Research On Power Quality Transient Disturbances Identification Based On Time-Frequency Analysis

With the improvement of power levels and the expanding of State grid scale,Powersystem is increasingly becoming complex. Due to the development of power control and transform technologies, numerous disturbances are poured in power system by primary equipment, secondary equipment and power consumers, which arouses different kinds of transient disturbance phenomenon, including voltage sag, swell, transient harmonic, simultaneous oscillation, etc.. Tinpot power quality has a bad influence on industrial manufacture and people’s daily life, which usually results in huge economic losses. Hence, it is especially essential to monitor and harness the power quality, but its prerequisite is to identify the categories of power quality disturbances(PQDs).Thus, this paper aims at the research of power quality disturbance identification.Analyzed the causes of PQDs, and established the PQDs’ mathematic model based on IEEE, IEC and GB. Then generated 6 kinds of PQDs data in MATLAB.Utilized the S transform(ST) and TT transform, and two tuning factors were proposed to improve the performance of ST and the signal power concentration, which improved the timefrequency resolution. Correspondingly, the time resolution of low frequency and frequency resolution of high frequency were both improved.TT transform is the reverse transform of ST. The signal energy would be concentrated near the dialog by adjust the tuning factors.At the same time, the time-time(t-t) resolution was improved, and the t-t span turned to be more detailed, which made TT transform perform better time localization capacity. Then the improved ST and TT transform were applied toPQDs feature extraction for 6 kinds of PQDs.Devised PQDs’ feature library which contained 15 categories PQDs’ feature information. It was hard to separate the crossed signal feature information. A new method was proposed by pair combining the PQDs’ feature information to screen out the best combination. Thus, the dimensions of PQDs’ feature were decreased to ease the burden of data storage, and improve the real-time identification of PQDs.Support Vector Machines(SVMs) were applied to identify the 6 kinds of transient PQDs. Designed the topology structure of SVMs and established the SVMs model. In accordance with magnitude-phase feature of PQDs,600 training instances were sampled to train the established classification model. And then 300 testing instances were fed in the trained model to test the correct rate of classification. The result demonstrates that the classifier’s good reliability, which can identify the transient PQDs.