A Power Quality Disturbance Identification Method Using Time-Domain Compression and Multi-resolution Fast S-Transform Feature Extraction

Abstract

The present invention provides a power quality disturbance recognition method using time-domain compression, multi-resolution fast S-transform feature extraction, which is applied to the analysis and research of power quality disturbance signals, and is characterized in that it includes simulation of power quality disturbance signals, and multiple disturbance signals Resolution fast S-transform processing, on the premise of determining the characteristics, the information required for the calculation of the windowed Fourier transform inverse transformation results of each main frequency point in the fast S-transform process is retained, and the intermediate matrix is ​​constructed, which is effective for the extraction of the intermediate matrix Features build feature vectors for classifier construction. It has the advantages of being scientific and reasonable, strong applicability, good effect, and being able to complete identification of complex power quality disturbance signals.

Description

technical field [0001] The invention is a power quality disturbance identification method using Time-domain Compression Multiresolution Fast S-transform (TCMFST) feature extraction, which is applied to the analysis and research of power quality disturbance signals. Background technique [0002] With the concept of sustainable development and low carbon, the scale of new energy power generation continues to expand, the grid connection of new energy power generation and the application of a large number of power electronic equipment will affect the power quality of the power grid (Power Quality, PQ), in science and technology and industry With the rapid development of production, the attention to the quality of power supply is increasing. Therefore, it is necessary to carry out in-depth monitoring and analysis of power quality, and the massive power quality data collected by a large number of monitoring points puts forward higher requirements for the power quality signal classi...

AI Technical Summary

Problems solved by technology

However, in practical applications, when the signal sampling rate is too high and the signal is too long, the scale of the modulus-time-frequency matrix obtained by the fast S transform is still large, the space complexity is high, the data storage space is large, and the hardware requirements are high.

In terms of classifiers, traditional support vector machines, BP neural networks, and decision trees have strong robustness, but cannot meet the needs of efficient and accurate classification, and extreme learning machines, as an optimization algorithm for single hidden layer neural networks, overcome this shortcoming

However, since the input weights and hidden layer nodes of the extreme learning machine are randomly selected, there must be a series of non-optimal or unnecessary values. On the one hand, it will increase the response time of the extreme learning machine algorithm to the test data. It will also increase the complexity of the network structure

Images