The molecular sieve catalyst of Ce-Cu-SSZ-13 has a good selective catalytic reduction (SCR) denitration performance at the temperature consistent with that of exhaust gas of industrial furnaces. It is proposed to be coated on the surface of honeycomb ceramics containing expansion and contraction channels to manufacture a high-temperature composite SCR regenerator, so that the integration of flue gas denitration and waste heat recovery equipment can be realized. To optimize the comprehensive performance of the new SCR regenerator, numerical simulation of SCR flue gas denitrification, coupled with regenerative heat transfer, is conducted with Fluent software, where regenerator length (L), expansion/contraction angle (θ), and flue gas flow rate (V) are adopted as design variables, and energy recovery ratio (ERR), denitrification efficiency (η), and flue gas pressure loss (?P) as objective functions. Based on the simulation results, an objective function regression model is established by using the response surface methodology (RSM). On this basis, the advanced non-dominated genetic algorithm (NSGA-II) is used to obtain the Pareto optimal solution set, from which the EWM (entropy weight method)-TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) method is employed to calculate the relative closeness between Pareto optimal solutions and negative ideal solutions, then the optimal solution is determined. Analysis of variance verifies the significance of the regression model, while response surface analysis reveals the synergistic effects of different design variables on the objective function. For the current new composite SCR regenerator, the optimal objective functions take the values of η=99.38%, ?P=52.57 Pa, and ERR=68.13%, while the corresponding design variables are V=4.03 m/s, θ=12.19°, and L=717.57 mm. At this time, the average temperature of flue gas outlet is about 150℃, and the mass concentration of NO is 4.3 mg/m3. Besides, after limiting the flue gas velocity to 5 and 6 m/s, it is found that a larger V results in a smaller optimal θ to prevent the pressure loss increase significantly.