当前位置：首页 > 过刊浏览->2025年第51卷第4期
输气管道管存量差异性分析及预测技术
Difference analysis and prediction technology of gas pipeline stock
浏览(602) 下载(4)
- DOI:
- 10.3969/j.issn.1001-2206.2025.04.006
- 作者:
- 赵鹤翔, 孟子钦, 张永继, 王泽宇, 洪毛毛
ZHAO Hexiang, MENG Ziqin, ZHANG Yongji, WANG Zeyu, HONG Maomao
- 作者单位:
- 1. 河北华北石油天成实业集团有限公司, 河北任丘 062500;2. 北京军区某部队, 北京 100000;3. 中国石油华北油田公司安全监督检测中心, 河北任丘 062500
- 关键词:
- 稳态管存; 动态管存; 数据驱动; 偏差模式; 顺序模式
steady-state pipeline stock;dynamic pipeline stock;data-driven;deviation mode;sequential mode
- 摘要:
管存气量是构成天然气管道运行输差的一个因素，为提高输气管道管存量的计算效率和计算精度，基于有限容积法，提出了一种动态管存量计算方法，在入口压力正弦变化的条件下，将动态管存量与稳态管存量计算结果进行了对比，分析了不同结果之间的差异性，最后将流动机理与运行数据相结合，构建了基于数据驱动的CNN-LSTM管存量预测模型。结果显示：动态管存量的振幅小于稳态管存量，相位滞后于稳态管存量，正弦周期越短则正弦振幅越大，管存量的累积偏差和相对输差越大；动态管存量可体现气体充装的过程，具有一定的科学性和合理性；偏差模式管存量计算模型在预测集上的平均绝对误差、均方根误差和决定系数分别为1.29×104 m3、0.51×104 m3、0.990 6，其预测表现优于顺序模式。研究结果可为日常管存量的精细化管理提供理论依据。
Pipeline stock is a part that consists of natural gas pipeline transmission differences. To improve the calculation efficiency and accuracy of pipeline stock, a dynamic pipeline stock calculation method was proposed based on the finite volume method. Under the condition of inlet pressure changing with the sinusoidal function, a comparison was made between the dynamic and steady-state pipeline stock results, while the variances between the different results were analyzed. Finally, the flow mechanism was combined with the operating data to construct a data-driven CNN-LSTM pipeline stock prediction model. The results showed that the amplitude of the dynamic pipeline stock was smaller than that of the steady-state pipeline stock, and the phase lagged behind that of the steady-state pipeline stock. As the sinusoidal period is shorter and the sinusoidal amplitude is larger, the cumulative deviation and relative transmission difference of pipeline stock are greater. The dynamic pipeline stock, which could reflect the gas filling process, was scientific and reasonable. The mean absolute error, root mean square error, and determination coefficient of the pipeline stick calculation model in deviation mode were 1.29×104 m3, 0.51×104 m3, and 0.990 6, respectively, better than the sequential mode. The research results can provide a theoretical basis for fine management of daily pipeline stock.