吴林峰,孟瑜炜,俞荣栋,张震伟,徐仙华,王豆,郭鼎
WU Linfeng,MENG Yuwei,YU Rongdong,ZHANG Zhengwei,XU Xianhua,WANG Dou,GUO Ding

• 1:浙江浙能技术研究院有限公司
• 2:浙江浙能乐清发电有限责任公司

摘要(Abstract)：

火电厂机炉协调控制系统的控制对象是一个多变量的复杂模型,具有非线性、强耦合、惯性大的特点。针对传统的建模方法缺乏灵活性的缺点,提出一种基于粒子群优化DBN(深度信念网络)的机炉协调系统数据驱动建模方法,以时序数据为基础,采用DBN的无监督贪婪逐层训练算法确定各层网络的权值,引入粒子群优化算法对DBN网络层的神经元数量进行寻优,提高模型精度,最后,结合BP网络在顶层设计联想记忆层实现预测回归分析功能。以660 MW燃煤机组协调系统为试验对象,结果表明,该方法建立的模型具有良好的非线性拟合能力,预测精度高。
The control object of the boiler-turbine coordinated control system in a power plant is a multivariable complex model, which is characterized by nonlinearity, strong coupling and large inertia. Given the poor flexibility of traditional modeling methods, the paper presents a data-driven modeling method for a boiler-turbine coordinated control system based on PSO(particle swarm optimization)-DBN(deep belief network).Based on the time series data, the network weights in each layer are determined by the unsupervised greedy layer by layer training algorithm of DBN, and the number of neurons in the DBN network layer is optimized through PSP to improve the model precision. Finally, the predictive regression analysis function is realized by designing the associative memory layer at the top level based on the BP network. A test on the coordinated control system of a 660 MW coal-fired power generating unit shows that the model established by this method has good nonlinear fitting ability and high prediction accuracy.

关键词(KeyWords)： 深度信念网络;粒子群优化;数据驱动建模;机炉协调系统
deep belief network;particle swarm optimization;data-driven modeling;boiler-turbine coordinated control system

Abstract:

Keywords:

基金项目(Foundation):

作者(Author): 吴林峰,孟瑜炜,俞荣栋,张震伟,徐仙华,王豆,郭鼎
WU Linfeng,MENG Yuwei,YU Rongdong,ZHANG Zhengwei,XU Xianhua,WANG Dou,GUO Ding

DOI: 10.19585/j.zjdl.202009010

参考文献(References)：

• [1]王富强,李晓理,张秋生,等.直流锅炉超超临界机组协调系统多变量预测控制方法[J].热力发电,2016,45(4):48-53.
• [2]李炳楠,朱峰,燕志伟,等.超临界火电机组协调系统建模及模型预测控制算法研究[J].热能动力工程,2020(2):117-125.
• [3]闫姝,曾德良,刘吉臻,等.直流炉机组简化非线性模型及仿真应用[J].中国电机工程学报,2012,32(11):126-134.
• [4]XI Y N,WANG Y S,SONG K B.Improved neural network models for coordinatedcon troller design of supercritical coal-fired power generating unit[C]//12th World Congress on Intelligent Control and Autom ation(WCICA),2016:1274-1279.
• [5]杨优生,王东风,智丹.基于动态模糊神经网络的机炉协调系统建模[J].电力科学与工程,2017,33(7):58-63.
• [6]马良玉,高志元.基于神经网络的超临界机组数学模型[J].动力工程学报,2013,33(7):517-522.
• [7]杨珺,佘佳丽,刘艳珍.基于深度置信网络的时间序列预测[J].深圳大学学报(理工版),2019,36(6):718-724.
• [8]杨国田,王英男,谢锐彪,等.基于变量选择的深度置信神经网络锅炉烟气NOx排放预测[J].热力发电,2020,49(4):34-40.
• [9]李嘉琳,何巍华,曲永志.PSO优化深度神经网络诊断齿轮早期点蚀故障[J].东北大学学报,2019,40(7):974-979.
• [10]方贤,铁治欣,崔仕文,等.基于粒子群优化双支持向量机的SCR烟气脱硝效率预测模型[J].热力发电,2018,47(1):53-58.
• [11]梁智,孙国强,李虎成,等.基于VMD与PSO优化深度信念网络的短期负荷预测[J].电网技术,2018,42(2):598-606.
• [12]徐春华,陈克绪,马建,等.基于深度置信网络的电力负荷识别[J].电工技术学报,2019,34(19):4135-4142.
• [13]NICOLAS L R,YOSHUA B.Representational power of restricted boltzmann machines and deep belief networks[J].Neural Computation,2008,20(6):1631-1649.
• [14]赵亮,刘友波,余莉娜,等.基于深度信念网络的光伏电站短期发电量预测[J].电力系统保护与控制,2019,47(18):11-19.
• [15]杨小明,崔雪,周斌,等.基于粒子群优化支持向量机的短期负荷预测[J].武汉大学学报(工学版),2018,51(8):715-720.
• [16]杨维,李歧强.粒子群优化算法综述[J].中国工程科学,2004(5):87-94.