Compared with the traditional AC distribution system,the flexible DC distribution network has the advantages of large transmission capacity,low line loss and high power supply reliability,which is conducive to the access of distributed new energy,energy storage devices and flexible DC load.It is the development direction of intelligent distribution system in the future.However,when the flexible DC distribution line fails,the short-circuit current rises rapidly and the amplitude is high,It will cause great damage to power electronic equipment in DC system.Therefore,the research on flexible direct distribution line protection scheme with reliable performance is of great significance to maintain the safe and stable operation of DC distribution system.The traditional flexible direct distribution protection method is easily disturbed by fault distance,transition resistance,line distributed capacitance current and noise,and has some problems,such as difficult threshold setting and complex calculation.To solve the above problems,this paper first analyzes the basic structure,wiring mode,operation characteristics and control strategy of key equipment of flexible direct medium voltage distribution system;Then the system simulation model is built in PSCAD/EMTDC;Finally,a single terminal protection scheme with high protection reliability and strong anti-interference performance for DC distribution lines and a pilot backup protection scheme with high selectivity and sensitivity are proposed.The main research contents are as follows:(1)Research on Single-ended Protection Scheme for Flexible DC Distribution Line Based on Gate Recurrent Unit NetworkFor the problems of the difficulty in setting the thresholds,the operation failures caused by high impedance faults and the insensitivity under noise interference in the protection of the flexible DC distribution lines,a protection scheme for the flexible DC distribution lines based on the GRU deep learning model is proposed in this paper.Firstly,the amplitude-frequency characteristics of the short circuit faults based on the current-limiting reactors are analyzed,and the fault characteristic quantities under the frequency domain features are extracted through the generalized S-transform,which are taken as the input data of the gate recurrent unit(GRU)network.Then,the GRU deep learning model is constructed and trained to extract the deep features of the fault samples under time series.Combined with the support vector machine(SVM)classifier,the fast fault identification and isolation are realized.Finally,the simulation model is built in the PSCAD/EMTDC to verify the feasibility of the protection scheme.The simulation results show that the scheme has strong anti-interference ability and high sensitivity,which meets the requirements of the reliability and quickness in DC line protection.(2)Research on Pilot Protection Scheme for Flexible DC Distribution Line Based on Currentlimiting Reactance VoltageFor the problems of the high resistance ground faults caused operation failure,susceptible to the influence of line distributed capacitance current and noise interference in flexible DC distribution line protection,a pilot protection scheme based on current-limiting reactance voltage is proposed in this paper.Firstly,the voltage polarity characteristics of current-limiting reactance on both sides of the line are analyzed,and the fault identification criterion is constructed by using Kendall correlation coefficient.Then,the ratio characteristics of low-frequency transient energy sum of two pole reactance voltage are extracted by variational mode decomposition(VMD)algorithm for fault pole selection.Finally,validation and analysis of simulation in PSCAD/EMTDC.The simulation results show that the scheme has strong ability to withstand transition resistance and noise interference,is not affected by the line distributed capacitance current,can act quickly and accurately under different fault conditions,has high sensitivity and reliability,and has certain applicability to the protection of four-terminal DC grid. |