Leakage Fault Diagnosis Of Centralized Heating Pipe Network Based On Change Point Theory And Pipe Network Model

With the increase of urbanization rate in China,the scale of heating pipe network is growing rapidly and the condition of pipe network is becoming more and more complicated.Leakage,as the main operational fault of the heat supply network,arises from many causes.And the network is buried deeper,if the time and location of the leak can’t be determined quickly,it will cause large economic losses and affect people’s normal life.In this thesis,based on the summarizing and analyzing the research results of leakage diagnosis of heating pipe network,we propose a leakage fault diagnosis method of centralized heating pipe network based on the classical method of variable point theory--Cumulative Sum（CUSUM）method and leakage simulation model,in order to diagnose small leaks in heating pipe network,quickly issue leakage alarm signals,and accurately diagnose the location of leaks.Firstly,based on Graph Theory,Kirchhoff laws and pipeline characteristic equations,etc.,this thesis obtains the simulation models of normal condition and two leakage conditions of heating pipe network,gives the leakage simulation modeling method and process systematically,and proposes the leakage fault diagnosis method of heating pipe network combined with CUSUM method,and gives the different selection methods of threshold h and drift parameter k in this method.The main contents of the leakage fault diagnosis method are: calculate the cumulative sum of the pipe network make-up water flow,when the cumulative sum exceeds the threshold value,a leakage alarm signal is issued;reverse the location of the variable point of occurrence,calculate the leakage flow;according to the actual pressure and simulation pressure changes before and after the leakage,determine the specific location of the leakage.Secondly,based on the experimental pipe network for heating,a simulation model of node leakage and pipe leakage of single heat source branch pipe network is established by MATLAB software,and in response to five different leakage conditions: J1 and J2 for nodal leakage,G1,G2 and G3 for segment leakage.G1,G2,and G3,are simulated and experimentally investigated respectively.The results show: the maximum absolute value of relative errors of user’s flow rates are 4.61%,4.74%,4.92%,4.80% and 4.90% in the five conditions,while the maximum absolute value of relative errors between user’s simulated and experimental pressure are 4.71%,3.71%,3.81%,4.56% and 3.93%,which verifies the validity and accuracy of the simulation model.Then,based on the heating experimental network,the experimental scheme was designed according to the complexity of the leakage conditions,and the suitable threshold h=5σ and drift parameter k=1.015 were obtained by combining the experimental data;after analysis,the Recursive Least Square（RLS）algorithm was selected as the experimental data pre-processing method;five different pressure collection point distribution schemes were designed for leak location diagnosis,and the best diagnosis effect was found when the pressure collection points were set at the supply and return branch nodes of users 1,10 and 18 after comparison.The results show: the leakage alarm time obtained from the diagnosis of the experimental pipe network is 2.2s after the start of the leakage,and no false alarm,missed alarm and delayed alarm occurred;the maximum relative error between the diagnosed maximum leakage flow and the experimental value is 1.60%,and the accuracy of the leakage location diagnosis reaches 73.1%.It proves the usability of the leakage fault diagnosis method.Finally,using the actual heating pipe network structure data,the actual operating pressure and flow rate data of each user,a simulation model of the normal condition of this pipe network and the leakage condition of nodes and pipe sections is established,the leakage diagnosis method is applied to the actual pipe network,and the leakage condition data analysis of the pipe network is carried out.The results show: that the absolute value of the relative error between the simulated pressure and the actual pressure of the normal condition of the pipe network is3.91%;the leakage diagnosis result shows that the alarm signal is issued 61 s after the leakage occurs,and the relative error between the maximum leakage flow and the data analysis result is1.10%,and the leakage location diagnosis result is also close to the data analysis result.It shows that the leakage fault diagnosis method proposed in this thesis has strong practicality.