摘要
针对现有单井产液量预测方法的局限性,从地质、压裂、开发等3个方面考虑,选取影响单井产量的12种影响因素,采用相关向量机(RVM)对多因素影响下的单井产量数据进行回归,并利用增强鲸鱼算法(IWOA)对RVM模型进行优化,建立IWOA-RVM的单井产量预测模型。在低渗透油田进行了实例分析和模型对比,结果表明,IWOA-RVM模型可以处理各影响因素之间的非线性、非正态关系,与BP、IWOA-ELM和SVM模型相比,其RMSE最小,MAPE最小,训练时间最短。该模型在预测低渗透油田油井产量上具有一定的科学性和有效性。
In view of the limitations of the single-well fluid production prediction methods,12 factors affecting single-well production were selected from three aspects of geology,fracturing and development.The single-well production data was regressed,the enhanced whale algorithm(IWOA)was used to optimize the RVM model,and the IWOA-RVM single-well production prediction model was established.Finally,an example is analyzed and the model is compared.The results show that compared with the BP,IWOA-ELM and SVM models,the model has the smallest RMSE,the smallest MAPE,and the shortest training time.IWOA-RVM model can deal with the nonlinear and non-normal relationship among various influencing factors.The model is scientific and effective in predicting single-well production.
作者
薛钊
XUE Zhao(Development Department,Huabei Oilfield Company,CNPC,Renqiu,Hebei 062550,China)
出处
《世界石油工业》
2022年第2期69-74,共6页
World Petroleum Industry
关键词
鲸鱼算法(WOA)
相关向量机(RVM)
产量预测
机器学习
低渗透油田
whale optimization algorithm(WOA)
relevance vector machine(RVM)
production prediction
machine learning
permeable oilfield
