基于差分进化-灰狼优化算法的支持向量机连铸漏钢预报系统研究  

Support Vector Machine Breakout Prediction System Based on DE-GWO Hybrid Algorithm Optimization

作　　者：吴迪 张本国[1] 生万宝 张瑞忠 Wu Di;Zhang Benguo;Sheng Wanbao;Zhang Ruizhong(School of Mechanical Engineering,Yancheng Institute of Technology;Technology Research Institute of Steel Research Institute of Hebei Iron and Steel Group)

机构地区：[1]盐城工学院机械工程学院 [2]河钢集团钢研总院工艺研究所

出　　处：《特种铸造及有色合金》2023年第2期174-177,共4页Special Casting & Nonferrous Alloys

基　　金：江苏省基础研究计划资助项目(BK20150429)。

摘　　要：针对小样本训练数据情况下BP神经网络模型难以获得较高预报准确率的问题,提出了一种基于差分进化-灰狼优化(DE-GWO)混合算法优化的支持向量机(SVM)模型,并将其应用到连铸漏钢预报系统领域。利用差分进化(DE)算法的交叉变异操作对灰狼优化(GWO)算法进行进化,得到DE-GWO混合算法,在解决了GWO易陷入局部最优问题的同时,提高了算法的寻优速度以及模型的准确性。结合某钢厂连铸生产数据,对DE-GWO-SVM漏钢预报模型进行测试。结果表明,该算法下的连铸漏钢预报系统的准确率为99.5%,报出率达到100%。Aiming at the difficulty to obtain high forecast accuracy for BP neural network models with small sample training data, a support vector machine(SVM) model based on a hybrid Differential Evolution-Grey Wolf Optimizer(DE-GWO) algorithm optimization was established and applied to the field of continuous cast steel breakout prediction system. The Grey Wolf Optimizer(GWO) algorithm was evolved by cross-variance operation of the Differential Evolution(DE) algorithm to obtain the DE-GWO hybrid algorithm, which solves the problem that GWO is prone to fall into local optima and improve the algorithm’s optimization finding speed as well as the accuracy of the model. The DE-GWO-SVM steel breakout prediction model was tested by combined with historical data from a steel mill continuous casting site, and the test results indicate that the accuracy of continuous cast steel breakout prediction system based on the algorithm is 99.5%, and the reporting rate reaches 100%.

关 键 词：连铸 漏钢预报 支持向量机 灰狼优化算法 差分进化 

分 类 号：TF777.7[冶金工程—钢铁冶金] TG249.7[金属学及工艺—铸造]