Electric energy quality mixed disturbance analysis method based on deep learning

Abstract

The invention discloses an electric energy quality mixed disturbance analysis method based on deep learning. The power quality mixed disturbance classification method can effectively classify the power quality mixed disturbance, and has higher robustness. The method comprises the steps of firstly adjusting the attenuation rate of a Gaussian window function to change the time resolution and the frequency resolution of S transformation so as to embody the characteristics of different types of electric quality signals better and improve the distinguishing strength of the different types of electric quality signals; then constructing a deep learning model based on the LSTM, and classifying the the electric quality signals, in a constructed deep learning model, preprocessing the data by adopting a multi-layer perceptron, carrying out preliminary analysis and feature extraction on the data, and then, carrying out semantic segmentation on the electric quality signal by utilizing an LSTM artificial neural network with stronger analysis capability on sequence data, and finally, performing supervised classification on the semantic segmentation of the LSTM by adopting a pooling layer and a multi-layer perceptron, thereby facilitating the improvement of the classification capability of the model.

Description

technical field [0001] The invention relates to the field of power electronics technology, in particular to a method for analyzing power quality mixed disturbances based on deep learning, which is applicable to the power quality classification method of multi-disturbance mixed signals and the construction of deep learning classification models based on LSTM networks. Background technique [0002] The rapid development of the industrial information society has also made the network structure of the distribution network complex, and the types and demands of users' electricity consumption have increased sharply. As a result, the efficiency and accuracy of state estimation of the distribution network and power quality management have also been higher. Requirements, the premise of distribution network state estimation is to accurately distinguish the power quality disturbance signal. [0003] The traditional power quality analysis method is mainly composed of two parts: feature e...

AI Technical Summary

Problems solved by technology

Both of these two feature classification methods have deficiencies. The artificial neural network is easy to fall into the local optimum and stop training, but cannot reach the global optimal solution. In the process of training the samples, the SVM method needs sufficient data and cannot Lack of certain specific data, such as signal incompleteness will lead to poor recognition rate. In addition, when solving multi-classification problems, the recognition effect of SVM is also very unsatisfactory.

In the previous literature, the combination of power quality analysis and restricted Boltzmann machine (RBM) in the field of deep learning has effectively solved the classification and identification of power quality compliance disturbance signals, but in processing multiple types of mixed disturbance signals, The sparse representation method first separates the mixed signal, and then processes the single disturbance signal separately, which has certain limitations for the classification of multi-type mixed disturbance signals.

Images