Classification and identification method of power quality multi-disturbance signals based on ga-svm

Abstract

The invention provides a GA-SVM-based electric energy quality multi-disturbance signal classification and identification method. Firstly, feature extraction is performed on various kinds of electric energy quality complex disturbance by using an HHT method based on an improved EEMD, namely, according to the characteristics of the EEMD algorithm, adding positive and negative Gaussian white noise with the same absolute value to reduce noise residues. A parameter selection process of the EEMD algorithm is optimized by using the parameter self-adaption characteristic of probability statistics. A method of adaptive threshold de-noising is proposed to perform de-noising treatment on IMF signals and the influence of noise in each IMF component is reduced. Then, the optimization of SVM parameters is achieved via the GA. An SVM classifier is used as the classification tool for electric energy quality multi-disturbance signals. For avoiding the defect of empirical parameter selection of SVMs in the prior art, a SVM parameter selection process is optimized by using the global optimization characteristic of the GA, so that the time of setting parameters according to experiments and repeated tests is greatly solved and accuracy and practicality of parameters are improved.

Description

technical field [0001] The invention relates to the technical field of power quality analysis in power systems, in particular to a GA-SVM-based classification and identification method for multiple disturbance signals of power quality. Background technique [0002] At present, one of the important goals and directions of power grid development is to ensure the power quality of power supply and provide corresponding services to different power users according to their needs; The timely and accurate identification and classification of the disturbances in the power quality of the power grid has become a hot topic of research by scholars in recent years. [0003] Since there are many types of power quality disturbances, and the disturbance signal itself occurs randomly and suddenly, it is difficult to perform feature extraction and classification identification; and in the process of compound disturbance formation, the eigenvalues ​​of each single disturbance signal are superim...

AI Technical Summary

Problems solved by technology

[0004] At present, it is difficult to extract the features of power quality multi-disturbance signals, and the extracted feature quantities are prone to failure and modal aliasing, and the extracted feature values ​​are not accurate enough; in addition, the current methods for classifying power quality composite disturbances mainly include: , decision tree and support vector machine, etc.; the neural network classifier has a simple structure and strong solving ability, but the training time is long, and it is prone to problems such as over-learning; the decision tree classifier is to simulate human thinking to construct classification rules, Although the classification speed is very fast, it is more complicated to establish rules in the classification process, and there will be error accumulation errors, and it is difficult to deal with multi-category classification models; therefore, it is urgent to study a new classification and recognition method for power quality multi-disturbance signals to solve the problem. above question

Images