高级检索

基于MI时序处理的GA-BP脱硫制浆系统能耗建模

金秀章, 李奕颖

金秀章, 李奕颖. 基于MI时序处理的GA-BP脱硫制浆系统能耗建模[J]. 南方能源建设, 2019, 6(4): 64-68. DOI: 10.16516/j.gedi.issn2095-8676.2019.04.010
引用本文: 金秀章, 李奕颖. 基于MI时序处理的GA-BP脱硫制浆系统能耗建模[J]. 南方能源建设, 2019, 6(4): 64-68. DOI: 10.16516/j.gedi.issn2095-8676.2019.04.010
JIN Xiuzhang, LI Yiying. Energy Consumption Modeling of GA-BP for Desulfurization Pulping System Based on MI Time Series Processing[J]. SOUTHERN ENERGY CONSTRUCTION, 2019, 6(4): 64-68. DOI: 10.16516/j.gedi.issn2095-8676.2019.04.010
Citation: JIN Xiuzhang, LI Yiying. Energy Consumption Modeling of GA-BP for Desulfurization Pulping System Based on MI Time Series Processing[J]. SOUTHERN ENERGY CONSTRUCTION, 2019, 6(4): 64-68. DOI: 10.16516/j.gedi.issn2095-8676.2019.04.010
金秀章, 李奕颖. 基于MI时序处理的GA-BP脱硫制浆系统能耗建模[J]. 南方能源建设, 2019, 6(4): 64-68. CSTR: 32391.14.j.gedi.issn2095-8676.2019.04.010
引用本文: 金秀章, 李奕颖. 基于MI时序处理的GA-BP脱硫制浆系统能耗建模[J]. 南方能源建设, 2019, 6(4): 64-68. CSTR: 32391.14.j.gedi.issn2095-8676.2019.04.010
JIN Xiuzhang, LI Yiying. Energy Consumption Modeling of GA-BP for Desulfurization Pulping System Based on MI Time Series Processing[J]. SOUTHERN ENERGY CONSTRUCTION, 2019, 6(4): 64-68. CSTR: 32391.14.j.gedi.issn2095-8676.2019.04.010
Citation: JIN Xiuzhang, LI Yiying. Energy Consumption Modeling of GA-BP for Desulfurization Pulping System Based on MI Time Series Processing[J]. SOUTHERN ENERGY CONSTRUCTION, 2019, 6(4): 64-68. CSTR: 32391.14.j.gedi.issn2095-8676.2019.04.010

基于MI时序处理的GA-BP脱硫制浆系统能耗建模

基金项目: 

国家科技重大专项资助“煤炭清洁高效利用和新型节能技术” 2016YFB0600701

详细信息
    作者简介:

    金秀章 1969-,男,河北衡水人,博士,华北电力大学控制与计算机工程学院副教授,硕士生导师。近期主要研究兴趣:先进控制策略在大型电力机组的控制研究、信息融合技术(e-mail)jinxzsys@ 163.com。

    李奕颖(通信作者) 1995-,女,吉林吉林人,华北电力大学硕士生,主要从事机器学习算法研究及火电厂控制策略研究(e-mail)15733220440@163.com。

  • 中图分类号: TM611,TM73

Energy Consumption Modeling of GA-BP for Desulfurization Pulping System Based on MI Time Series ProcessingEn

  • 摘要:
      [目的]  石灰石—石膏湿法脱硫是目前燃煤电厂应用最广泛、技术最为成熟的一种烟气脱硫技术,石灰石浆液制备作为其中一道高耗能工序,具有生产过程复杂,物耗与能耗间存在非线性关系等特点,缺乏合理有效的能耗模型。为建立能够对生产参数优化提供指导的可靠的制浆系统能耗模型。
      [方法]  基于某600 MW电厂实际运行数据,选择生产过程中的可控制量作为输入,并基于互信息(Mutual Information)理论调整各输入变量间的时滞关系,采用结合遗传算法(Genetic algorithm,GA)的改进BP神经网络建立了制浆系统的单位制浆能耗模型。
      [结果]  试验结果表明:与未调整时序的GA-BP模型和标准BP算法的模型相比,经过时序调整的GA-BP模型的计算结果能够更为准确地接近制浆系统生产实际数据。
      [结论]  所建立的模型可以应用到浆液制备过程的能耗优化研究中。
    Abstract:
      [Introduction]  Limestone-gypsum wet desulfurization is a flue gas desulfurization technology with the most extensive application and the most mature technology in coal-fired power plants. Limestone slurry preparation is one of the high-energy -consuming processes, which has complex production process and difficult to correspond to material consumption and energy consumption. There is no reasonable and effective energy consumption prediction model. In order to establish a reliable pulping system energy consumption model that can guide the optimization of production parameters.
      [Method]  Based on actual operating data of a 600 MW power plant, the controllable quantity in the production process was selected as input, and the time-delay relationship between each input variable was adjusted by mutual information theory. The improved BP neural network combined with genetic algorithm (GA) was used to establish the unit pulp energy consumption model of the pulping system.
      [Result]  The experimental results show that compared with the unadjusted timing GA-BP model and the standard BP algorithm model, the calculation results of the GA-BP model with time series adjustment can more accurately approach the actual production data of the pulping system.
      [Conclusion]  The established model can be applied to the energy optimization study of the slurry preparation process.
  • 图  1   GA-BP算法流程图

