基于改进随机森林的海上油气生产设备运行数据清洗方法

Operation Data Cleaning Method of Offshore Oil and Gas Production Equipment Based on Improved Random Forest

常规的生产设备运行数据清洗方法多数采用多阶段递进识别的原理,对冗余、异常数据进行剔除与清洗,效果较差,无法显著降低运行数据信噪比。基于此,引入改进随机森林原理,针对海上油气生产设备,提出了一种全新的运行数据清洗方法。首先,构建生产设备运行数据采样模型,进行运行数据信息采样。在此基础上,基于改进随机森林原理,对运行数据中的异常数据进行多维度清洗。从实验结果可以看出,新的清洗方法应用后,运行数据信噪比大幅度减少,清洗效果优势显著。

Conventional production equipment operation data cleaning methods mostly use the principle of multi-stage progressive recognition to eliminate and clean redundant and abnormal data,which has poor effect and can not significantly reduce the signal-to-noise ratio of operation data.Based on this,a new operation data cleaning method for offshore oil and gas production equipment is proposed by introducing the improved random forest principle.First of all,build the production equipment operation data sampling model to sample the operation data information.On this basis,based on the improved random forest principle,the abnormal data in the operation data are multi-dimensionally cleaned.From the experimental results,it can be seen that after the application of the new cleaning method,the signal-to-noise ratio of the operation data is significantly reduced,and the cleaning effect has significant advantages.