Comprehensive energy metering working condition extraction method and system

Abstract

The invention discloses a comprehensive energy metering working condition extraction method. The method comprises the following steps: inputting a comprehensive energy metering working condition signal into a working condition one-dimensional convolutional network model to obtain a first feature vector of comprehensive energy metering; inputting the visualized comprehensive energy metering working condition signal into a working condition one-dimensional convolutional network model to obtain a second characteristic vector of comprehensive energy metering; and according to the first feature vector and the second feature vector, obtaining a working condition dimensionality reduction feature vector for comprehensive energy metering. The method has higher working condition classification accuracy. On one hand, the working condition feature extraction method can extract working condition signal features and working condition matrix features in a targeted manner, can optimize feature data through a classification algorithm, and has a stronger theoretical basis. And on the other hand, the method has very strong non-linear capability, can effectively extract features in various working condition scenes, is simple in operation steps and high in operation speed, and has a relatively good application prospect.

Description

technical field [0001] The invention belongs to the technical field of energy metering, and in particular relates to a method and system for extracting working conditions of comprehensive energy metering. Background technique [0002] With the access of new energy power generation systems and the expansion of power grid business, nonlinear loads and other loads in energy metering lead to more complex and variable metering conditions, and traditional feature extraction methods for working conditions are difficult to meet the needs of complex scenarios. . At present, the feature extraction of traditional energy metering conditions in my country mainly relies on wavelet transform, Fourier transform, S transform, Hilbert-Huang transform, singular value decomposition and other variant models. These traditional methods expand the way and method of working condition feature extraction through fitting or algorithm, but there are problems such as slow extraction speed and low accura...

AI Technical Summary

Problems solved by technology

However, due to the inevitable two-sidedness of research algorithms, research in these related fields generally only studies a certain problem in signal processing, feature selection and classification methods, and the consideration of feature engineering problems is not comprehensive enough.

For example, efficient classification methods generally use systematic models without explicit feature extraction. Traditional feature extraction methods do not link the feature extraction process with classification results, and are not effective in classification.

Images