Research On Pipeline Leakage Technology Based On QPSO-FCM Algorithm

In order to solve the disadvantages of low accuracy,high cost and great influence of terrain factors in the leakage detection of long-haul oil and gas pipelines,this paper proposes to use machine learning algorithm to build a model for sake of replacing the traditional leakage detection method.By means of the analysis and prcessing of the relevant data during the operation of long-haul oil and gas pipeline,a deep neural network model(fuzzy C-means)based on particle swarm optimization(PSO-FCM)is proposed,so as to apply to the detection of pipeline operation information and determine whether the pipeline has leakage accident.This study not only reduces the cost of detection and the influence of terrain factors on leakage detection,but also improves the accuracy of leakage detection.The main contents of this paper are as follows:First,several machine learning algorithms are comprehensively analyzed and compared,and the fuzzy C-means algorithm is the most suitable algorithm for the requirements of this paper which is selected to build the model.In allusion to the disadvantage of low speed,high request and non-convergence under some particular cases for using gradient descent method in training process,the particle swarm optimization algorithm is introduced to replace thegraient descent algorithm of the traditional fuzzy C-means algorith in order to improve the capacity,so that the fuzzy C-means algorithm can better achieve the purpose of pipeline leakage detection.Second,by applying the inertia linerly decreasing algorithm and the quantizer to the troditional particle swarm optimization algorithm,a mdified particle swarm optimization algorithm is proposed in this paper,which improves the global searching ability as well as insures the original convergency compared with the troditional particle swarm optimization algorithm.Several classical benchmark functions are used for compring the proposed particle swarm optimization algorithm with the troditional particle swarm optimization algorithm in various eignvalues.And the experiment result demonstrates that the proposed particle swarm optimization algorithm has a better global searching capacity and higher speed rather than the troditional particle swarm optimization algorithm.Finally,the PSO-based fuzzy C-means algorithm is applied to long-haul oil and gas pipeline leakage detection.The proposed PSO-based fuzzy C-means algorithm is compared with troditional fuzzy C-means algorithm and the widly used 3-layer BP neural network in processing and analyzing the data of simulated pipeline leakage experiment.And the experiment result proves that the proposed PSO-based fuzzy C-means algorithm has a higher accuracy,a stronger stability and a better speed compared with another two algorithms,which means that the method proposed in this paper is worthy of testing in the real long-haul pipeline.