基于随机森林回归算法的抽油机井系统效率分析与预测  

Analysis and prediction of system efficiency for pumping well based on random forest regression algorithm

作　　者：王薇[1] WANG Wei(No.3 Oil Production Plant of Daqing Oilfield Co.,Ltd.)

机构地区：[1]大庆油田有限责任公司第三采油厂

出　　处：《石油石化节能与计量》2024年第8期1-5,共5页Energy Conservation and Measurement in Petroleum & Petrochemical Industry

摘　　要：抽油机井系统效率过低,则无功损耗大,必然造成能耗的浪费,因此有必要对系统效率进行分析研究。首先根据系统效率计算公式进行分析,构建了12类属性指标数据集,用于有功功率的分析和回归;采用随机森林回归算法,对指标数据集进行训练集回归,并对测试集测试;最后,采用随机森林回归算法对现场的抽油机井系统效率进行了预测。对训练集2560口抽油机井进行回归,得出系统效率主要受日产液量影响,其次为有功功率,二者重要性占72.2%;全样本特征属性预测和缺样本特征属性预测的测试集的确定系数分别为0.852和0.701,说明在有功功率缺失时,拟合质量降低,但系统效率的变异中可由各属性指标参数解释部分的占比仍较大;根据缺样本特征属性预测回归模型,在现场对系统效率低于15%的188口井进行措施调整,累计节电15.28×10^(4)kWh,折合经济效益9.73万元。The system efficiency of pumping unit is too low,and the reactive power loss will be large,causing the waste of energy consumption.Hence it is necessary to analyze and study the system efficiency.Firstly,the calculation formula of system efficiency has been analyzed,and a dataset of twelve types of attribute index has been constructed for the analysis and regression of active power.Then the random forest regression algorithm is used to perform training set regression on the index data set and test the testing set.Finally,the system efficiency of pumping well based on random forest regression algorithm is applied and predicted in the field.Based on the regression of 2560 pumping wells in the training set,it is concluded that the system efficiency is mainly affected by daily fluid production,followed by active power,and the importance of both accounts for 72.2%.The coefficients of determination of the testing set for full sample feature attribute prediction and missing sample feature attribute prediction are 0.852 and 0.701 respectively,indicating that the fitting quality decreases when active power is missing.However,the proportion of the variation of system efficiency can still be explained by each attribute index parameters.According to the regression model of missing sample feature attribute prediction,the 188 wells with a system efficiency of less than 15%in the field are adjusted,the cumulative power saving is 15.28×104 kWh,saving 97300 yuan of economic benefits.

关 键 词：系统效率 随机森林 确定系数 回归模型 全样本特征属性 缺样本特征属性 

分 类 号：TE933.1[石油与天然气工程—石油机械设备] TP18[自动化与计算机技术—控制理论与控制工程]