秸秆颗粒燃料成型参数模糊神经网络预测系统研究

Study on Fuzzy Neural Network Predictive System of Straw Briquetting Fuel Parameters

为寻找秸秆颗粒燃料成型过程的最佳生产参数,通过构建成型参数预测系统,采用正交实验分析原料含水率、电机转速2个成型参数对成型燃料松弛密度、单位时间产量和成品燃料热值3个成型技术指标的影响规律,基于实验数据建立成型工艺参数数据库,利用Matlab建立密度、热值、生产率的BP(Back Propagation)神经网络预测模型和模糊神经网络预测模型,将2种模型的预测值与实验值进行对比,得出模糊神经网络预测模型的相对误差平均值为0.72%,预测效果优于BP神经网络预测模型。对建立的模糊神经网络成型参数预测系统进行验证实验,预测值与实验值的相对误差最大为2.08%,绝大部分预测值的相对误差小于1%。研究表明,该模糊神经网络预测系统预测效果较好,能够满足实际预测要求,可运用到秸秆燃料成型实际生产预测中。

In order to find the optimal production parameters of straw pellet fuel molding process,through the construction of molding parameter prediction system,the orthogonal experiment was used to analyze the influence of two molding parameters of raw material moisture content and motor speed on three molding technical indexes of molding fuel relaxation density,unit time yield and finished fuel calorific value.Based on the experimental data,the database of molding process parameters was established,and the BP(Back Propagation)neural network prediction model and fuzzy neural network prediction model of density,calorific value and productivity were established by Matlab.The predicted values of the two models were compared with the experimental values.The average relative error of fuzzy neural network prediction model was 0.72%,and the prediction effect was better than that of BP neural network prediction model.The established fuzzy neural network molding parameter prediction system was verified by experiments.The maximum relative error between the predicted value and the experimental value was 2.08%,and the relative error of most predicted values was less than 1%.The research showed that the fuzzy neural network prediction system had good prediction effect and can meet the actual prediction requirements,which can be applied to the actual production prediction of straw fuel molding.