基于样本生成策略的成品油管道混油浓度预测

Sample generation strategy-based method for predicting mixed oil concentration in multi-product pipeline

摘要

成品油管道顺序输送不可避免会产生混油,准确预测混油浓度是保障油品质量、降低处理能耗的重要措施;然而工程管道工况变化复杂,混油发展受多因素影响,现有模型挖掘变量关联信息有限,存在准确性和有效性较低的弊端。充分考虑多变量影响机制,提取混油浓度分布机制特征;构建基于自学习策略的虚拟样本生成模型,填补原始样本缺失信息;耦合原始与生成样本,充分表征混油浓度特征变量高维非线性关联信息,提高模型的准确性。结果表明:本方法性能优于现有混油浓度预测方法,均方误差降低了40%;与现有样本生成方法相比,本方法生成样本质量更高,浓度预测准确度分别提高了33%、23%和29%,可实现小样本下不同工况特性的混油浓度精确预测,兼具准确性和通用性,有助于提高站场混油监测与控制智能化水平。

Abstract

The sequential transportation of multi-product pipelines inevitably forms mixed oil segment. Accurate prediction of mixed oil concentration is an important measure to ensure oil quality and reduce processing energy consumption. However, the engineering pipeline conditions change complexly, and the development of oil mixing is affected by multiple factors. The variable association information mined by existing models is limited, resulting in lower accuracy and effectiveness. In this work, the multivariate influence mechanism was fully considered to extract the mechanism characteristics of mixed oil concentration distribution. Then, a self-learning-based virtual sample generation model was established to fill in the missing information of the original samples. The generated and original samples were coupled to fully characterize the high-dimensional nonlinear correlation information of the mixed oil concentration variables, and the model accuracy was improved. The results show that the performance of the proposed method is superior to the existing methods for predicting oil mixture concentration, with a mean square error reduction of 40%. Compared to existing sample generation methods, the proposed method generates samples of higher quality and improves prediction accuracy by 33%, 23% and 29%, respectively. The proposed model can achieve accurate prediction of mixed oil concentration under different operating conditions and characteristics under small samples, with desirable accuracy and versatility, and can help to improve the intelligent level of station mixed oil monitoring and control.

关键词

成品油管道 / 混油浓度预测 / 特征表征 / 自学习机制 / 虚拟样本生成

Key words

multi-product pipeline / mixed oil concentration prediction / feature characterization / self-learning / virtual sample generation

参考文献

[1] 郑坚钦, 杜渐, 梁永图. 成品油管道停输压力实时监控研究[J]. 石油科学通报, 2021, 6(4): 648-656.

[2] ZHENG Jianqin, DU Jian, LIANG Yongtu. Research into real-time monitoring of shutdown pressures in multi-product pipelines[J]. Petroleum Science Bulletin, 2021, 6(4): 648-656.

[3] YUAN Z, CHEN L, LIU G, et al. Knowledge-informed variational Bayesian Gaussian mixture regression model for predicting mixed oil length[J]. Energy, 2023, 285: 129248.

[4] 闫亚敏, 董烈武, 任泽, 等 . 通用成品油管道调度模型建立及求解[J]. 油气储运, 2020, 39(5): 549-557.

[5] YAN Yamin, DONG Liewu, REN Ze, et al. Model establishment and solution of general product pipeline scheduling[J]. Oil & Gas Storage and Transportation, 2020, 39(5): 549-557.

[6] DU J, ZHENG J, LIANG Y, et al. Deeppipe: a two-stage physics-informed neural network for predicting mixed oil concentration distribution[J]. Energy, 2023, 276: 127452.

[7] 郑坚钦. 机制与数据融合的成品油管道运行工况监测研究[D]. 中国石油大学(北京), 2023.

[8] ZHENG Jianqin. Research on monitoring the operating conditions of multi-product pipelines through mechanism and data fusion[D]. China University of Petroleum (Beijing), 2023.

[9] 杜渐, 郑坚钦, 夏玉恒, 等 . 耦合混油发展机制与数据修正的成品油管道混油浓度预测[J]. 油气储运, 2024, 43(7): 796-808.

[10] DU Jian, ZHENG Jianqin, XIA Yuheng, et al. Prediction of mixed oil concentration in product oil pipeline coupling oil mixing mechanism and data correction[J]. Oil & Gas Storage and Transportation, 2024, 43(7): 796-808.

[11] 徐斌, 白霄梦, 徐宁, 等 . 成品油管道混油接收及处理优化方法[J]. 油气与新能源, 2025, 37(2): 65-73, 81.

[12] XU Bin, BAI Xiaomeng, XU Ning, et al. Optimization method on mixed oil reception and treatment in multiproduct pipelines[J]. Petroleum and New Energy, 2025, 37(2): 65-73, 81.

[13] 胡煊, 陈世一. 管道顺序输送混油浓度一维模型的应用[J]. 油气储运, 2010, 29(12): 913-915,876.

[14] HU Xuan, CHEN Shiyi. Application of one-dimensional model of contaminated concentration in pipeline batch transportation[J]. Oil & Gas Storage and Transportation, 2010, 29(12): 913-915,876.

[15] 吕坦. 成品油管道顺序输送混油分析与模拟研究[D]. 西安: 西安石油大学, 2014.

