论文部分内容阅读
在文献基础上分析了热虹吸式再沸器的两段机理模型,分析了影响再沸器运行的各变量关系。选择了液位、导热油入口温度与流量、精馏塔釜温度作为数据模型预测再沸器热量的输入变量。采用浅层神经网络方法建立数据模型,当隐含层只有两层时表现出了较好的运算速度与拟合精度的平衡。采用工业现场数据建立再沸器热量的软测量模型。考虑到再沸器的性能在不断变化,比较了几种数据采集时间策略。当训练样本不包含检测样本时,可以看到再沸器运行过程中,拟合精度将会随着设备参数、控制参数的变化而逐渐下降,均方根误差依然保持在2.5%以内。但当现场设备参数发生较大变化后,数据模型的适用性变差,需要重新训练以达到原来的精度。
Based on the literature, the two-stage model of the thermosyphon reboiler was analyzed and the relationship between variables affecting the reboiler operation was analyzed. The choice of liquid level, HTF inlet temperature and flow rate, distillation tower temperature as the data model to predict reboiler calorie input variables. Using shallow neural network method to establish the data model, when the hidden layer only two layers showed a good balance between computing speed and fitting accuracy. Soft-sensing model for reboiler heat build-up using industrial field data. Considering that the performance of the reboiler is constantly changing, several data collection time strategies are compared. When the training sample does not contain the test sample, it can be seen that during the operation of the reboiler, the fitting accuracy will gradually decrease as the equipment parameters and control parameters change, and the root mean square error remains within 2.5%. However, when the parameters of the field equipment changes greatly, the applicability of the data model becomes worse and needs to be re-trained to achieve the original accuracy.