Research On Soft Measurement Method Of Oil-well Fluid Production Based On Dynamometer Card

As an important production energy,oil plays an important role in people’s life and economic development.Oil well production volume is an important indicator in oil well production management.For the rod pumping system used in most oil fields,the measurement records are not timely and costly using traditional mechanical measurement methods,while the mechanistic model of fluid production volume measurement based on dynamometer card has problems such as low accuracy and weak generalization.This thesis analyzes and studys the shortcomings of downhole working condition identification and schematic oil production mechanism model with the production process of rod pumping system as the research background,and proposes the hybrid prediction model of soft measurement of oil well production based on dynamometer card and a dynamic update strategy of the model which is updated instantly according to the working condition change,and its main research contents are as follows.(1)The hybrid model of oil well fluid production prediction based on the dynamometer card is proposed.The hybrid model is based on the effective stroke method of oil measurement with the support vector machine algorithm,and is supplemented by the compensation model of fluid production.The compensation model reduces the impact of the calculation error of the mechanistic model on the accuracy of the fluid production prediction model,and improves the generalization and accuracy of the oil well geophysical fluid production measurement model.(2)The traditional working condition identification algorithm is susceptible to factors such as unbalanced data sets when identifying the working conditions of a rod pumping system,leading to partial misjudgment of the working conditions and resulting in inaccurate working conditions when the soft measurement model of fluid production is subsequently updated according to the working condition changes.This thesis proposes the random forest downhole working condition identification model with data imbalance processing.The gray matrix feature extraction algorithm is used to extract features from the pump dynamometer card,and the extracted gray features are upsampled by synthetic minority oversampling technique to equalize the unbalanced data,and then the random forest parameters are optimally selected by the dragonfly optimization algorithm to recognize the pumping well conditions.(3)To address the problem that the soft measurement model of fluid production fails with the change of oil field working conditions,resulting in inaccurate and poorly generalized fluid production prediction results.The method is proposed to update the modeling data according to the just-in-time learning algorithm when the working conditions change,and the compensation model is updated online using similar sample numbers.It improves the accuracy of modeling sample selection of compensation model,strengthens the updating ability of compensation model to changes in working conditions,and avoids the failure of fluid production mixing model.By establishing the hybrid prediction model for soft measurement of oil well production based on dynamometer card and the dynamic update module for the model that updates instantly according to the change of working conditions,the overall model of soft measurement of oil well production based on dynamometer card is obtained.