Scheduled System Maintenance:
Some services will be unavailable Sunday, March 29th through Monday, March 30th. We apologize for the inconvenience.
By Topic

Multiclass support vector machines for power system disturbances classification based on wide-area frequency measurements

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

The purchase and pricing options are temporarily unavailable. Please try again later.
2 Author(s)
Gang Zheng ; ECE Dept., Tennessee Tech Univ., Cookeville, TN, USA ; Craven, R.

The intelligent, robust and fast multi-class classification of power system disturbances is very important to improve control algorithms for ensuring power system security and reliability, an essential function for smart grid infrastructure. Moreover, in a future power system mostly consisting of distributed generators and renewable energy resources on which the disturbance has more impact, the analysis of disturbances by classifying and categorizing real-time frequency data is rather critical. Fortunately, wide area frequency data from a nation-wide frequency monitoring network (FNET) provides a means by which disturbances can be detected. However, so far none of strategies reported to date has good performance at classifying the disturbances although many of them are used currently in on-line analysis. The complex and irregular pattern characteristics of each kind of disturbance are the main reason. Artificial intelligence methods could be one of the solutions, but the large number of input values and an insufficient number of training examples has slowed the reduction of artificial intelligence methods to practice. Therefore, a mathematical model of common disturbances is proposed to generate a training database for artificial intelligence method and feature extraction by computing the wavelet coefficients, parameterizing the results and computer generating the data. This paper uses a multi-class support vector machine model to be trained on the extracted features to discern the otherwise hard-to-classify disturbances pattern and upon testing, yields good performance.

Published in:

Southeastcon, 2011 Proceedings of IEEE

Date of Conference:

17-20 March 2011