Abstract: Under the background of carbon peaking and carbon neutrality goals and clean and low-carbon transformation of coal, green hydrogen is an important secondary energy supporting the carbon reduction upgrading of coal chemical industry and the construction of clean coal technology system. Anion exchange membrane water electrolysis (AEMWE) has the advantages of low cost with non-noble metal catalysts and high efficiency with the high current density, which is a green hydrogen production technology with great engineering application potential. However, its internal multi-physics fields are strongly coupled, the interaction between structural and operating parameters is complex, and there is an inherent objective conflict between hydrogen production rate and product purity. The traditional computational fluid dynamics(CFD) point-by-point optimization has extremely high computational cost, which is difficult to support multi-parameter collaborative optimization. This paper proposes a rapid optimization framework integrating analysis of variance(ANOVA), Latin hypercube sampling(LHS), BP neural network surrogate model and multi-objective particle swarm optimization(MOPSO). Firstly, 5 key decision variables including electrode conductivity, electrode active specific surface area, electrolyte concentration, diffusion layer permeability and porosity are screened out from 7 candidate parameters via ANOVA. A BP surrogate model is constructed based on 360 groups of LHS samples and multi-physics simulation data. After verifying its prediction accuracy, the MOPSO algorithm is coupled to solve the Pareto optimal solution set, and the weighted ideal point method is adopted to select the compromise optimal scheme considering both objectives. The optimization results show that the determination coefficients(R2) of the surrogate model for current density and hydrogen mole fraction reach 0.9913 and 0.9389 respectively, which can efficiently replace high-fidelity CFD simulation for parameter optimization. A total of 16 uniformly distributed non-dominated solutions are obtained by MOPSO optimization. The final selected optimal operating and structural parameters are as follows: electrode conductivity of 9811S/m, active specific surface area of 5314m?1, electrolyte concentration of 3733mol/m3, diffusion layer permeability of 1.32×10?1?m2, and porosity of 0.50. Simulation verification results show that the average current density of the optimized electrolyzer reaches 1.78A/cm2, increased by 3.73% compared with the baseline condition; the hydrogen mole fraction reaches 0.267, increased by 5.95%. Meanwhile, the uniformity of current density distribution and the stability of gas-liquid two-phase flow are significantly improved, and their coefficients of variation(CV) are reduced by 19.7% and 10.4% respectively, which effectively inhibits local current hot spots and bubble accumulation. This study reveals the performance evolution law of AEM electrolyzers under multi-parameter coupling. The proposed optimization framework can realize rapid collaborative optimization of structural and operating parameters, provide theoretical basis and technical path for efficient and stable operation and intelligent parameter regulation of AEM electrolyzers, and has reference value for promoting the engineering application of green hydrogen in the field of clean coal utilization.