CAJ | 학술논문

以某甲醇合成塔为研究对象，综合考虑甲醇生产过程各因素对甲醇转化率的影响，采用模糊神经网络技术，建立了甲醇转化率的软测量模型。为了提高模糊神经网络建模的精度，提出了一种改进的差分进化算法优化模糊神经网络的参数，从而获得了具有较高精度的软测量模型。以某甲醇合成塔的实际运行数据为样本进行甲醇转化率的软测量建模，结果表明该模型具有较高的精度，能够实现甲醇转化率的实时监测。
이모갑순합성탑위연구대상，종합고필갑순생산과정각인소대갑순전화솔적영향，채용모호신경망락기술，건립료갑순전화솔적연측량모형。위료제고모호신경망락건모적정도，제출료일충개진적차분진화산법우화모호신경망락적삼수，종이획득료구유교고정도적연측량모형。이모갑순합성탑적실제운행수거위양본진행갑순전화솔적연측량건모，결과표명해모형구유교고적정도，능구실현갑순전화솔적실시감측。
The soft sensor model for methanol conversion is constructed by adopting the fuzzy neural network technology with consideration of all the influence factors on the methanol conversion rate during the manufacture process with one methanol synthesis column as the research object. To increase the accuracy for constructing fuzzy neural network model, one improved differential evolution algorithm （DE） is proposed to optimize fuzzy neural network parameters and obtain the soft sensor model with higher accuracy. The soft sensor model for methanol conversion rate is built by using actual running data of one methanol synthesis column as the sample. The simulation results show that the soft-sensor model is of great accuracy, and can be used to realize real-time monitoring of methanol conversion rate.