摘要
对高后果区的管道实施应力监测是保证其安全运行的重要措施。针对目前应力预测模型精度低的问题,采用自回归滑动平均(ARIMA)模型与长短时记忆(LSTM)模型对管道等效应力的发展趋势进行预测,并采用自适应人工鱼群算法(IAFSA)对LSTM模型的关键参数进行求解。结果表明,ARIMA模型对非线性时间序列的适应性不好;等效应力与轴向应变、周向应变、测试点温度、压力等因素有关,LSTM模型的预测结果中存在部分离群点,且出现了放弃和平移现象;组合模型的均方根误差最小,决定系数最大,说明组合模型可用于长期应力监测数据的趋势分析。研究结果可为管道完整性管理和评价提供技术支撑。
Stress monitoring is an important measure to ensure the safe operation of pipelines in high-consequence areas. Considering the low accuracy of current stress prediction models, this paper uses the autoregressive integrated moving average(ARIMA) model and the long short-term memory(LSTM) model to predict the development trend of pipeline equivalent stress, and the key parameters of the LSTM model are solved by the improved artificial fish swarm algorithm(IAFSA). The results reveal that the ARIMA model has poor adaptability to nonlinear time series. The equivalent stress is related to factors such as the axial strain, circumferential strain, as well as temperature and pressure at the test point. Some outliers exist in the prediction results of the LSTM model, and the phenomenon of abandonment and translation appears. The combined model has the smallest root mean square error and the largest coefficient of determination, which indicates that the combined model can be used for trend analysis of long-term stress monitoring data. The research results can provide technical support for pipeline integrity management and evaluation.
作者
刘翔
LIU Xiang(PetroChina Huabei Oilfield Company,Renqiu 062550,China)
出处
《石油工程建设》
2022年第5期38-43,共6页
Petroleum Engineering Construction
