Abstract:

The real-time monitoring of the flame in the furnace of coal-fired power plants is crucial for both the economics of power generation and the safe operation of the boiler. Traditional fire detection techniques based on energy signals such as light, heat and radiation can only detect the presence or absence of flame. These techniques are gradually unable to meet the increasingly stringent requirements for fine-grained "peaking" of thermal power generation. In this study, the features of flame images from actual power plants were analyzed from multiple perspectives. By leveraging an improved version of the Inception deep convolutional neural network (DCNN) for flame state classification, the multi-dimensional characteristics of flame were extracted. And a dataset was made by in-depth analysis of the flame image characteristics of the burner. At the same time, the preprocessed images of different categories of flames were used to create a flame image dataset. The Inception DCNN models were constructed to achieve flame state classification based on automatic feature extraction. It was proposed to classify the flame state of the burner based on the Siamese-Inception DCNN. It was found that the improved Siamese-Inception DCNN model, which converted the flame state classification problem into an evaluation of state similarity, was proposed indirectly to achieve the classification objective. The recognition accuracy of the network architecture reached 99.86%.

Key words: burner flame state monitoring, coal-fired power plants, Inception convolutional neural network, siamese network

摘要：

燃煤电厂炉膛火焰的实时监测关系到发电经济性和锅炉的安全运行，而基于光能、热能和辐射能等能量信号的传统火检技术仅能探测火焰的有无，已无法满足日益迫切的火力发电精细化“调峰”需求。本文以实际电厂燃烧器火焰图像为研究对象，应用基于改进的Inception深度卷积神经网络（deep convolutional neural network, DCNN）的火焰状态分类方法，通过深入分析燃烧器火焰图像特点，对火焰多维度特征进行提取并制作数据集，同时将预处理后的不同类别火焰图像制作成火焰图像数据集，构建Inception DCNN，实现自动特征提取的火焰状态分类，并提出基于孪生Inception DCNN对燃烧器火焰状态进行分类。结果表明，改进的孪生Inception DCNN网络模型将火焰的状态分类问题转化为评价状态相似度问题，间接实现分类目标，识别准确率达到99.86%。

关键词: 燃烧器火焰状态监测, 燃煤电厂, Inception深度卷积神经网络, 孪生网络