Deep wavelet twin network fault diagnosis method for modular multilevel converter

Abstract

The invention discloses a deep wavelet twin network fault diagnosis method for a modular multilevel converter. The method comprises the steps of firstly carrying out the fusion and enhancement of fault data in an MMC rectification mode and an inversion mode, enabling the data to be mapped to a low-dimensional space through the powerful feature extraction capability of a deep wavelet twin network, and carrying out the fault recognition based on the Euclidean distance, secondly, introducing incremental learning to carry out re-learning on the fault diagnosis model after MMC parameter drifting to update model parameters, and improving the generalization ability of the fault diagnosis model, and finally, building an MMC rectification and inversion simulation model, and verifying the effectiveness of the method. According to the method, the advantages of the twin network and incremental learning are combined, the fault can be rapidly diagnosed, the operation reliability of the converter is improved, and the fault diagnosis accuracy of the MMC under the condition of small sample and parameter drift is improved.

Description

technical field [0001] The invention relates to the field of power grids, in particular to a deep wavelet twin network fault diagnosis method for modular multilevel converters. Background technique [0002] Modular multilevel converter (MMC) is widely used in flexible DC transmission, new energy grid connection and electric power due to its excellent control performance, good output characteristics and flexible AC-DC interface. Electronic transformer and other fields. The modular structure of MMC makes it possible to bypass or replace the faulty sub-module through the strategy of redundant protection when the sub-module fails, thereby improving the reliability and safety of the system operation, and accurate fault diagnosis is for redundant protection premise. Therefore, it is of great significance to quickly and accurately diagnose the fault when the sub-module fails. [0003] The fault diagnosis process based on machine learning is usually divided into two parts: featur...

AI Technical Summary

Problems solved by technology

In addition, after the model is trained offline, the model parameters cannot be relearned according to the actual situation, which may lead to unsatisfactory fault diagnosis results after the circuit parameters change.

However, due to the complex topology of MMC, there are many types of open-circuit faults of power devices, it is difficult to collect a large amount of data in each fault situation, and MMC is inevitably affected by the environment during operation, resulting in drift of device parameters

Images