基于神经网络的固体未完全燃烧热损失研究

The Study of Mechanical Incomplete Combustion Heat Loss based on Neural Network

燃煤工业锅炉的固体未完全燃烧热损失q4的常规测试方法,需现场采集飞灰、漏煤和炉渣并由实验室化验完成,测试成本高,周期长;对于运行中的锅炉,很难对其燃烧状况进行在线有效判断。为了实时监测燃煤锅炉的固体未完全燃烧热损失,现基于MATLAB的BP神经网络,建立锅炉固体未完全燃烧热损失q4的预测模型,可快速预测出q4。并在此基础上,建立模糊神经网络,完成对q4的优化。对于锅炉q4的预测模型,结果显示预测值与期望值之间的平均误差在0.538 8%,基本能满足实时在线检测时计算固体未完全燃烧热损失q4和及时调整锅炉运行工况所需,从而有效提高锅炉运行的经济性。

The conventional testing methods for the mechanical incomplete combustion heat loss（ q4）,need to collect fly ash drain coal,slag from the field and send it to the laboratory. The cost of which is too high and last long period. For the boiler in operation,it is hard to make a reasonable judgment about its combustion on-line. In order to monitor the mechanical incomplete combustion loss of the boiler in time. the prediction model of boiler heat efficiency is established based on MATLAB BP neural network which q4 can be predicted quickly. On the basis of it,the fuzzy neural network is established to optimize the q4. The results show that the error between the predicted value and the real value is 0. 538 8%. It can meet the requirement of on-line calculating the mechanical incomplete combustion heat loss and adjust the operating mode in time to improve the efficiency of boiler operation.