Industrial energy conservation and CO2 emission reduction (ECCER) management is a multi-objective optimization problem with multiple uncertainty factors. However, most studies have used deterministic optimization approaches, and neglected the uncertainty factors that affect the effectiveness of the management strategies. This study adopts a multi-objective optimization model under uncertainty to solve the ECCER management problem in China's iron and steel industry. Three objectives: minimum energy intensity, maximum CO2 emission reduction, and minimum cost, are optimized simultaneously. This study simulates the perturbation of the uncertainty parameters within their fluctuation ranges via Latin Hypercube Sampling, adopts the mean objective function value mechanism to calculate the objective value, and obtains the optimized results using the second generation of the Non-dominated Sorting Genetic Algorithm (NSGA-II). Lastly, this study sets three types of preferences to generate final decision strategies via a Vague set-based approach. Results show: (1) The algorithm is reliable as per the verification of Hypervolume indicator and Spacing Metric; (2) The average values of energy intensity and CO2 emission reduction amount in optimal solutions are 524.00 kgce and 125.03 kg per ton steel respectively, which are 6.3% and 7.6% lower than the deterministic optimal ones; (3) The decision strategies encourage the wider application of large-sized process equipment, identify 8-9 advanced technology and eight reutilization approaches as key measures, but find the use of renewable energy will still be in low level. This study aims to solve the industrial ECCER optimization problem under uncertainty, and put forward policy suggestions in sustainable manufacturing in this industry.