A novel optimization method via Box-Behnken Design integrated with back propagation neural network - genetic algorithm on hydrogen purification

Abstract

High purity hydrogen is a necessary need for fuel cell. Pressure swing adsorption (PSA) technology is one of the effective methods for hydrogen purification. The layered bed PSA model is built and validated. To simplify the calculation of PSA purification performance, the quadratic regression equations are obtained by Box-Behnken design (BBD) method.With adsorption time, pressure equalization time and feed flow rate from PSA process as independent optimization parameters of the BBD method, the hydrogen purity and productivity as two responses. The genetic algorithm (GA) is introduced to the back propagation neural network (BPNN) to solve the optimization problem of the PSA process. In order to explore the performance of optimization algorithms, a novel optimization method is proposed in this work. With the BBD method is integrated with BPNN-GA model to optimize the structure (BBD-BPNN-GA). The results showed that the BBD-BPNN-GA model have a better performance with the MSE of 0.0005, while the mean square error (MSE) of BPNN-GA model is 0.0035. And the correlation coefficient of R-values are much closer to 1of the BBD-BPNN-GA model, which is illustrated that the BBD-BPNN-GA model can be effectively applied to the prediction and optimization of PSA process.