摘要
转炉炼钢的钢水终点成分及温度应由供氧进行控制,由于炉内存在物理化学反应复杂难以预测和冶炼过程中所得数据非线性的问题,传统方法不能准确预测。为提高预测精度,提出采用极限学习机来建立耗氧量预测模型的方法。针对极限学习机的权值和阈值随机确定所导致的网络结构稳定性差的问题,采用遗传算法进行优化,并对神经网络隐含层数量和隐含层激励函数的不同选择对仿真结果的影响做出了具体的分析。仿真结果表明优化算法模型预测精度有明显提高,验证了上述优化方法的有效性。
In order to improve the accuracy of prediction, an extreme learning machine was put forward to build up the model of oxygen consumption prediction. In view of poor stability of ELM network structure caused by random determination of the weight and threshold value, a genetic algorithm was used to solve the problem. The influences of different choices of the number of hidden layers and the activation function of the hidden layers of the neural network on the simulation results were analyzed exactly. The final simulation results show that the prediction accuracy of GA-ELM is significantly higher than that of ELM. The effectiveness of the optimization method is verified.
出处
《计算机仿真》
北大核心
2017年第1期380-383,422,共5页
Computer Simulation