Research Of Voltage Sag Identification On Electric Railway Based On S-transform And RBF Neural Network

With the rapid development of electrified railway towards intelligence, high efficiency and safety, an increasing number of power electronic devices have been applied to the power supply system in the electrified railway. It has a great influence on the power quality so the higher requirements for power quality of power supply system are put forward. As one of the ordinary and influential power quality problems, voltage sag gets more and more attention, due to its characteristics of great harm, large influence scope and high occurrence frequency. It is of great significance for ensuring the safe and efficient operation of railway to efficiently determine the disturbance sources. In this paper,based on S-transform and RBF neural network algorithm, applied to electrified railway to achieve the detection, classification and recognition of voltage sag sources.Firstly, introduces the definition, sag reasons, harm and characteristic-index of voltage sag. And the general detection methods of voltage sag characteristics are analyzed and summarized. For the different characteristics of voltage sag sources of power supply system in the electrified railway, this paper analyzes the characteristics of traction power supply system and 10 kV automatic blocking transmission lines respectively. By software MATLAB/Simulink, the model of power supply system in the electrified railway is conducted and six data samples for characteristic analysis of voltage sag sources are acquired. As the method of characteristic analysis and detection for voltage sag, compared with other time-frequency detection methods, S-transform has the advantages of variable window function, high resolution of time, amplitude and frequency and low noise sensitivity. The disturbance signal becomes a complex matrix via S transform, so the curves such as 3D contour representing signal characteristic, amplitude envelope, amplitude envelope of time, amplitude envelope of frequency etc. are obtained. By means of processing and constructing those curves, energy, entropy and the statistics features on every envelope are extracted. The extracted statistic features include ten statistics which distinguish different voltage sag sources, for example mean, standard deviation, skewness, root mean square etc. Secondly, by the comparison with BP neural network, RBF network model in neural network is adopted as the recognition method of voltage sag sources. The statistic data extracted from samples are imported for automatic learning and test. It provides a basis for long-term detection and recognition of disturbance signal in the electrified railway.Finally, based on the different types of sample data of sag disturbance from the simulation model of voltage sag of electrified railway, the validity of the algorithm is verified. With the combination of the algorithm and the prototype of power quality detection and analysis device for electrified railway power supply system, portable device for power quality detection and analysis in the electrified railway power supply system has been developed to detect and recognize the voltage sag source. And the field experiments are conducted by the use of prototype for traction substation and 10 kV automatic blocking transmission lines in Beijing Railway Administration.The experimental results indicate that themethod proposed in this paper can accurately detect amplitude of voltage sag and determine the disturbance time. Besides, the method can precisely recognize the type of voltage sag disturbance source and ensure the safe and stable operation of power supply system in the electrified railway.