Optimization For Hydrogen Purification Performance Of Multi-Layer Bed Pressure Swing Adsorption Based On Artificial Neural Network

Hydrogen energy is a renewable energy that has invested the most resources in research and development to achieve the goal of "carbon neutrality" in the world.As one of the typical application scenarios of hydrogen energy,hydrogen fuel vehicles require high purity of hydrogen used as fuel.Hydrogen purified from industrial gas mixtures by pressure swing adsorption(PSA: Pressure Swing Adsorption)is a key technology in hydrogen energy applications.In this work,a four-layer adsorption bed model was established.The performance of the adsorbent was studied,and the structure of the adsorption bed was designed.First,the mathematical model of heat and mass transfer required to establish the PSA model is described in detail,and the principle of each formula and the meaning of its parameters are clarified.The corresponding adsorption isotherm models of activated carbon,zeolite,Cu-BTC and UTSA-16 were studied,and the isotherm models were verified with reference to references.The breakthrough curve was established in the simulation software and compared with the data in the breakthrough experiment to verify the correctness of the PSA cycle model.Secondly,a single layer adsorption bed model was established for the above four adsorbents on Aspen Adsorption,and the adsorption performance of the four adsorbents was compared under the same cycle conditions.It is concluded that UTSA-16 has the highest hydrogen purity and zeolite has the highest hydrogen recovery rate.The effect of adsorbent characteristics on adsorbent performance was also studied,and it was found that the saturated adsorption capacity had the greatest impact on the purification performance of adsorbent.The study of adsorbent characteristics also provides a reference for the selection of PSA adsorbents.Thirdly,a four-layer adsorption bed PSA cycle model composed of the above four adsorbents was established.The circulation process was designed,and the effect of adsorbent placement order on purification performance was studied.A model with activated carbon as the first layer and zeolite as the last layer was determined to have the best hydrogen purification effect.A comparison was made with the three-bed model without UTSA-16 under the same cycling conditions,and the results demonstrated that the four-bed model had better purification performance under various cycling conditions.The parameters of several cycle parameters were studied to obtain the influence of the cycle parameters on hydrogen purification effect.Finally,this work uses the simulation data obtained by the four-layer bed to train the ANN and designs the parameters and structure of the ANN model.The tuned ANN can accurately predict the PSA cycle outcome.Then,the multi-objective optimization algorithm is used to optimize the two optimization objectives of hydrogen purity and recovery at the same time,and a set of optimal solutions that maximize these two values are obtained,which can make the purity of hydrogen achieve 99.99% and the recovery of hydrogen more than 70%...