基于极限学习机的采煤机功率预测算法研究

为减少对领域专家的过分依赖，实现企业专家经验知识的继承，面向采煤机方案设计中总体技术参数的确定过程，结合采煤机条件属性与决策属性间的映射关系，提出了基于极限学习机的采煤机功率预测模型。采用遗传算法选定最佳隐层神经元个数，利用递进方式比选确定激励函数，随机产生输入权值及隐元偏置，由此计算隐层节点输出矩阵、隐层与输出层连接权重，进而完成建模与优化。模型可根据用户输入的不同原始设计条件输出采煤机功率的预测值。选用某煤机企业的实例数据进行算例分析，将其与基于支持向量机回归预测模型进行对比，实验结果表明，ELM模型可实现600 ms内完成单次功率预测，预测值与真实值平均相对误差在2.5%以内。其预测精度优于SVM模型，且在学习速度方面优势明显，推理效率显著提高。

Research on the algorithm of shearer power prediction based on extreme learning machine

In order to not excessively rely on domain experts and the inherit experimental knowledge of the experts,this paper presents a prediction model of shearer power based on Extreme Learning Machine (ELM),and combined with the mapping relationship between condition attributes and decision attributes for the overall design technical parameters determination in the design process.The model is built and optimized by identifying the optimum number of neurons on hidden layer with genetic algorithm,determining the excitation function with progressive comparison,randomly generating input weights and hidden elements bias and calculating hidden layer nodes output matrix,hidden layer and output layer connection weights.The model could output the predictive values of shearer power according to different original conditions input by users.Real design data were adopted to do algorithm analysis and contrast experiment with the reasoning model based on Support Vector Machine (SVM).The results show that ELM model can be used to complete a single power prediction in 600 ms.The average relative error of predicted values and real value is within 2.5%.The prediction accuracy of the proposed model is better than that of SVM model and it is of an apparent advantage over SVM model in learning speed.The reasoning efficiency has been improved significantly.