摘要
针对高炉冶炼过程的复杂、多变以及非线性等因素,提出了一种基于粒子群算法(Particleswarmoptimization,PSO)和遗传算法(Geneticalgorithm,GA)相结合来优化极限学习机(Extremelearningmachine,ELM)的高炉铁水硅含量预测模型。PSO-GA-ELM预测模型,主要是在PSO算法进行适应度值计算、粒子的速度更新和位置更新时将GA算法中的选择、交叉和变异等操作融入其中,使其输出最优的连接权值和阈值代入到ELM模型中。通过对4种不同的预测模型进行实验验证,结果表明,优化后的PSO-GA-ELM模型在进行铁水硅含量预测时的预测精度、学习能力和泛化性能均高于其他三种预测模型。
Aiming at the complex,versatile and nonlinear factors of blast furnace smelting process,a combination ofParticle Swarm Optimization(PSO)and Genetic Algorithm(GA)is proposed to optimize Extreme learning machine(ELM)blast furnace molten iron silicon content prediction model. The PSO-GA-ELM prediction model mainly integrates the selection,crossover and mutation of the GA algorithm into the PSO algorithm for fitness value calculation,particle velocity update and position update,so that it outputs the optimal connection right values and thresholds are substituted into the ELM model. The experimental results of four different prediction models show that the optimized PSO-GA-ELM model has higher prediction accuracy,learning ability and generalization performance than other three prediction models.
作者
孙洁
崔婷婷
刘晓悦
徐彬
SUN Jie;CUI Ting-ting;LIU Xiao-yue;XU Bin(School of Electrical Engineering,North China University of Science and Technology,Hebei Tangshan 063210,China;Ironmaking Department,Shougang Jingtang United Iron&Steel Company Limited,Hebei Tangshan 063210,China)
出处
《机械设计与制造》
北大核心
2022年第3期228-232,237,共6页
Machinery Design & Manufacture
基金
河北省自然科学基金(E2019209492)。
关键词
粒子群算法
遗传算法
极限学习机
硅含量预测
预测精度
泛化性能
Particle Swarm Optimization
Genetic Algorithm
Extreme Learning Machine
Silicon Content Prediction
Prediction Accuracy
Generalization Performance
