改进的多种群算法优化隐马尔可夫模型预测篦压趋势
Prediction of the tendency of grate pressure based on hidden Markov model which is optimized by the improved multiple population genetic algorithm
摘要点击 48  全文点击 70  投稿时间:2018-07-03  修订日期:2019-03-05
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2018.80490
  2019,36(8):1217-1226
中文关键词  主成分分析  遗传算法  隐马尔可夫模型  篦冷机  预测  篦下压力  改进的多种群算法
英文关键词  principal component analysis  genetic algorithm  hidden Markov model  grate  prediction  the grate pressure  improved multiple population genetic algorithm
基金项目  国家自然科学基金(51641609);河北省自然基金项目(F2016203354)
学科分类代码  
作者单位E-mail
刘兆伦 燕山大学信息科学与工程学院 liuzhaolun@ysu.edu.cn 
张春兰 燕山大学信息科学与工程学院  
郭长江 燕山大学里仁学院  
王海羽 燕山大学信息科学与工程学院  
武尤 燕山大学信息科学与工程学院  
刘彬 燕山大学信息科学与工程学院  
中文摘要
      以篦冷机关键参数篦下压力为研究对象,提出一种篦压变化趋势预测模型。利用主成分分析对数据降维,以主元序列作为观测序列,构建改进的多种群优化隐马尔可夫模型参数,种群内利用轮盘赌算子选择个体,设计双区与均匀行交叉结合的自适应交叉算子避免局部收敛,进行动态变异率的多项式变异操作提高收敛速度,种群间提出混合师生交流机制的自适应移民算子保证多种群协同进化。仿真表明本文算法可收敛到全局最优,能提高收敛精度和速度,利用该算法建立的模型跟踪性能好,预测精度高,能满足对篦压趋势预测的要求。
英文摘要
      A model for predicting the variation tendency of grate pressure is proposed in this paper, taking the grate pressure as the research object which is the key parameters of grate cooler. Principal component analysis is used to reduce the dimension of data. The principal component feature sequence is used as the observation sequence. An algorithm combined with the improved multiple population genetic algorithm and hidden Markov model is constructed. Individuals are selected by roulette selection operator to avoid local convergence, the adaptive crossover operator is designed, which is combined with double zone crossover and uniform line crossover. The polynomial mutation operation of dynamic mutation rate is applied to improve the convergence speed within population. An immigration operator is presented which is mixed by communication mechanism between teachers and students to ensure the co-evolution of multiple populations. The research results show that the improved algorithm can converge to the global optimum and improve convergence accuracy and speed. The model established by this algorithm exhibits good tracking performance and high prediction accuracy, which is suitable for predicting the variation tendency of grate pressure.