基于改进PSO优化RBF的页岩气TOC含量预测  

Prediction of Total Organic Carbon Content of Shale Gas Based on RBF Neural Network Optimized by Improved PSO Algorithm

作　　者：邵越 黄诚[1] SHAO Yue;HUANG Cheng(School of Computer Science,Southwest Petroleum University,Chengdu Sichuan 610500,China)

机构地区：[1]西南石油大学计算机科学学院,四川成都610500

出　　处：《计算机仿真》2025年第2期83-89,共7页Computer Simulation

基　　金：中石油-西南石油大学创新联合体项目(2020CX030200)。

摘　　要：针对页岩气井中“甜点”参数总有机碳(TOC)含量的预测,传统机器学习方法和许多神经网络模型由于计算代价大、时间成本高等问题,难以在勘探开发过程中进行应用。为保证预测精度和减少运算时间,提出一种改进粒子群(PSO)优化径向基(RBF)神经网络的页岩气TOC含量预测模型。模型通过PSO算法对RBF神经网络的隐含层中心点、场域宽度以及连接权值进行优化,并且在PSO算法中结合Tent混沌映射、差分进化算法以及自适应动态调整惯性权重和学习因子,提高粒子的寻优能力。经实验表明,使用上述模型预测的TOC曲线与实际的TOC曲线拟合程度高,测试集R2为97.4%,具有良好的预测效果。For the prediction of total organic carbon(TOC)content of the"sweet spot"parameter in shale gas Wells,traditional machine learning methods and many neural network models are difficult to apply in the process of exploration and development due to the high computational cost and time cost.In order to ensure prediction accuracy and reduce the operation time,an improved particle swarm optimization(PSO)radial basis(RBF)neural network was proposed to predict the TOC content of shale gas.The model optimizes the hidden layer center point,field width and connection weight of the RBF neural network by using the PSO algorithm,and combines Tent chaotic mapping,differential evolution algorithm and adaptive dynamic adjustment of inertia weight and learning factor in the PSO algorithm to improve the particle optimization ability.The experiment shows that the TOC curve predicted by this model fits well with the actual TOC curve,and the R2 of the test set is 97.4%,which has a good prediction effect.

关 键 词：页岩气 总有机碳 粒子群优化 径向基函数 

分 类 号：TE319[石油与天然气工程—油气田开发工程] TP391.9[自动化与计算机技术—计算机应用技术]