锅炉燃烧系统变分模态分解建模方法研究  

Research on Modeling Method of Variational Modal Decomposition of Boiler Combustion System

作　　者：张效源 刘利军 辛超 孙凯 鲍旭奇 李益国[2] ZHANG Xiaoyuan;LIU Lijun;XIN Chao;SUN Kai;BAO Xuqi;LI Yiguo(CHN Energy Feixian Power Generation Co.,Ltd.,Feixian 276001,China;School of Energy and Environment,Southeast University,Nanjing 210096,China)

机构地区：[1]国家能源费县发电有限公司,山东费县276001 [2]东南大学能源与环境学院,江苏南京210096

出　　处：《自动化仪表》2024年第6期6-10,共5页Process Automation Instrumentation

基　　金：国家自然科学基金资助项目(51476027)。

摘　　要：建立准确的燃烧系统模型是实现锅炉燃烧优化的基础。针对现有建模方法多步预测精度不足的问题,提出一种锅炉燃烧系统变分模态分解(VMD)建模方法。该方法将传统单个模型的预测方式转换为基于数据分解的混合模型预测方式。首先,采用VMD将输出信号分解为多个规律变化的平稳序列,以降低数据的复杂度。然后,针对每个模态信号,利用在线最小二乘支持向量机算法和时间序列预测算法分别建立预测模型。最后,通过合成得到最终的预测结果。仿真结果表明,该方法有效提升了燃烧模型的单步和多步预测精度,可以为实现锅炉闭环燃烧优化控制奠定良好基础。该方法有助于解决常规燃烧系统建模方法中只关注单步预测误差而忽视多步预测误差的问题。Establishing an accurate combustion system model is the basis for achieving boiler combustion optimization.Aiming at the problem of insufficient multi-step prediction accuracy of existing modeling methods,a modeling method of variational modal decomposition(VMD)of boiler combustion system is proposed.The method converts the traditional single model prediction method into a hybrid model prediction method based on data decomposition.Firstly,the VMD is used to decompose the output signal into multiple regularly varying smooth sequences to reduce the complexity of the data.Then,for each modal signal,the prediction model is built by using online least squares support vector machine algorithm and time series prediction algorithm respectively.Finally,the final prediction results are obtained by synthesis.Simulation results show that the method effectively improves the single-step and multi-step prediction accuracy of the combustion model,which can lay a good foundation for realizing the closed-loop combustion optimization control of the boiler.The method helps to solve the problem of conventional combustion system modeling methods that only focus on single-step prediction errors but neglect multi-step prediction errors.

关 键 词：锅炉燃烧系统 建模方法 锅炉效率 NOx生成量 变分模态分解 多步预测 

分 类 号：TH861[机械工程—仪器科学与技术]