Aiming at the complicated problems of green and high efficiency recovery for metal zinc in the process of resource utilization for zinc leaching residues by rotary kiln,it is crucial to optimize the zinc volatilization efficiency (VEZn) with minimizing the emission intensity of CO2 (EICO2) and kiln slag (EIks). A quantitative regulation approach of key parameters coupled with neural network simulations and multi-objective optimization algorithm is proposed in this study. The index prediction models of VEZn, EICO2 and EIks have been established through the back propagation neural network (BP-NNP) and its two improved algorithm, respectively. Through training by long time series of actual operating data, the results showed that the BP-NNP model improved by particle swarm optimization (PSO) could more accurately describe the complex nonlinear mapping relationship between wide fluctuating parameters and VEZn, EICO2 and EIks, and the mean absolute percentage errors (MAPE) could be reduced to 6.00 × 10−3, 8.52 × 10−3 and 6.08 × 10−2, respectively. Furthermore, the developed index prediction model was incorporated into cuckoo search (CS) and non-dominated sorting genetic algorithm-II (NSGA-Ⅱ) that performed Pareto dominance-based multi-objective optimization. In experiments conducted, the significant advantages of NSGA-Ⅱ were revealed that the corresponding Pareto-optimal solutions with VEZn distribution between 93.32% and 94.52% were obtained. Meanwhile, EICO2 and EIks could be reduced to 1.525 t/t and 0.317 t/t, respectively. And comprehensive benefits by Pareto-optimal solutions were analyzed to provide various representative countermeasures to assist decision-makers. New insights gained herein could enable clean, efficient and low-carbon process regulations for recycling of solid waste in zinc hydrometallurgy industry.