基于GCN-GWO混合优化的地热水换热式掺水集输工艺研究
Research on geothermal water heat exchange-based water-blending gathering and transportation process based on GCN-GWO hybrid optimization
浏览(138) 下载(0)
- 引用格式:
-
李瑞钊,魏立新,邓海平,胡世丰,胡金梅,文宇豪.基于GCN-GWO混合优化的地热水换热式掺水集输工艺研究[J].天然气与石油,2026,44(3):142-150.doi:10.3969/j.issn.1006-5539.2026.03.018
Li Ruizhao, Wei Lixin, Deng Haiping, Hu Shifeng, Hu Jinmei, Wen Yuhao.Research on geothermal water heat exchange-based water-blending gathering and transportation process based on GCN-GWO hybrid optimization[J].Natural Gas and Oil,2026,44(3):142-150.doi:10.3969/j.issn.1006-5539.2026.03.018
- DOI:
- 10.3969/j.issn.1006-5539.2026.03.018
- 作者:
- 李瑞钊1 魏立新1 邓海平2 胡世丰1 胡金梅1 文宇豪1
Li Ruizhao1, Wei Lixin1, Deng Haiping2, Hu Shifeng1, Hu Jinmei1, Wen Yuhao1
- 作者单位:
- 1. 东北石油大学石油工程学院, 黑龙江 大庆 163318; 2. 大庆油田有限责任公司第七采油厂, 黑龙江 大庆 163517
1. College of Petroleum Engineering, Northeastern Petroleum University, Daqing, Heilongjiang, 163318, China; 2. No.7 Oil Production Plant of Daqing Oilfield Co., Ltd., Daqing, Heilongjiang, 163517, China
- 关键词:
- 地热;油气集输;图卷积神经网络;灰狼优化算法;参数优化
Geothermal energy; Oil and gas gathering and transportation; Graph Convolutional Network(GCN); Grey Wolf Optimizer(GWO); Parameter optimization
- 摘要:
- 针对转油站集输系统能耗高、运行效率低等问题,结合“双碳”背景用地热能替代传统加热炉的实际应用场景,以地热水换热式掺水集输工艺流程为研究对象,建立了以能耗最低为目标的运行参数优化模型。结合模型的结构特点,提出一种融合图卷积神经网络的改进灰狼优化算法。算法通过k近邻算法在灰狼种群个体间构建图结构,利用图卷积操作聚合邻居节点的位置信息生成候选更新方向,与原始灰狼算法引导方向进行加权融合,并引入随迭代次数线性衰减的动态权重因子,平衡收敛速度与全局寻优能力。采用罚函数法将约束优化问题转化为无约束问题进行求解。GS转油站集输系统应用表明,优化后日掺水量减少12.80%,日均耗电量降低9.57%,优化效果显著。研究结果为地热供热条件下油田掺水集输工艺的低碳化、精细化运行提供了可量化的优化模型与可行的求解方法。
In response to the high energy consumption and low operational efficiency of gathering and transportation systems at oil transfer stations, and against the backdrop of China's dual-carbon goals, this study explores the practical application of replacing traditional heating furnaces with geothermal energy. Taking the geothermal water heat exchange-based water-blending gathering and transportation process as the case study, an operational parameter optimization model is established with the objective of minimizing energy consumption. Based on the structural characteristics of the model, an improved Grey Wolf Optimizer(GWO) incorporating a Graph Convolutional Network(GCN) is proposed. In the algorithm, a graph structure is constructed among individual grey wolf population members using the k-nearest neighbors method. Graph convolution operations are then applied to aggregate the positional information of neighboring nodes, generating candidate update directions. These directions are fused with the original GWO guidance directions through weighted combination, and a dynamic weighting factor with linear decay over iterations is introduced to balance convergence speed and global search capability. A penalty function method is adopted to convert the constrained optimization problem into an unconstrained one for solving. Application to the GS oil transfer station demonstrates significant optimization results:the daily water-blending volume is reduced by 12.80% and the average daily power consumption decreases by 9.57%. The proposed approach provides a quantifiable optimization model and a viable solution method for achieving low-carbon and fine-grained operation of oilfield water-blending gathering and transportation processes under geothermal heating conditions.
