| The new generation of urban rail flexible traction power supply system adopts fully controlled voltage source converter(VSC)to completely replace diode uncontrolled rectifier unit.Due to better adjustment ability,energy-saving effect and simplified structure,it has become the key development direction under the carbon peaking and carbon neutrality strategy.The operation of urban rail flexible traction power supply system involves the process of high-frequency switching and low-frequency power flow,and the power distribution and waveform distortion are usually described by power flow(second level)and harmonic(millisecond level)at different time observation scales,which are important characteristics to reflect the steady-state operation of the system.At present,model based design is the main means of comprehensive analysis,control and optimization of system power flow and harmonic.However,the traditional modeling methods have inherent defects such as time scale separation,off-line update and ideal input.There are still deficiencies in the consistency,fidelity,calculation efficiency and accuracy of the model.Thus based on the digital twin technology,the digital twin architecture,multi-time scale steady-state modeling and real-time simulation method,power flow model autonomous fidelity evolution method and harmonic source model consistency optimization method were studied in this thesis,so as to realize the accurate deduction and intelligent control of multi-time scale state,and ensure the best operation performance of urban rail flexible traction power supply system.In the aspect of digital twin architecture,a digital twin multi-level architecture of urban rail flexible traction power supply system was proposed,and the core technical schemes at all levels including physical perception,data interaction,model simulation and functional application were designed,which establishes a complete digital twin closed-loop system with virtual and real two-way mapping.Focusing on the functional requirements of system steady-state operation control,the digital twin function application methods aiming at power flow regulation optimization,power flow anomaly detection and harmonic impact assessment were proposed,and the relationship between function and various models based on digital thread was established,which realizes the unified digital chain management;the data acquisition,communication and processing methods for digital twin were proposed to ensure the accurate perception and analysis of power flow and harmonic state;a complete modeling system of digital twin generalized form was proposed,which provides a top-level modeling framework for digital twin of urban rail traction power supply system.In terms of multi-time scale modeling and simulation,aiming at the separation of power flow and harmonic simulation in traditional modeling methods,a multi-time scale modeling and real-time simulation method of urban rail flexible traction power supply system was proposed,which ensures the real-time synchronous simulation of power flow and harmonic,and realizes the multi-time scale operation deduction.Firstly,based on the mechanism modeling method,the power flow model and harmonic model were established,and the multi-time scale operation characteristics of urban rail flexible traction power supply system were revealed.Then,a modeling method of external characteristics of harmonic source based on multi-layer perceptron(MLP)was proposed,which improves the harmonic calculation speed and solves the simulation efficiency contradiction between traditional power flow and harmonic model.Finally,according to the harmonic coupling and multi-time scale state correlation,a mult-time scale simulation method based on collaborative interaction was proposed.The influence law of control parameters on flexible traction power supply system was revealed,and the power flow regulation ability and energy-saving improvement effect were verified.For the fidelity evolution of power flow model,in order to solve the problem that the traditional power flow model is difficult to adapt to the changes of the actual system,an autonomous fidelity evolution method based on cluster intelligence was proposed,which ensures the consistency between the model parameters and the actual system and improves the accuracy of power flow calculation.It mainly includes: the autonomous evolution framework and strategy were designed,which ensures the accurate load input and online power flow perception in the evolution process of power flow model;A parameter hybrid correction algorithm based on particle swarm optimization was proposed,which effectively improves the convergence and optimization ability of parameter correction of power flow model,and realizes the high fidelity and high consistency modeling of power flow.In the aspect of harmonic source model consistency optimization,aiming at the characteristic difference between harmonic simulation data and measured data,a harmonic source model consistency optimization method based on transfer learning was proposed to solve the consistency deviation between the model and real equipment in harmonic characteristics.It mainly includes: a consistency optimization strategy based on transfer learning was proposed,which adopts online transfer learning with closed-loop feedback of measured data to continuously optimize the consistency of the model;A finetune transfer learning adaptive algorithm based on DDC and Adam was designed to reduce the tranfer learning loss caused by the difference of data characteristics between the source domain and the target domain,and ensure the accuracy and efficiency of harmonic source model consistency optimization. |
PDF Full Text Request