【作者】 祖春光；

【导师】 周怀春；

【作者基本信息】 华中科技大学 ， 热能工程， 2006， 硕士

【摘要】 电站锅炉燃烧的稳定性与经济性一直是热能工程领域中一个复杂而又非常重要的问题,而炉膛内煤粉的燃烧是一个发生在大空间范围、剧烈脉动的、具有明显三维特征的复杂物理化学过程。研究合理有效的全炉膛火焰检测、诊断及分析的手段,并建立可靠的锅炉燃烧效率计算模型对优化锅炉运行具有非常重要的意义。本文通过在一台200MW机组上已安装的炉膛三维温度场监测系统的火焰检测装置基础上,从拍摄的火焰图像中提取出反映全炉膛燃烧的辐射能信息,建立其与机组功率之间的动态模型,以进行燃烧监测与控制,同时进行稳定性的分析。进而发现该炉在负荷波动时,炉内辐射能剧烈变动,表明该炉燃烧调整过度,是导致经常灭火的主要原因。同时,从综合得出的反映全炉膛火焰图像中提取特征量,进行稳定性的分析和灭火的判断,并用锅炉灭火记录进行验证,结果表明这种方法能够很好地诊断燃烧状况。同时,锅炉燃烧的经济性及优化运行需要准确的锅炉效率计算模型,而运行中其影响因素繁多且复杂。BP人工神经网络作为一种非线性映射手段,具有传统优化求解法所不具备的很多优点,适合于解决该类问题。本文应用VC调用Matlab的方法实现在线训练BP神经网络,并用具体的试验数据说明训练的过程,得到了很好的锅炉效率预测模型。最后,用训练好的BP网络对检验样本进行验证,验证结果是比较令人满意的。并在此基础上,针对具体电厂开发了选煤及燃烧优化专家系统,并在现场得到了初步的应用,从而更大程度地提高锅炉运行的经济性。更多还原

【Abstract】 It is a complex and important issue that the combustion stability and economy of power station boilers in the field of thermal energy engineering. While the combustion of pulverized-coal in furnace is a complex physical and chemical reaction, which is occurred under the condition of large space, frequent fluctuation and having conspicuous three- dimension character. Therefore, it is significant to study on effective methods of flame detection, diagnosis and analysis, and to establish reliable computing-models of boiler combustion efficiency to optimize the operation of boilers.In order to monitor and control the state of combustion and analyze its stability at the same time, the essay establish a dynamic model between sets’power and radiant energy signals which reflects the combustion status of whole furnace from flame images derived from the flame detection equipment of the three-dimension temperature field monitoring system built in a 200MW boiler unit. It is found that the radiant energy always fluctuates violently when the boiler’s power changes, indicating that the main reason for frequent flame failures is the excessive combustion adjustment. Besides, analyzing the stability and diagnosing the combustion status are carried out by distilling characteristic parameters from the flame images which reflect the whole furnace. According to a record of the whole furnace extinction accident, it has been verified that the method could diagnose the combustion status well.Generally, economical combustion and optimized operation of boilers need accurate model of boiler efficiency, while there are many complex influencing factors during the running period of boilers. As a non-line mapping means, compared with traditional optimizing methods, BP neural networks has much more advantages which can help to deal with this kind of problem. Herein, by carrying out online training BP neural networks using VC to call MatLab, and illuminating the process with specific test data, a nice prediction model for boiler efficiency is obtained. Finally, through verifying check data using trained network the result is found to be satisfactory and based on which, an expert system of selecting coals and optimizing combustion is explored according to a certain power plant. The system has applied to the field of a power plant preliminarily, proving that it could improve the economy of boiler operation to further degree.更多还原