Research On The Identification Of Over-voltage And The Decomposition Of Mixed Over-voltage

In recent years, the smart grid technologies is developmented rapidly, a lot of countries and regions in the world have taken the construction of smart grid into their planning. China also has been upgrading the constricution of smart grid as the national strategy. State grid corporation of china proposes to build the“Strong and smart grid”based on the UHV power grid, which has the features of information, automation, and interactive. One of the main features of smart grid is the self-healing, which can continuously carry out on-line self-evaluation to predict the problems the grid may meet, discover existing faults, and immediately take control to ensure the reliability, safety and efficiency of the grid. The over-voltage caused by switching, lightning or system faults affects the security and stability of the power system. And it will become more dangerous to the system equipment with the increasing of voltage level. In order to improve the saftiy and staibility of power system, the research of over-voltage online monitoring and identification is ver important and has great practical significance for the protection of the safe operation of power grid. At present, a lot of over-voltage online monitoring equipment and waveform recorder have been put into field operation. However, accurate and effective over-voltage identification methold is still a problem need to be solved.In this paper, a whole complete over-voltage identification system is built, which has ability to identify and classify the metallic grounding, sub-frequency ferroresonance, power frequency over-voltage, high frequency ferroresonance, switching line, switching capators, lightning, arc grounding over-voltage. Meanwhile, aiming at the problem of power frequency over-voltage and mixed over-voltage identification, two recognition subsystems are built out.The main single-type identification system adopts the S-transform as the signals time-frequency analysis and feature extration algorithm, identify 8 kinds of over-voltages. And in this system, 3 types of power frequency over-voltage: fundmental ferroresonance, single phase-to-ground and wire breakage, which are highly similar in the voltage signal's features, are combined into one type to be dealed with. Based on the main features analysis of each kind of over-voltage, 6 different characteristic quantities are proposed.Based on the feature extraction of large mount of field over-voltage signals, the value distribution of each kind of characteristic quantity are obtained by statistical analysis. A complete initial identification system is built by employing the fuzzy expert system and support vector machine, which can meet the requirement of practical applications.Aiming at the identification problem of three kinds of power frequency over-voltages, fundmental ferroresonance, single phase-to-ground and wire breakage, an identicication system is built by compareing their major features in voltage and current signals, which is obtained through deducing their voltage and current analytical expressions. Meanwhiling, a new identification criterion based on zero sequence current is presented in this paper for the difficulty of classification of single phase-to-ground and fundamental ferroresonance, which is also called‘false grounding'.Through large amount of field over-voltage signals, this paper researches the decomposition algorithm and identification method of mixed over-voltage. In order to recougnize each kind of over-voltage in the mixed signals, based on the analysis of posbale type of mixed over-voltage, a decomposition algorithm based on actomix decomposition and damped sinusoidal atom dictionary is proposed.In order to reduce the computational complexity of the atomic decomposition and improve the accuracy of the decomposition, the FFT, GA and PSO are employed in the application.Aiming at the possible incorrect decomposition coursed by simply using of atomic decomposition, an improved time support searching method based on STFT and Hilbert transform and a double-atoms decomposition algorithm are proposed. Based on these methods, a well designed mixed over-voltage identification system is proposed based on fractal, and the acutrace of this system is vrified by field mixed over-voltage signals.Based on the iffective identification and classification of each kind of over-voltage through severl algorithms, a single type over-voltage identification system, power frequency over-voltage identification system and a mixed over-voltage decomposition and identification system are composed into a complete practicality identification system.