基于机器学习的螺旋管圈水冷壁壁温预测模型研究  

Research on prediction model for spiral water-cooled wall temperature based on machine learning

作　　者：袁茂博 邓磊[1] 刘雪敏[2] 杨凯镟 梁永 刘虎 笪耀东[2] 车得福[1] YUAN Maobo;DENG Lei;LIU Xuemin;YANG Kaixuan;LIANG Yong;LIU Hu;DA Yaodong;CHE Defu(State Key Laboratory of Multiphase Flow in Power Engineering,Xi’an Jiaotong University,Xi’an 710049,China;China Special Equipment Inspection and Research Institute,Beijing 100029,China;Shanghai Power Equipment Research Institute Co.,Ltd.,Shanghai 200240,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,陕西西安710049 [2]中国特种设备检测研究院,北京100029 [3]上海发电设备成套设计研究院有限责任公司,上海200240

出　　处：《热力发电》2023年第8期32-39,共8页Thermal Power Generation

基　　金：国家重点研发计划项目(2017YFB0602102)。

摘　　要：针对某600 MW机组四角切圆锅炉建立了数值计算模型,综合考虑了一次风率、主燃区过量空气系数、燃烧器摆角以及SOFA喷嘴竖直摆角等运行参数的影响,设计L16(45)正交工况获得了100%BMCR、75%THA、50%THA以及35%BMCR负荷下锅炉水冷壁壁面的热流量,各工况下螺旋管圈水冷壁的壁温分布通过耦合壁面吸热量、水动力特性与壁温计算得到。由于正交工况参数设置的不连续性,建立了机器学习模型,实现了正交工况覆盖参数范围内的螺旋管圈水冷壁壁温分布预测。研究结果表明:在亚临界工况下,螺旋管圈水冷壁在燃烧器高度区域内出现730 K的温度峰值;锅炉在变负荷过程中,当炉膛火焰中心高度与管内工质相变起始高度重合时,易发生传热恶化导致壁温激升;机器学习算法中集成学习算法在壁温数据的训练集和测试集上拟合优度R2均达到了0.99,能够适用于宽负荷下锅炉水冷壁壁温预测。同时,机器学习算法建立了壁温分布与锅炉运行参数之间的映射关系,后续研究可通过优化算法合理调整优化运行参数,保障水冷壁的壁温安全。In the study,a computational fluid dynamics(CFD)model based on a 600 MW tangentially coal-fired boiler was established.According to orthogonal conditions(L16(45)),the heat flux distributions of the water-cooled wall under 100%BMCR,75%THA,50%THA and 35%BMCR loads were obtained.In addition,the factors also included:primary to secondary air rate,degree of air-staging,swing angles of burners and SOFA nozzles.Then,the spiral water-cooled wall temperature distributions under various conditions were calculated through coupling the heat absorption,temperature calculation and hydrodynamic characteristics of the water-cooled wall.Due to the discontinuity of orthogonal condition,the machine learning was used for predicting the spiral water-cooled wall temperature distribution within the range of parameters covered by orthogonal conditions.The results showed that a wall temperature peak up to 730 K would appear in the area among burner system.The heat transfer deterioration was easy to occur when the flame center height in furnace coincided with the phase change height of the working fluid during the boiler load adjusting process.The goodness of fit R2 of the ensemble learning on the training set and the test set of the wall temperature data had reached 0.99,which could be used to predict the wall temperature of the boiler under wide load.At the same time,the machine learning established the mapping relationship between the wall temperature distribution and the operating parameters of the boiler.In the future study,the wall temperature safety of the water wall can be guaranteed by reasonably adjusting and optimizing the operating parameters through the optimization algorithm.

关 键 词：螺旋管圈水冷壁 热流密度分布 壁温分布 正交工况 机器学习 

分 类 号：TM621.2[电气工程—电力系统及自动化] TP181[自动化与计算机技术—控制理论与控制工程]