Abstract: Based on dynamic flame image of furnace, the way to analyze the combustion stability is presented. By analyzing the characteristic of the flame image, the character values of image are proposed. A method judging combustion stability by online clustering is presented by using these character values of image. Characteristic of dynamic image is utilized, and the states of flame are judged. The results show that the accuracy of online method is increased 5.3% than that of offline learning method. This method can judge the combustion stability effectively and has practical significance to monitor the combustion states automatically and ensure the boiler operation is safe.

Key words: boiler, combustion, flame stabilization, online fuzzy clustering method

摘要： 针对锅炉燃烧监控系统所采集的火焰动态图像，提出了一种基于在线模糊聚类算法的炉内火焰燃烧诊断方法。该方法分析了火焰图像的特点，提取了判别火焰稳定性的特征量，以提取的特征量作为在线模糊聚类算法的输入参数，分析燃烧图像的隶属度，给出判别标准对燃烧稳定性进行综合评估。将在线算法与离线算法进行比较，实验结果表明，在线算法比离线算法的准确率提高了5.3%，验证了算法的有效性。该方法对实现燃烧状态自动监测，保障锅炉安全运行具有重要意义。

关键词: 锅炉, 燃烧, 火焰稳定性判别, 在线模糊聚类算法