Research On Key Technology Of Leak Detection Of Natural Gas Pipeline Based On VMD

The laying of natural gas pipelines is getting longer and longer,and various pipeline safety hazards follow.Due to the overage operation of many pipelines,environmental corrosion and aging of pipelines,and man-made perforation theft of oil and gas,etc.,pipeline leakage often occurs.If the detection is falsely reported and the leakage is not found in time,it will not only cause a lot of waste of resources,environmental pollution,and economic loss.It will also endanger our personal safety and even cause disasters.Therefore,timely and accurate detection of pipeline leakage is one of the research hotspots.The preprocessing of pipeline working condition signal,feature extraction and pattern recognition of pipeline working condition are designed.Specific plans are as follows:There are a lot of noises in pipeline leakage signals,which will seriously affect the accuracy of pipeline leakage detection.Therefore,a new denoising method combining Variational Mode Decomposition(VMD)and improved Bhattacharyy Distance(BD)is proposed.Firstly,pipeline leakage signal is decomposed using VMD to obtain several Intrinsic Mode Functions(IMFs).Then,the similarity between each IMF and pipeline leakage signal is calculated using the improved BD,and the turning point was determined by the maximum slope of the BD curve to determine whether IMFS is effective.Compared with other denoising methods of EMD and VMD,the proposed method not only has higher denoising waveform reduction degree,but also has better denoising performance.The parameters of VMD are set by human experience.If the settings are not accurate,it will seriously affect the performance of VMD.Therefore,an improved VMD adaptive signal denoising method is proposed,which uses Salp Swarm Algorithm(SSA)to search for the best parameters of VMD.The permutation entropy is used to construct a new fitness function,that is,the ratio of the mean value to the variance of permutation entropy,which is used as the evaluation index of SSA.Finally,the best parameters obtained by SSA search are input into VMD.Compared with other methods,SSA-VMD improves the ability of decomposing the signal in terms of separating the useful signal and noise in the original signal.The feature extraction of pipeline signals will also affect the accuracy of natural gas pipeline leak detection.For this reason,a feature extraction method based on VMD and Locally Linear Embedding(LLE)is proposed.First,use VMD to decompose the signal to obtain a number of IMFs,and the characteristic modal components are selected by using the dispersion entropy.Then,the time-frequency domain features of the characteristic modal components are extracted to construct a high-dimensional feature matrix,and the low-dimensional feature vector is obtained through LLE dimensionality reduction.Finally,the support vector machine(SVM)model is established through the extracted feature vector,and the performance test is performed.The control experiment shows that the scheme has a high rate of condition recognition for pipeline signals,up to 95%.The selection of SVM parameters will directly affect the performance of the model.Therefore,this paper uses the GA algorithm to search for the best parameters of the SVM,namely the penalty parameter c and the kernel function parameter g.Control experiments show that the GA optimized SVM has a higher recognition rate than the general SVM,and improves the performance of the SVM model.The GA-SVM method has good applicability in pipeline leak detection,with a classification recognition rate of 98.33%,which verifies the feasibility and superiority of the GA-SVM method in pipeline leak detection.Effectively solve the problems of false positives and false negatives in pipeline leak detection.GA-SVM method has good applicability in pipeline leak detection,and the classification recognition rate reaches 98.33%,which is 3.33% higher than SVM and 1.66% higher than PSO-SVM.GA-SVM method can identify the running state of pipelines with high accuracy.