Abstract <p>Aiming at the problem that the traditional sintering batching model lacks cost iterative optimization and is not associated with sinter ore quality indicators, the study proposes an intelligent sintering batching optimization model incorporating Least Squares Support Vector Machine Regression (LSSVR) and Multi-Objective Gray Wolf Optimization (MOGWO) algorithm. Based on the sintering cup experimental data of a company in the past two years, a dual-index prediction model of sinter yield rate and drum index was constructed, and key parameters such as combustion consumption, assimilation temperature, and raw ore composition were screened out as model input variables by Spearman correlation analysis, feature ranking and similarity computation, and the accurate prediction of sinter ore quality indicators was realized by combining with the LSSVR algorithm, of which the model fitting of yield rate the R2 of the yield model reached 0.92, and the R2 of the drum index model reached 0.96. A multi-objective optimization model is established to minimize the cost of sintering batching, and the yield rate and drum index are qualified, taking into account the constraints of material composition, process parameters and production indexes, and the MOGWO algorithm is adopted to realize the optimal solution of batching scheme; when the model output scheme quality indicators are not satisfied, problem analysis and iterative optimization are performed with the help of the constructed large language reasoning model. The actual case validation shows that the optimization scheme obtained by this sintering batching optimization model can ensure that the cost of ore batching for ton of iron is reduced by RMB 4.77 under the condition that the yield rate and drum index reach the standard, which provides an effective decision-making reference for the fine management of the sintering batching process and the reduction of cost and increase of efficiency.</p> Graphical Abstract <p></p>