链箅机预热段温度场模型的贝叶斯-BP神经网络系统辨识

Identification on Bayes-BP neural network system for temperature field model of preheating section of chain grate

针对球团链箅机预热段温度场因非线性、时滞性、不确定性等特点难以通过传统的理论分析方法建立数学模型的问题,建立贝叶斯-BP神经网络,对该温度场模型进行系统辨识,对比模型预测输出值与实际系统输出值,通过仿真与实验分析该辨识模型的拟合效果。结果表明:贝叶斯-BP神经网络拟合效果较好,其线性拟合度近似为1,最终预测误差约为0.014 K,预测相对误差在5%范围内,构建的预热段温度场模型准确可靠且适用性强,可为预热段温度场均衡稳定控制提供理论指导。

Aiming at the problem that the temperature field of the preheating section of the chain grate for pellet features non-linearity,time lag,uncertainty,etc.,which is difficult to establish its mathematical model through traditional theoretical analysis methods,a Bayesian-BP neural network is established to systematically identify the temperature field model,compare the predicted output value of the model with the actual system output value,and conduct simulation and experiment on the fitting effect of the identification model.The results show that the Bayesian-BP neural network has a good fitting effect,the linearity is approximate 1,the final prediction error is about 0.014 K,and the relative prediction error is within 5%.The constructed temperature field model of preheating section is accurate,reliable and applicable,which can provide theoretical guidance for the balanced and stable control of the temperature field of the preheating section.