The Study On Dynamic Operation Characteristics Models Of Natural Gas Pipeline Network

Currently, the majority of the mathematical models used in the optimization of thenatural gas pipeline network operation are obtained by linearization of the steady-stateequations or by solving partial differential equations with difference method. However, suchmodels can not keep track of the time-varying impact of pipeline network parameters and thesolving speed can not meet the requirements of real-time simulation. Furthermore, therealization of optimal operation method is in need of the control theory, which can not beapplied without the dynamic operating characteristics model. Therefore, the accurate dynamiccharacteristics models have important significance in the optimization operation of the naturalgas pipeline network.In order to obtain the mathematical models of the dynamic operating characteristics ofthe natural gas pipeline network, several kinds of methods for modeling was analyzed and amethod of obtaining the model by applying the system identification techniques was proposed.First, a natural gas pipeline was built using the pipeline simulation software known asTGNET, the input signal was designed and the dynamic simulations were done. Second, theparameters of the model were estimated by using the least square method. FPE, AIC andseveral other methods were combined to determine the order and pure delay of the models.Finally, dynamic operating characteristics of single-stage pipeline, pipeline network withbranches as well as centrifugal compressor were modeled in this thesis.Input and output signals should be sampled before the identification of a continuoussystem. To solve the problem of how to determine the sample interval, this thesis proposed amethod for searching the optimal sample interval based on the existing method which uses thedistance between the original and the characteristic value of the identified model in the w-plane, combined with the binary search. The optimal sample interval which minimizes theoutput error of the identified model can be obtained by using this method without knowingthe characteristic value of the actual system, and the simulation result verified the efficiencyof the novel method.The application of the models was also studied in this thesis. Firstly, the operationdispatch of a simulative natural gas pipeline was analyzed, and the scheduling scheme wasgiven. Secondly, the optimal control of natural gas pipeline using Generalized PredictiveControl method was studied, and the scheduling scheme was implemented. Lastly, a methodof pipeline roughness detection using the step response curve of dynamic models wasproposed, and the simulation result verified the efficiency of this method.