Proposed scheme for characterization of power quality disturbances

Abstract EN

A voluminous amount of disturbance waveforms are captured and recorded by power quality survey projects.

These disturbances need to be automatically classified and characterized to provide informative and useful results about the power quality condition of the system.

Intensive research is conducted to accomplish efficient automatic classification tools.

There is still a notable scarcity in apt techniques for characterization or quantification of disturbances.

In this paper, a scheme based on discrete wavelet transform and neural networks is proposed to characterize the recorded power quality disturbances.

A routine is presented to compute the disturbance duration.

A dedicated neural network is used to estimate the duration-magnitude product of the disturbance.

The design and structure of the neural estimator is addressed.

An alternative scheme for designing the estimator is also proposed and described.

The performance of the two methods is tested with many disturbances of 6 different types.

The results are compared to select the best estimators relevant to each disturbance type.