基于神经网络的锅炉燃烧控制系统建模设计  

Modeling and design of boiler combustion control system based on neural network

作　　者：孙德良[1] 王田[1] 周宇哲 柏雪 程崇俊[1] SUN Deliang;WANG Tian;ZHOU Yuzhe;BAI Xue;CHENG Chongjun(Xi’an Thermal Power Reserch Institute Co.,Ltd.,Xi’an 710054,China)

机构地区：[1]西安热工研究院有限公司,陕西西安710054

出　　处：《电子设计工程》2023年第10期20-24,共5页Electronic Design Engineering

基　　金：华能集团总部科技项目(HNKJ20-H80-01)。

摘　　要：为了提升锅炉控制系统的智能化水平,文中对锅炉内的燃烧过程进行了数学建模。基于锅炉的历史数据进行了关联分析,提取影响锅炉燃烧效率与危害物排放的主要因素,且采用神经网络模型实现了过程描述。在建模过程中,引入遗传算法对神经网络的隐含层节点数选择以及传递权重与直流偏置等参数进行了优化,以神经网络训练中的误差函数作为种群迭代的适应度,有效提升了神经网络训练过程中的收敛效率,避免在训练时陷入梯度下降的局部最优。基于某火电厂两组机组的真实运行数据进行了模型的训练与测试,为评估模型在遗传算法下的优化效果,将普通神经网络作为对照组。仿真结果表明,所设计模型的MAE和RMSE分别达到了4.63%及5.17%,在训练时间基本不变的情况下,MAE、RMSE相较于普通神经网络模型分别提升了5.60%和6.45%。In order to improve the intelligent level of the boiler control system,this paper makes a mathematical model of the combustion process in the boiler,carries out correlation analysis based on the historical data of the boiler,extracts the main factors affecting the boiler combustion efficiency and hazardous emission,and uses the neural network model to realize the process description.In the process of modeling,genetic algorithm is introduced to optimize the parameters such as the number of hidden layer nodes,transfer weight and DC bias of neural network.The error function in neural network training is used as the fitness of population iteration,which effectively improves the convergence efficiency in neural network training and avoids falling into the local optimization of gradient descent during training.Based on the real operation data of two groups of units in a thermal power plant,the model is trained and tested.In order to evaluate the optimization effect of the model under genetic algorithm,the ordinary neural network is used as the control group.The simulation results show that the MAE and RMSE models designed in this paper reach 4.63%and 5.17%respectively.When the training time is basically unchanged,MAE and RMSE is improved by 5.60%and 6.45%compared with the ordinary neural network model.

关 键 词：神经网络 遗传算法 系统建模 锅炉燃烧 智能控制 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]