基于BP神经网络锅炉炉膛火焰可视化监测方法研究

RESEARCH ON VISUALISED MONITORING METHOD FOR BOILER FURNACE FLAME BASED ON BP NEURAL NETWORK

为了可视化监测炉膛火焰燃烧状况,提出一种基于级联前向BP神经网络模型的锅炉炉膛火焰可视化监测方法。通过比较选取级联前向BP神经网络作为炉膛温度预测模型,利用图像处理技术得到炉膛火焰辐射能图像对应二维温度场,并采用正则化方法重建炉膛火焰三维温度场。仿真结果表明,根据二维温度场可得火焰等温线走向和分布,根据三维温度场易得火焰中心分布及全炉最高温度点信息,实现锅炉运行控制和异常温度报警,满足燃烧诊断要求并实现炉膛火焰可视化监测。

In order to monitor the Combustion status of furnace flame in visualised way, we propose a visualised boiler furnace flame monitoring method which is based on cascade forward BP neural network model. Through comparison we select the cascade forward BP neural network as the furnace temperature prediction model, and use image processing technique to get the furnace flame radiation images to correspond to the two-dimensional temperature field. Furthermore, we use regularisation method to reconstruct the three-dimensional temperature field of furnace flame. Simulation results show that by using two-dimensional temperature field it is able to get the trend and distribution of the flame isotherm, and by using three-dimensional temperature field it is easy to get the distribution of flame centres and the information of highest temperature point in whole furnace, thus the boiler operation control and abnormal temperature alarm are achieved, this meets the eombustion diagnostic requirements and realises the visualised furnace flame monitoring.