论文部分内容阅读
为了精确控制青海盐湖结晶器的在线工艺过程,研发基于BP神经网络的波美度测量模型。寻求可全局收敛的快速学习算法,以满足系统实时控制和良好的性能需要,通过分析结晶过程的物理化学反应,确定影响波美度预测的主要因素,根据现场提取的KCL、温度、离心机电流等历史数据对神经网络模型进行训练。通过仿真以及对实际值和神经网络预测值的均方差分析,表明该模型可以准确地预测和确定波美度,最终提高结晶器的控制精度。
In order to precisely control the on-line process of Qinghai Salt Lake crystallizer, Baume degree measurement model based on BP neural network was developed. In order to meet the need of real-time control and good performance of the system, a fast learning algorithm that can globally converge is proposed. By analyzing the physicochemical reaction of the crystallization process, the main factors affecting the prediction of Baume degree are determined. According to KCL, temperature, centrifuge current Such as historical data to train neural network model. Through simulation and analysis of mean square error between the actual value and the predicted value of neural network, it is shown that the model can accurately predict and determine the Baume degree, and finally improve the control accuracy of the mold.