Photovoltaic fault arc feature selection method combining filtering type evaluation strategy and packaging type evaluation strategy

Abstract

The invention discloses a photovoltaic fault arc feature selection method combining a filtering type evaluation strategy and a packaging type evaluation strategy, and the method comprises the steps: carrying out the processing of a to-be-selected multi-feature set through employing a ReliefF algorithm, endowing different features with corresponding weights based on fault arc feature correlation, screening out the features with lower weights, and if the number of needed features is given, obtaining a required optimal feature set by adopting a maximum correlation and minimum redundancy algorithm, otherwise, obtaining a series of non-redundant feature sets by adopting the maximum correlation and minimum redundancy algorithm, and performing solving by adopting multi-objective optimization of classifier accuracy and the number of features to obtain the optimal feature set. The optimal data set meeting different requirements is constructed by combining multiple feature selection methods, the optimal features used for fault arc data analysis are fully mined, the dimensionality of the fault arc feature set is reduced, the classifier training time is shortened, and the rapidity and reliability of a fault arc detection algorithm are improved.

Description

technical field [0001] The invention belongs to the technical field of photovoltaic fault arc feature selection, and specifically relates to a method for the characteristics of photovoltaic system DC fault arc time domain and time-frequency domain, by constructing a multi-time-frequency feature set, fully mining the characteristics of fault data, and combining filtering and packaging A method for feature selection based on a formula evaluation strategy. Background technique [0002] In recent years, photovoltaic power generation has developed rapidly due to its clean and renewable nature. However, with the aging of equipment and external factors, the problem of DC arc faults in photovoltaic systems is becoming more and more serious. At present, large-scale photovoltaic power plants are mainly analyzed based on the time domain and frequency domain characteristics of DC fault arcs. In the time domain, the voltage and current waveforms on the line when the arc occurs can be u...

AI Technical Summary

Problems solved by technology

[0004] In the arc fault feature selection algorithm, the computational complexity is low and the speed is fast when the filter-type feature selection is used. However, since the filter method mainly considers the expression differences of features in different categories of samples, no classifier is involved in the process of feature selection. Therefore, when using the filtering method for feature selection, it is difficult to determine the optimal number of features of the feature set. The encapsulation feature selection process is related to the learner, and the performance of the learner is used as the evaluation criterion for feature selection, so the versatility is poor. , when using embedded feature selection, feature selection can be performed automatically during the learner training process

However, in the application process of the hybrid feature selection method, there are four problems in the application process: the classification accuracy needs to be improved, the data is high-dimensional, the candidate feature subset is single, and the proportion of relevance and redundancy is the same.

Images