论文部分内容阅读
基于发酵生产的特点及建模要求,以某企业燃料乙醇生产过程为研究对象,利用工业生产中的参数及数据,建立了以乙醇发酵效果为目标的BP神经网络模型,以静态模型反映复杂的动态问题。探讨了乙醇发酵生产模型的误差产生原因,并提出改进方案,根据已有经验将相关参数进行了合理组合,调整神经网络模型的输入输出参数结构,以提高仿真模拟效果。通过多次模型改进,使模拟的平均相对误差从10%提高至5.4%,为进一步研究发酵生产建模提供了思路。
Based on the characteristics of fermentation production and the modeling requirements, a BP neural network model targeting at ethanol fermentation was established by taking the fuel ethanol production process of a certain enterprise as the research object. Based on the parameters and data of industrial production, the static model was used to reflect the complex Dynamic issues. The reasons for the error of the ethanol fermentation production model were discussed, and the improvement scheme was proposed. Based on the existing experience, the relevant parameters were rationally combined to adjust the input / output parameter structure of the neural network model to improve the simulation results. Through multiple model improvement, the average relative error of simulation was increased from 10% to 5.4%, which provided a new idea for further research on fermentation production modeling.