燃煤电站锅炉炉膛结渣特性计算分析  

Calculation and Analysis on the Slagging Performance of Coal-Fired Boilers

作　　者：刘丹[1] 周熙宏 杨冬[1] 刘朝晖[1] 裘立春 滕敏华 LIU Dan;ZHOU Xihong;YANG Dong;LIU Zhaohui;QIU Lichun;TENG Minhua(State Key Laboratory of Multiphase Flow in Power Engineering,Xi’an Jiaotong University,Xi’an 710049,China;Zhejiang Energy Group R&D Institute Co.,Ltd.,Hangzhou 311121,China)

机构地区：[1]西安交通大学动力工程多相流国家重点实验室,西安710049 [2]浙江浙能技术研究院有限公司,杭州311121

出　　处：《西安交通大学学报》2019年第9期150-158,共9页Journal of Xi'an Jiaotong University

基　　金：国家重点研发计划资助项目(2018YFB0604400)

摘　　要：为对炉膛结渣情况进行有效预测,通过基于燃煤特性的单一指标与多指标综合预测模型和模糊神经网络分别对一台300 MW级亚临界、一台600 MW级亚临界以及两台1 000 MW级超超临界锅炉机组炉膛结渣情况进行了计算分析;针对300 MW级亚临界锅炉机组建立了膜式水冷壁实际热流密度的计算模型,并利用基于污染系数的神经网络对该电站锅炉炉膛结渣情况进行了预测。3种预测模型的结果表明:单一指标和多指标综合预测模型一定程度上可对炉膛结渣情况进行预测,但其分辨率较低,且模型中各指标对于不同煤种和炉型的分辨率存在差异;模糊神经网络相对于上述模型和传统神经网络分辨率较高,所构建的4种模糊神经网络分辨率可分别达到92%、92%、92%以及100%,且统计结果的分辨率也可达到100%,对不同炉型和煤种的适用性更强。另外,基于污染系数的神经网络可根据电站运行数据对炉膛局部结渣情况进行实时预测,误差在3%以内,均方误差为0.013 4,预测结果可为吹灰提供指导。To effectively predict the slagging performance in furnace, a 300 MW sub-critical boiler, a 600 MW sub-critical boiler and two 1 000 MW ultra-supercritical boilers were taken as the test subjects to conduct slagging predication by a single-index prediction modal and a multi-index comprehensive prediction model based on coal-burning behavior and a fuzzy neural network. The calculation model of the actual heat flux density of the membrane water wall was established for the 300 MW subcritical boiler. The pollution coefficient based neural network was used to predict the slagging of the furnace. The results of three prediction models show that the single index and multi-index comprehensive prediction model can predict the slagging of the furnace to some extent but with a low resolution, and the models’ resolutions are different for different coal types and furnace types;the fuzzy neural network has higher resolution than the furnace slagging prediction model based on coal-burning behavior and the traditional neural network. The resolutions of the four kinds of fuzzy neural networks constructed in this study can reach 92%, 92%, 92% and 100%, respectively. In addition, the accuracy of the statistical results can reach 100%, and the fuzzy neural network is more applicable to different furnace types and coal types. The above two methods can only predict the overall slagging situation of the furnace. The neural network based on pollution coefficient can predict local slagging situation of furnace according to the real-time operation data of the boilers, where the error is within 3% and the mean square error is 0.013 4, and it can be used to guide the soot blowing.

关 键 词：炉膛结渣 模糊神经网络 污染系数 

分 类 号：TK229[动力工程及工程热物理—动力机械及工程]