance, it performs much better than the traditional SMO, the correctness and effectiveness of the proposed method are both verified.