Multi-fault diagnosis modeling method of hybrid convolution neural network-driving lithium battery

Abstract

The invention relates to a multi-fault diagnosis modeling method of a hybrid convolution neural network-driving lithium battery, and belongs to the technical field of batteries. The method comprises the steps of acquiring actually-measured and sieved lithium battery fine fault spectrum by fractional order Fourier conversion to form a hybrid big data label sample for lithium batter diagnosis; designing a global convolution neural network for a hybrid lithium battery fault sample and a local convolution neural network for the actually-measured and sieved lithium battery fault data to form a hybrid convolution neural network lithium battery fault diagnosis model; and achieving multi-classification and positioning of lithium battery faults by learning global and local lithium battery fault characteristics in the convolution neural network and by full-connection classification mapping. By the method, the reliability and the safety of a battery management system are improved, the calculationcomplexity caused by various parameters is reduced, and the problem that only a certain lithium battery fault diagnosis task can be handled by a multi-physical coupling diagnosis model during actualapplication is solved.

Description

technical field [0001] The invention belongs to the technical field of batteries, and relates to a multi-type fault diagnosis modeling method for lithium batteries driven by a hybrid convolutional neural network. Background technique [0002] In recent years, lithium batteries, as green and clean secondary batteries, have been widely used in various electronic devices, such as automobiles, ships, airplanes and even some military electronic devices. How to effectively evaluate the reliability of lithium-ion batteries, in order to avoid the failure of lithium-ion batteries and cause serious consequences ranging from operational damage to performance degradation or even catastrophic failure, requires accurate diagnosis and analysis of multiple types of lithium-ion battery failures. The lithium battery fault diagnosis coupling model established by commonly used physical and chemical laws is difficult to reduce the computational complexity caused by the large number of parameters...

AI Technical Summary

Problems solved by technology

The lithium battery fault diagnosis coupling model established by commonly used physical and chemical laws is difficult to reduce the computational complexity caused by the large number of parameters, and the multi-physics coupling diagnosis model can only be used for a certain type of lithium battery fault diagnosis task in practical applications; in the face of multiple types of For fault diagnosis, the existing neural network method can learn the interrelated behavior of various faults from the measured fault data to a certain extent, but due to incomplete high-value fault data and insufficient fine extraction of various fault features , resulting in the deep neural network not playing its due simultaneous and spatial diagnostic energy efficiency in the field of lithium battery fault diagnosis

Images