Research On Pipeline Leakage Detection And Localization Technology Based On Data Drive

As one of the main transportation methods,pipeline transportation has the advantages of low transportation cost,good safety,and freedom from seasonal constraints.With the rapid development of my country’s economy,the demand for petroleum and related products is also rapidly increasing.With the mileage of pipeline laying is increasing and the structure of pipelines is becoming more complicated,pipeline construction has entered a new stage of development.As the length of the pipeline increases year by year and the service life of the old pipeline increases year by year,the probability of pipeline leakage will increase.Once the leakage is not detected in time,it will lead to waste of resources,loss of property,and environmental pollution.The efficiency,reliability and safety of long-distance pipelines are of vital importance to pipeline transportation systems.The main purpose of leakage detection and localization technology is to detect leakages in time and accurately locate them when the pipeline leaks.The data-driven pipeline leakage detection and localization technology proposed in this thesis is the embodiment of digitization and informatization in the field of pipeline leakage detection.It is an important mean to maintain the integrity of the pipeline.It can greatly improve the agility,robustness and accuracy of the detection system.The pipeline leakage detection and localization system based on the data driven uses sensors to monitor the changes of the internal parameters of the pipeline in real time.These internal parameters include pressure,flow,sound pressure,ultrasonic sound velocity and so on.Then the pipeline leakage occurs or not will be deduced by intelligent algorithm.From the view of reducing data calculation amount,speeding up algorithm and balancing unbalanced data,sparse kernel principal component analysis(SKPCA)and improved grey wolf optimization(IGWO)are proposed in this thesis.The small leakage aperture can also be identified accurately by applying IGWO to optimize the setting parameters of LSTSVM;The fault diagnosis observer is used to detect multi-point leakage,then the pipeline is considered to be broken or not by comparing the actual measured value with the estimated value which is got by the fault diagnosis observer,and the multi-point leakage is detected effectively;The expression between the position of leakage and ultrasonic sound velocity is established,which is based on the relationship between ultrasonic sound velocity and pipeline pressure.The method proposed in this thesis can better solve the problem of small leakage and multi-point leakage that are difficult to identify and locate.It can provide some theoretical and experimental support for pipeline leakage detection technology applied in engineering.Its main contents are as follows:(1)Aiming at the problem that the pipeline system is nonlinear,and the real-time data operation has high dimension,and the data is unbalanced due to that the normal operating time is much longer than the leakage occurrence time,a sparse kernel principal component analysis algorithm is proposed in this thesis.This algorithm brings the sparse principal component analysis(SPCA)algorithm into the kernel principal component analysis(KPCA)algorithm.It achieves non-linear dimensionality reduction between data,simplifies the structure of the kernel principal component,and increases the sparsity between the nuclear principal variables.When the sparse data enters IGWO-LSTSVM,it can effectively improve the anti-interference ability of IGWO-LSTSVM and speed up the calculation.(2)Aiming at the problem that small leakage is difficult to find and locate,the IGWO-LSTSVM algorithm is proposed,which uses IGWO to optimize the kernel parameter and penalty parameter of LSTSVM.It improves the recognition performance and operation speed of LSTSVM.The change of fluid parameters in pipeline caused by small leakage is small and disappears quickly.The speed of the algorithm will affect the detection rate of small leakage directly.Based on grey wolf optimization(GWO),IGWO is proposed in this thesis,which brings particle swarm optimization(PSO)and genetic algorithm(GA)into GWO.The previous traversal optimal position information of PSO is brought into IGWO to enhance the local ability;The selection,crossover and mutation of GA is brought into IGWO to enhance global ability.The optimization time of the IGWO is shortened obviously,and the IGWO-LSTSVM algorithm has a good identification rate for pipeline operating conditions,and the average identification rate is up to 98%.(3)Aiming at the problem that multi-point leakage is difficult to find and locate,a leakage detection method based on fault diagnosis observer is proposed.It expands the leakage coefficient into state space as a state variable,and makes the output estimation of the fault diagnosis observer is only affected by the leakage coefficient through decoupling technology.It can detect leakages by comparing the estimation value with the actual value of pressure and flow,then multi-point leakage detection and localization are realized.The system has high robustness.(4)Aiming at the problem that the noise in the signal can weaken the identification and localization accuracy of leakage,variational mode decomposition(VMD)and continuous wavelet transform(CWT)are used to eliminate noise.The decomposition layers and penalty parameters of VMD and the decomposition layers of CWT have a great influence on the effect of noise reduction.In order to get the optimal value of these parameters,IGWO is used to optimize them(write as IGWO-VMD,IGWO-CWT).At the same time,a localization formula is proposed,which forms the relationship between the position of leakage and the change of ultrasonic sound velocity.The leakage signal which is processed by IGWO-VMD and IGWO-CWT will eliminate most of the noise,so the identification and localization accuracy of leakage is promoted.The experiment results show that the localization error of the annular pipeline test device is less than6 m.The data driven pipeline leakage detection and localization technology utilizes the classification function of IGWO-LSTSVM and the recognition function of the fault diagnosis observer.It takes single-point leakage detection method based on IGWO-LSTSVM as the core,and multi-point leakage detection method based on fault diagnosis observer as the auxiliary.It is proved to be effective and accurate by 2800 m annular pipeline test device.IGWO-LSTSVM is used to identify pipeline operating condition,which includes non-leakage conditions,different leakage aperture conditions and adjusting valve conditions.The experiment results show that even the small leakage of 2mm aperture can be identified accurately,so it is effective to single-point leakage,furthermore it accurately classifies all the leakage aperture of 2mm,3mm,6mm.It also can accurately identify and locate three points leakage by the fault diagnosis observer,which are established on 400 m,800m and 1200 m.Its maximum estimation error of leakage position is 14.16 m.The experiment shows that the method has good performance of identification and localization for small leakage of single-point and large leakage of multi-point.It is feasible.