摘要
针对秸秆燃料送往循环流化床锅炉燃烧发电时,其热值无法在线估量,计量实效性低的问题,设计了一套用于秸秆燃料热值估计的图像分析系统;该系统由工业摄像机、图像分析主机、服务器和工业互联通信网络总线组成;系统通过工业摄像机在秸秆燃料进入锅炉前进行图像采集,并将采集到的图像通过高速差分信号传送给图像分析主机,主机采用改进的U-Net深度学习网络对图像进行分割;得到的分类结果结合从服务器读取的秸秆燃料的成分组成和组分的热值,加之实时返回的秸秆燃料含水率等参数,通过热值计算公式实时计算出燃料热值;测试结果表明,基于改进U-Net深度学习网络的图像分割算法分割效果较好,平均精度(mean Average Precision)达到0.86,平均重合度(mean Intersection over Union)达到0.68,可以满足燃料热值的在线估量要求。
n view of the difficulty and low effectiveness of calorific value estimation while the straw fuel is sent to CFB boiler for combustion power generation,an image analysis system is designed to resolve the problem.The system consists of an industrial camera,an image analysis host,a server and an industrial interconnected communication network bus.The system uses industrial camera to gather images of straw fuel before entering the boiler,and the collected images are sent to the image analysis host through a high speed differential signal,the host uses the improved U-Net deep learning network to segment the image.The classification results are combined with the composition of the straw fuel and the calorific value obtained from the server,together with the water content of straw fuel returned in real time,then the calorific value of fuel can be calculated in real time based on caloric value formula.The test results show that the image segmentation algorithm based on improved U-Net deep learning network has nice segmentation effect.The mean Average Precision is over 0.86 and the mean Intersection over Union is over 0.68,which can meet the online estimation requirement of fuel calorific value.
作者
郑雅羽
陈超
梁圣浩
何德峰
李廉明
Zheng Yayu;Chen Chao;Liang Shenghao;He Defeng;Li Lianming(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023,China;Jiaxing New Jies Thermal Power Co.,Ltd.,Jiaxing 314016,China)
出处
《计算机测量与控制》
2018年第11期261-266,共6页
Computer Measurement &Control
基金
国家自然科学基金(61773345)
浙江省重点研发计划项目(2017C01073)
关键词
秸秆
热值估计
图像分割
深度学习
U-Net
straw fuel
calorific value estimation
image segmentation
deep learning
U-Net