[16] LÜ Tan. The analysis and simulation study about batching transportation of mix oil by products pipeline[D]. Xi̍an: Xi̍an Shiyou University, 2014.

[17] 夏增艳, 刘青泉. 顺序输送混油过程的二维数值分析[J]. 力学与实践, 2010, 32(6): 13-17, 48.

[18] XIA Zengyan, LIU Qingquan. Numerical simulation of the contamination between batches in multi-product pipeline transport[J]. Mechanics in Engineering, 2010, 32(6): 13-17,48.

[19] ZHENG J, DU J, LIANG Y, et al. Deeppipe: theory-guided LSTM method for monitoring pressure after multi-product pipeline shutdown[J]. Process Safety and Environmental Protection, 2021, 155: 518-531.

[20] 韩冬. 顺序输送管道中航煤质量指标变化规律研究[D]. 北京: 中国石油大学(北京), 2020.

[21] HAN Dong. Study on the quality evolution of jet fuel transported sequentially in the multi-product pipelines[D]. Beijing: China University of Petroleum (Beijing), 2020.

[22] 汤东东. 成品油管道混油浓度在线预测及基于质量指标的切割方法研究[D]. 北京: 中国石油大学(北京), 2019.

[23] TANG Dongdong. On-line prediction of mixed oil concentration in product pipelines and study on cutting method based on quality index[D]. Beijing: China University of Petroleum (Beijing), 2019.

[24] 赵云鹏. 基于数据分析的成品油管道混油特性研究[D]. 北京: 中国石油大学(北京), 2020.

[25] ZHAO Yunpeng. Study on the characteristics of oil mixing in multi-product pipelines based on data analysis[D]. Beijing: China University of Petroleum (Beijing), 2020.

[26] 杜渐, 郑坚钦, 赵伟, 等 . 成品油管网批次在线跟踪技术[J]. 油气与新能源, 2024, 36(4): 67-75.

[27] DU Jian, ZHENG Jianqin, ZHAO Wei, et al, Online batch tracking technology for multi-product pipeline network[J]. Petroleum and New Energy, 2024, 36(4): 67-75.

[28] 杜渐, 李昊翀, 廖绮, 等 . 耦合流动机制与运行数据的成品油管道瞬态仿真[J]. 油气储运, 2024, 43(10): 1157-1172.

[29] DU Jian, LI Haochong, LIAO Qi, et al. Transient simulation of multi-product pipeline driven by flow mechanisms and operational data[J]. Oil & Gas Storage and Transportation, 2024, 43(10): 1157-1172.

[30] 邢志晟, 韩国庆, 贾友亮, 等 . 机制与数据双驱动柱塞气举瞬态仿真[J]. 石油机械, 2025, 53(12): 119-128.

[31] XING Zhisheng, HAN Guoqing, JIA Youliang, et al. Mechanism and data dual-driven transient simulation of plunger gas lift[J]. China Petroleum Machinery, 2025, 53(12): 119-128.

[32] ZHENG J, DU J, LIANG Y, et al. Deeppipe: a semi-supervised learning for operating condition recognition of multi-product pipelines[J]. Process Safety and Environmental Protection, 2021, 150: 510-521.

[33] ZHENG J, WANG C, LIANG Y, et al. Deeppipe: a deep-learning method for anomaly detection of multi-product pipelines[J]. Energy, 2022, 259: 125025.

[34] 陈溪铭, 李坤, 付孟楷, 等 . 口径管道裂纹缺陷漏磁数据生成与评估方法[J]. 石油机械, 2025, 53(7): 1-11.

[35] CHEN Ximing, LI Kun, FU Mengkai, et al. A method for generating and evaluating magnetic flux leakage data for large-diameter pipeline crack defects based on numerical simulation and diffusion algorithm[J]. China Petroleum Machinery, 2025, 53(7): 1-11.

[36] 何国玺. 成品油管道输送过程混油特性及泄漏测算方法研究[D]. 北京: 中国石油大学(北京), 2018.

[37] HE Guoxi. Study on the oil mixing characteristics and leakage estimation method during batch transportation process in a multi-product pipeline[D]. Beijing: China University of Petroleum (Beijing), 2018.

[38] 王宝莹. 成品油管道混油浓度非对称性分布现象的特性研究[D]. 北京: 中国石油大学(北京), 2020.

[39] WANG Baoying. Characteristics research of asymmetry concentration curve in products pipeline[D]. Beijing: China University of Petroleum (Beijing), 2020.

[40] 卢克东, 徐良骥, 牛亚超. 基于GA-PSO融合算法的开采沉陷Richards预计模型参数优化[J]. 金属矿山, 2021, 536(2): 155-160.

[41] LU Kedong, XU Liangji, NIU Yachao. Parameter optimization on richards model of mining subsidence based on GA-PSO hybrid algorithm[J]. Metal Mine, 2021, 536(2): 155-160.

[42] 郑坚钦, 杜渐, 梁永图, 等 . 基于CAE-TSNE的成品油管道运行工况识别[J]. 石油科学通报, 2024, 9(1): 148-157.

[43] ZHENG Jianqin, DU Jian, LIANG Yongtu, et al. Identification of operating conditions of multi-product pipelines based on CAE-TSNE[J]. Petroleum Science Bulletin, 2024, 9(1): 148-157.

基金资助

国家自然科学基金项目(52202405,5234100047)

国家重点研发计划政府间重点专项(2024YFE0100800)