Study On Fault Classification And Fault Location Methods Based On Transient Signals For High Voltage Transmission Lines

With the continuous improvement of comprehensive national strength and the rapid development of power industry, China's electric power system has entered a time of high-voltage, great gird and large generators. However, with the unceasing increase of transmission capacity and voltage level, high-voltage transmission line fault will also bring more serious harm and impact to electric power system and stable operation, national economic construction and people's daily life. Therefore, the study of fast and accurate fault classification and fault location methods for transmission lines, can not only shorten the power-off time and reduce the strength of maintenance staff for power operation, but also have a major practical significance in ensuring the security and economical efficiency of power system. Although the transient duration of fault on transmission line is short, it contains a wealth of fault information, which makes it possible to achieve a fast and accurate classification and location. Based on this, in this thesis, in order to build a rapid fault diagnosis system for high-voltage transmission lines, fault classification and fault location methods based on transient are studied.In fault classification, the thesis first presents three different fault classification methods based on artificial intelligence algorithms for high-voltage transmission line. One is based on rough membership neural network (RMNN), which takes13different time domain and time-frequency domain characteristics of fault transient current signals as feature values for fault classification, using the classifier for fault classification based on ten different RMNN to classify and identify the ten common short circuit faults for high-voltage transmission lines. The second one is based on adaptive-network-based fuzzy inference system (ANFIS), which takes both time domain standard deviation and interquartile range of transient current fault components as the feature values for fault classification, and the fault classifier is based on two different ANFIS, which are designed according to the characteristics of the feature values. The third one is based on negative selection algorithm, which takes high-frequency transient energy of transient current signals as the feature values, and uses improved negative selection algorithm as the transmission line fault classifier. The simulation data from PSCAD/EMTDC is used to verify and test the proposed three kinds of fault classification methods for high-voltage transmission line, indicating that: these three methods of fault classification can classify and identify different fault of transmission line fast, accurately and reliably. Moreover, their classification effects are unaffected by fault resistance, fault distance, fault inception angle, and they also have a good adaptability to noise interference.In addition, the thesis proposes a method for transmission line fault classification based on time-frequency characteristics of the fault transient signal. The method considers time-frequency correlation coefficient and characteristics of time-frequency energy of fault transient current signals to define time-frequency characteristic correlation coefficient, through which different fault can be characterized. Moreover, the classification criterion is designed based on time-frequency characteristic correlation coefficient directly, which avoids structural problems of training samples contained by fault classification methods based on artificial intelligence algorithms. The proposed method is verified and tested by large numbers of simulation data. The simulation results show that this method can classify the different fault of transmission line fast, accurately and reliably. In addition, fault resistance, fault distance and fault inception angle cannot affect the classification effects.In the aspect of fault location, this thesis proposes two kinds of single-end fault location methods based on the time-frequency characteristics of transient traveling wave to improve the accuracy of current fault location methods using single-end traveling wave and traveling wave natural frequency. One of them is time-domain location method that considers frequency domain characteristics of transient traveling wave. It uses single-end traveling wave method as a springboard. Then on the basis of considering the influence of frequency characteristics of transient current traveling wave on the wave propagation velocity and the arrival of the first wave-front, Lipschitz exponent of transient signal is used to link the time domain and frequency domain characteristics of transient wave-front. Thus a more precise and reasonable method is achieved. Another is a frequency-domain fault location method, which considers the time-domain characteristics of transient traveling wave. The starting point of this method is traveling wave natural frequency. A new method of extracting natural frequency is proposed, which compensates for the natural frequency of transient traveling wave by time-domain period characteristic of signal to result in more accurate results of fault location. A large number of PSCAD/EMTDC simulation results show that the location accuracy of the two methods raised in this thesis have a significantly improvement over existing traveling wave method and traveling wave natural frequency method. Moreover, their reliability and accuracy will not be affected by fault resistance, fault distance, fault inception angle or fault type. It also has a good adaptability to noise interference.Finally, the frequency-domain location method considering transient traveling wave time-domain characteristics is extended to the grid fault location. Based on the mathematical relationship between the natural frequency of transient traveling wave signal with its propagation path and boundary conditions, a grid fault location method based on propagation path of transient traveling wave is proposed. First, the method determines the fault line according to natural frequency or the distribution of the transient traveling wave signal accurately, and then the natural frequency that reflects traveling wave propagation path including fault point is used to compute the fault distance accurately. Large numbers of simulation show that, in the situation of certain power network topology, the method can not only figure out fault line, but also calculate the distance accurately, and the location accuracy is not affected by fault resistance, fault inception angle, fault type and fault distance.Overall, this thesis eventually forms a rapid fault diagnosis system for high-voltage transmission line from fault classification to fault location.This thesis is supported by National Natural Science Foundation of China:'Information theory based power network fault diagnosis of multi-sourced signal'(No.50877068,2009-2011) and Doctoral Fund of Ministry of Education of China:'A new transmission line fault location method based on the extracted natural frequency of single-ended traveling wave'(No.200806130004,2009-2011).