基于神经网络的钢铁企业电厂煤汽比预测模型

FORECAST MODEL OF GAS-STEAM RATIO BASED ON NEURAL NETWORK IN POWER PLANT OF IRON AND STEEL WORKS

以钢铁企业自备电厂锅炉煤汽比为研究对象,以灰色关联度分析为基础,从理论和数据两方面分析了影响煤汽比的主要因素及各因素对煤汽比影响程度的大小.结果表明,各因素对煤汽比的影响程度为排烟温度>热风温度>给水温度>空燃比>烟气含氧量.基于BP神经网络预测方法,建立了自备电厂锅炉煤汽比预测模型,此BP神经网络为5-12-1结构,隐含层和输出层分别用tansig,purelin函数传递,利用改进动量梯度下降优化算法traingdm训练网格.预测结果表明,该模型网络训练值与实际值较吻合,相关系数R达到0.993 7,用建立的网络进行预测,预测的相关系数为0.976 2,平均误差绝对值为3.9%,在可控范围之内,证明了网络的可靠性与良好的泛化推广能力,可用来指导实际生产.

With the gas-steam ratio of self-provided power plant in an iron and steel works taken as an object,flue gas temperature,hot air temperature,feed water temperature,air fuel ratio and oxygen content in flue gas are the major factors influencing gas-steam ratio,which is analyzed by grey relation analysis.A prediction model of gas-steam ration of self-provided power plant is established on the basis of BP neural network,which is a 5-12-1 network structure,the hidden layer and output layer is transferred by tansig and purelin function respectively,momentum gradient descent optimization algorithm,traingdm is also used to train network.The results show that the model can effectively predict the gas-steam ratio of boiler,the correlation coefficient of actual values and training ones is 0.993 7,and the correlation coefficient of actual values and prediction ones is 0.976 2,the mean absolute error is also controlled within the scope.Showing a good generalization ability and outreach capacity,we can provide a theoretical basis and guide for the real production.