摘要
针对回采工作面瓦斯涌出这样复杂的动态变化系统,提出了改进的和声搜索算法(IHS)与正则极速学习机(RELM)相结合的预测方法。对和声搜索算法的基本原理进行了研究,通过采用动态变化的PAR和BW值,优化和声搜索算法的全局搜索能力;利用IHS选取RELM中的输入层权值(IW)和隐含层阈值(B),以均方根误差为目标函数,提高了算法的预测精度。仿真实验结果表明,通过与已有的BP神经网络和SVM预测模型作对比,该方法具有更好的预测效果。
According to the dynamic systems of gas emission in working face, this paper proposed a prediction method that improved harmony search algorithm(IHS) combined with regularized extreme learning machine(RELM) ,and researched on basic principle of the harmony search algo- rithm, optimized its global search ability by using dynamically values PAR and BW, and selected the input weights and hidden layer of the RELM by the algorithm. Using the root mean square error as objective function, this paper improved the prediction accuracy of algorithm. Simulation results showed that, by comparison with other existing prediction methods, this algorithm had better prediction results.
出处
《资源开发与市场》
CAS
CSSCI
2015年第3期262-265,300,共5页
Resource Development & Market
基金
国家自然科学基金项目(编号:70971059)
关键词
和声搜索算法
极速学习机
回采工作面
瓦斯涌出量
harmony search algorithm(HSA)
extreme learning machine(ELM)
working face
gas emission quantity