    Figure  1.   GA-BP algorithm flow chart

    图  2   BP模型(未调时序)预测结果

    Figure  2.   BP model (unregulated timing) prediction results

    图  3   GA-BP模型(未调时序)预测结果

    Figure  3.   GA-BP model (unregulated timing) prediction results

    图  4   GA-BP模型(调整时序)预测结果

    Figure  4.   GA-BP model (adjusted timing) prediction results

    表  1   辅助变量与耗电量互信息值及时延表

    Table  1   Auxiliary variable and power consumption mutual information value and time delay

    辅助变量 MI最大值 时延/s
    给料量 0.004 2 275
    稀释水流量 0.127 5 247
    研磨水流量 0.113 8 283
    再循环箱浆液密度 0.001 1 244
    循环泵频率 0.003 6 0
    旋流器出口压力 0.002 9 261
    球磨机电流 0.012 8 236
    下载: 导出CSV

    表  2   预测性能对比

    Table  2   Comparison of predicted performance

    预测样本数 RMSE/(kWh·t-1) MRE/% 迭代步数
    BP(未调时序) 0.062 4 5.19 278
    GA-BP(未调时序) 0.042 0 3.92 126
    GA-BP(调整时序) 0.018 8 1.55 133
    下载: 导出CSV
  • [1] 吕新锋. 石灰石-石膏湿法烟气脱硫设施常见故障及影响脱硫效率因素分析 [J]. 电力科技与环保,2018,34(2):27-29.
    [2]

    HAO R,ZHANG Y,WANG Z,et al. An advanced wet method for simultaneous removal of SO2 and NO from coal-fired flue gas by utilizing a complex absorbent [J]. Chemical Engineering Journal,2017,307:562-571.

    [3] 李伟科. 火电厂排放烟气白雾治理新方案及节能分析 [J]. 南方能源建设,2018,(4):50-54.
    [4] 杨伟,倪黔东,吴军基. BP神经网络权值初始值与收敛性问题研究 [J]. 电力系统及其自动化学报,2002(1):20-22.
    [5] 史坤鹏,乔颖,赵伟,等. 计及历史数据熵关联信息挖掘的短期风电功率预测 [J]. 电力系统自动化,2017,41(3):13-18.
    [6]

    SHANNON C. A mathematical theory of communication [J]. Bell System Tech. J.,1948,27(4):379-423.

    [7]

    WANG Z C,LI M Q,LI J Z. A multi-objective evolutionary algorithm for feature selection based on mutual information with a new redundancy measure [J].Information Sciences,2015,307(1):73-78.

    [8] 赵彦涛,单泽宇,常跃进,等. 基于MI-LSSVM的水泥生料细度软测量建模 [J]. 仪器仪表学报,2017,38(2):487-496.
    [9] 刘吉臻,秦天牧,杨婷婷,等. 基于偏互信息的变量选择方法及其在火电厂SCR系统建模中的应用 [J]. 中国电机工程学报,2016,36(9):2438-2443.
    [10] 马平,李珍,梁薇. 基于互信息的辅助变量筛选及在火电厂NOx软测量模型中的应用 [J]. 科学技术与工程,2017,17(22):249-254.
    [11]

    CELLUCCI C J,ALBANO A M,RAPP P E. Statistical validation of mutual information calculations:comparison of alternative numerical algorithms [J]. Physical Review E,2005,71(6):066208.

    [12] 冯长敏,张炳江. 基于BP神经网络的分段函数连续优化处理 [J].北京信息科技大学学报(自然科学版),2019,34(1):18-22.
    [13]

    SHAN Y,YI J L,F J G. The Application of BP neural network algorithm in optical fiber fault diagnosis [C]// IEEE. International Symposium on Distributed Computing & Applications for Business Engineering & Science,Guiyang,2015. Guiyang:IEEE,2015.

    [14] 陈浩广,王银河. 基于扩展BP神经网络的一类非线性系统自适应控制设计 [J]. 计算机应用,2017,37(6):1670-1673.
    [15] 张铃,张钹. 遗传算法机理的研究 [J]. 软件学报,2000,11(7):945-952.
    [16] 邬小龙,郭兵,沈艳. 改进BP神经网络在软件能耗分析中的应用 [J]. 计算机工程,2012,38(10):185-187.
图(4)  /  表(2)
计量
  • 文章访问数: 
  • HTML全文浏览量: 
  • PDF下载量: 
  • 被引次数: 0
出版历程
  • 收稿日期:  2019-03-20
  • 修回日期:  2019-05-19
  • 刊出日期:  2020-07-10

目录

    LI Yiying

    1. On this Site
    2. On Google Scholar
    3. On PubMed

    /

    返回文章
    返回