Scholarlink

<p>A small sample datasets jet pump efficiency prediction model based on SHAP-LSSVR is proposed to address the issue of determining multiple structural parameters in the design of jet pumps used in drain sand of coal-bed methane wells. We use 30 sets of CFD numerical simulation results a-s samples for this experiment, and use the SHAP values of each sample as the weight of the penalty factor to improve the least squares support vector regression (LSSVR) algorithm. The average relative errors obtained by LSSVR and SHAP- LSSVR were 2% and 1.1%, respectively, with r-squared values of 0.74 and 0.92. The results show that SHAP-LSSVR can enable different samples to play different roles in the fitting process, and is suitable for optimizing the structure of jet pumps-with small sample set. Finally, we use the particle swarm optimization algorithm to search for the global optimum in the structural parameter space, and the optimal parameter combination was obtained. The error between the predicted results and CFD numerical simulation was 1.8%.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimization Method for the Structure of Drain Sand Jet Pump in Deep Coalbed Methane Wells

  • 期刊论文

  • 发表日期：

    2025年5月1日

查看全文
  • Yin Decai,
  • Wu Jianjun,
  • Liu Zhen,
  • Tian Zhida,
  • Luo Xiangjie

摘要

A small sample datasets jet pump efficiency prediction model based on SHAP-LSSVR is proposed to address the issue of determining multiple structural parameters in the design of jet pumps used in drain sand of coal-bed methane wells. We use 30 sets of CFD numerical simulation results a-s samples for this experiment, and use the SHAP values of each sample as the weight of the penalty factor to improve the least squares support vector regression (LSSVR) algorithm. The average relative errors obtained by LSSVR and SHAP- LSSVR were 2% and 1.1%, respectively, with r-squared values of 0.74 and 0.92. The results show that SHAP-LSSVR can enable different samples to play different roles in the fitting process, and is suitable for optimizing the structure of jet pumps-with small sample set. Finally, we use the particle swarm optimization algorithm to search for the global optimum in the structural parameter space, and the optimal parameter combination was obtained. The error between the predicted results and CFD numerical simulation was 1.8%.