利用麻雀算法优化随机森林提高持液率预测精度

Using Sparrow Algorithm to Optimize Random Forest to Improve the Accuracy of Liquid Holdup Prediction and Identification

摘要: 为提高气液两相流持液率的预测精度，本文提出一种基于麻雀搜索算法(SSA)优化随机森林(RF)模型的持液率预测方法。通过整合1219组来自不同学者多源实验数据，构建了包含管径、气相折算速度、液相折算速度、压力、温度及管道倾角等多维特征的高质量持液率数据库。采用3σ法则进行异常值处理，确保模型训练数据的稳健性。利用SSA算法自动优化RF模型的关键超参数，避免传统调参的主观性和局限性，提升模型拟合能力与泛化性能。实验结果表明，SSA-RF模型在训练集与测试集上分别实现了R²为0.935和0.913，RPD值分别为3.81和3.18。研究成果可为复杂气液两相流工况下持液率的高精度预测及工程优化设计提供重要技术支持。

Abstract: To improve the prediction accuracy of the liquid holding rate of gas-liquid two-phase flow, this paper proposes a liquid holding rate prediction method based on the Sparrow Search algorithm (SSA) to optimize the random forest (RF) model. By integrating 1219 sets of multi-source experimental data from different scholars, a high-quality liquid holdup database was constructed, which includes multi-dimensional features such as pipe diameter, gas phase conversion rate, liquid phase conversion rate, pressure, temperature and pipe inclination Angle. The 3σ rule is adopted for outlier handling to ensure the robustness of the model training data. The SSA algorithm is utilized to automatically optimize the key hyperparameters of the RF model, avoiding the subjectivity and limitations of traditional parameter adjustment, and enhancing the model’s fitting ability and generalization performance. The experimental results show that the SSA-RF model achieves R² of 0.935 and 0.913 respectively on the training set and the test set, and the RPD values are 3.81 and 3.18 respectively. The research results can provide important technical support for the high-precision prediction of liquid holdup and engineering optimization design under complex gas-liquid two-phase flow conditions.