Abstract: Reasonable blasting system is conducive to the smooth condition of blast furnace, and can promote energy saving and carbon reduction of blast furnaces. At present, the mapping relationship between blasting system and fuel consumption parameters is not perfect. To solve this problem, we propose an optimization method of blast furnace blasting system based on K-means clustering, support vector regression(SVR) and non-dominated sorting genetic algorithm(NSGA-Ⅱ). Firstly, K-Means and fuzzy C-means algorithm are used to cluster blast blast parameters, and K-Means model with good clustering effect is selected to cluster the blast furnace condition. Then, combining K-Means clustering results and feature selection, we extracte the key parameters of the blasting system, and use SVR, radial basis neural network, random forest regression and extreme gradient lifting model to predict the fuel ratio, pulverized coal ratio and daily iron output of blast furnace. The SVR model with the most accurate prediction was selected as the prediction model. Finally, a multi-objective programming model is constructed based on SVR model, and NSGA-Ⅱ algorithm is used to find the non-inferior solution set with minimum fuel ratio and maximum pulverized coal ratio, and the results are analyzed. The results show that the method can improve the blasting system, reduce the fuel ratio, and promote the energy saving and carbon reduction of blast furnaces.