Abstract:

The load signal in coal-fired power plants is calculated based on steam parameters, but the thermal inertia of the furnace process causes a lag in steam parameter changes, affecting the load monitoring and adjustment decisions of the operators. Therefore, this paper proposed an improved load monitoring method based on the time series alignment of flame image features, leveraging the ability of flame images to reflect the actual conditions inside the furnace. This method firstly converted boiler flame monitoring videos into continuous image frames, then extracted flame temporal features through a series of adaptive preprocessing steps, including filtering and segmentation. Dimensionality reduction was then applied to obtain the flame image indicators, and finally, dynamic time warping (DTW) algorithm was used to align the flame image indicators with the load signal, generating an alignment path and time-delay curve between them. Analysis showed that the DTW algorithm significantly enhanced the correlation between flame image indicators and load signals under typical operating conditions (50%, 80%, and 100% load), providing an efficient solution for monitoring load fluctuations in coal-fired power plants.

Key words: coal-fired power plant, flame image processing, dynamic time warping, time-delay analysis, load monitoring

摘要：

燃煤发电厂的负荷信号往往基于蒸汽侧参数计算得到，而炉膛热力过程的热惯性会引起蒸汽参数的变化存在滞后性，从而影响运行人员的负荷监测和调整判断。因此，本文从炉膛火焰图像反映炉膛内部实际状况的特点出发，提出了一种基于火焰图像时序特征规整的改进锅炉负荷监测方法。该方法首先对锅炉火焰监控视频提取转换成连续图像帧，并经过滤波、分割等一系列自适应预处理步骤提取火焰时序特征，再通过降维得到火焰强度指标，最后将强度指标与负荷信号两时序信号进行动态时序规整计算，从而获得强度指标与负荷信号的规整对齐路径和时间延迟曲线。分析表明，动态时序规整算法在50%、80%、100%等各个典型负荷工况下使用图像监测指标对负荷信号的弹性关联提升效果明显，为燃煤发电厂等复杂工业系统的锅炉负荷变动监测和分析提供了一种高效准确的方法。

关键词: 燃煤发电厂, 火焰图像处理, 动态时间规整, 时延分析, 负荷监测