采用数据融合方法诊断电站锅炉的燃烧稳定性

A Way of Diagnosing Combustion Stability of Utility Boilers by Data Fusion

电站锅炉燃烧稳定性易受锅炉负荷、煤质等因素影响,单一的火检信号或炉膛压力信号所反映的燃烧状态信息,既不准确也不完整。为提高对炉内燃烧状态的判别质量,引入了数据融合的方法。通过采用D-S证据理论方法对常规火检信号和炉膛压力压信号进行融合,得到综合诊断结果,并提出了一种简单可行的基本概率分配方法。对某电厂监控信息系统中保存的某一时间段的运行数据进行分析验证。结果表明:基于证据理论融合后的诊断结果较各传感器单独决策的结果精度高、可靠性强,不确定性大大减小,从而能对燃烧状态作出更有效的诊断。

Combustion stability of utility boilers is easily affected by a number of factors like boiler output, coal quality etc. The use of single flame monitoring signals, or furnace pressure signals, to reflect states of combustion can neither be called correct nor sufficient. Therefore, for improving the discrimination quality concerning the state of combustion in the furnace, the concept of data fusion is introduced. Judgment is obtained by fusing traditional flame monitoring signals with furnace pressure signals by making use of the Dempster-Shafer （D-S） Evidence Theory. Thus a simple feasible way of basic probability assignment is presented. An analysis and verification of operational data, stored during a certain time interval in the supervisory information system of a certain power plant, shows that a diagnosis result based on fusion, performed in accordance with the evidence theory, is more accurate and reliable, with less uncertainty, than any decision made by relying solely on informations obtained from any single sensor, and thus a more effective judgment concerning the state of combustion is offered.