Study On Digital Twin Method For Oil Pipeline Station Process Based On Hybrid Model

With the advancement of digital pipeline construction in China,the concept of digital twin has gradually entered the field of oil and gas industry.Integrating multi model state estimation methods and data processing methods,pipeline system process digitial twin is expected to mirror system operation parameters and dynamic operation behavior.As an intelligent node,it can provide a simulation platform for intelligent pipeline scheduling algorithm,which promotes the development of pipeline system from digitization to intelligence.Comparing with gas pipeline,the fast transient characteristics of liquid pipelines put forward higher requirements for the accuracy and efficiency of state estimation of process digital twin model.As the core control component of oil pipeline operation,the construction of oil pipeline station process digital twin is of great significance for the safe operation of system.This paper focuses on the difficulties in the construction of station process digital twin,which involves the coupling of data-driven method and model-driven method,nonideal boundary data processing,inverse problem of system dynamic control process and so on.In the construction of system state model,the unsteady state estimation models of external pipeline and station are derived by time-domain discretization and frequencydomain analysis respectively,based on the spatial scale difference of station pipeline and long-distance pipeline,the characteristics of components and instrument monitoring.And the solution methods of these flow models are improved with the consideration of noise interference.The particle filter method is adpted to the time-domain discretize model to conduct on-line unsteady flow parameter estimation of external pipeline.The generalized predictive control theory is innovatively introduced to solve the station system state estimation model based on frequency domain analysis method,which realized the on-line estimation of unsteady parameters of station system with anti colored process noise interference characteristics.By this way,the problem of system computational efficiency caused by the difference of spatial scale between the station and the external pipeline is solved.In terms of SCADA data processing and application,many methods including lifting wavelet analysis,fast Fourier transform,outlier detection,mechanism and data model fusion are used to deal with the problems of data noise idenfication,missing data and abnormal data.Using the redundant data of the oil pipeline station system,the system state model is modified by variety of data drive modeling methods and nonlinear optimization methods.According to the characteristics of station data sampling,an online state model parameters calibration method has been proposed by using multiobjective optimization method.Moreover,the on-line construction of data driven model for external pipeline transient flow process is realized by comprehensive application of NARX open-loop model and closed-loop model.Using the interactive multi model fusion method,a hybrid on-line state estimation method with parameter adaptive calibration is established for external pipeline system.For the state estimation model of external pipeline system,combined with the characteristics of station data,the on-line correction method of model parameters based on multi-objective optimization method and the construction method of on-line data drive state model based on NARX model are proposed,forming the mechanism of on-line adaptive correction and data-driven state estimation method.For the station system state model,the parameters of the multi condition model are modified by introducing Bayesian estimation and particle swarm optimization,based on the characteristics of quasi steady state and dynamic regulation conditions.Coupling adaptive control theory and Kalman state estimation model,a station state estimation model with on-line correction of model deviation is established.Combining this method with the state estimation method based on generalized predictive control,the state estimation and model adaptive correction of unsteady process parameters under multi rate observation are realized.For the oil pipeline station,Bayesian estimation and particle swarm optimization methods are introducted for off-line state model parameter modification methods of quasi steady state and dynamic regulation conditions respectively.The on-line correction of model deviation is realized by coupling adaptive control theory and Kalman state esetimation method.Combining this method with the state estimation method based on generalized predictive control,the state parameter estimation and model adaptive correction are realized under multirate observation unsteady process.In the aspect of dynamic control and operation behavior inverse problem,this paper focuses on the parameter tuning of the existing controller in the station,the application method of modern control theory and solution method of system operation process based on actual data.Combining with linearized state estimation model and nonlinear optimization algorithm,a parameter tuning method of PID controller for oil pipeline station with initial parameter selection is proposed.In addition,we explore the feasibility of the model free adaptive controller method to replace the traditional PID method.And the statblity of system and controller method has been proved.In this way,the parameter control of a single component is extended to the overall output operation parameter control of the station,which considers the overall unsteady hydraulic characteristics of the pipeline system.Based on the historical dynamic operation data and system state model,the full format dynamic linearization model free adaptive control method is innovatively introduced to solve the inverse problem of dynamic operation,which realizes the twining of dispatchers’ dymaic operation behavior of single component control and multi-station componets joint control process.