基于BP神经网络的低温提氦工艺优化

氦气是国家重要性战略物资之一,目前氦气的主要工业来源仍是从天然气中提取。为进一步优化低温提氦工艺,降低工艺能耗水平,对已有低温提氦工艺进行了改进,以一级提氦塔进料温度、压力、回流比、制冷剂高压、低压压力和制冷剂流量6个参数为变量,建立基于BP神经网络算法的综合能耗及提氦浓度预测模型,并对模型进行检验,并运用训练好的BP神经网络对改进工艺的综合能耗及粗氦浓度进行了预测。研究表明:BP模型训练效果较好,可用于综合能耗和粗氦体积分数的预测;通过训练误差分析,确定了模型隐藏层节点数为8时BP模型预测结果最优;利用确定好的BP神经网络预测出最优工艺生产参数,在满足粗氦体积分数不小于63.5%的基础上,综合能耗降低了18.08%。

Optimization of low temperature extraction process based on BP neural network

Helium is one of the national importance strategic materials. At present, the main industrial source of helium is still extracted from natural gas. In order to further optimize the low temperature helium extraction process and reduce its energy consumption, the existing low temperature helium extraction process was improved. Taking the feed temperature, pressure, reflux ratio and the high pressure, low pressure and flow rate of refrigerant in the first stage helium extraction column as variable parameters, a comprehensive energy consumption and helium concentration prediction model has been established based on BP neural network algorithm. The model is tested and the trained BP neural network is used to predict the comprehensive energy consumption and the concentration of crude helium of the improved process. The results show that BP model has good training effect and can be used for predicting the comprehensive energy consumption and the volume fraction of crude helium. Through training error analysis, it is determined that BP model has the best prediction results when the number of hidden layer nodes is 8. Using the determined BP neural network to predict the optimal process parameters, on the basis that the crude helium volume fraction is not less than 63.5%,the comprehensive energy consumption is reduced by 18.08%.