Neural network-based lithium battery ultrasonic information characteristic extraction method

Abstract

The invention belongs to the field of battery related technologies, and specifically discloses a neural network-based lithium battery ultrasonic information characteristic extraction method. The method comprises the following steps of constructing two neural networks, taking ultrasonic information as input vectors and respectively taking health states and charge states as output vectors; acquiringultrasonic information as a training sample and training the two neural networks to obtain two corresponding neural network models; and inputting the acquired ultrasonic information of a lithium battery into the two neural network models so as to obtain a health state and a charge state of the lithium battery. According to the method, the two neural networks are constructed and then trained to obtain the two neural network models, so that the health state and the charge state of the lithium battery can be obtained at the same time by utilizing the ultrasonic information, thereby overcoming the bottleneck that the users cannot judge which states cause the changes of the ultrasonic signal amplitudes.

Description

technical field [0001] The invention belongs to the battery-related technical field, and more specifically, relates to a method for extracting ultrasonic information characteristics of a lithium battery based on a neural network. Background technique [0002] The state of charge of the battery, also known as the remaining power, represents the ratio of the remaining capacity of the battery to the capacity of the fully charged state after it has been used for a period of time or left unused for a long time, referred to as SOC; the state of health of the battery is the comparison between the battery and its ideal state A quality factor, referred to as SOH. Generally, the health status of the battery decreases with the increase of the use time and frequency, so it is usually necessary to set a threshold for the battery health status. When the battery health status is lower than this threshold, it means that the battery is not suitable for continued use. It is an effective meas...

AI Technical Summary

Problems solved by technology

[0004] Among them, the accuracy of the method of extracting spectrum peaks as features is not high. Experimental results show that it is difficult to directly establish a one-to-one linear relationship between peaks and SOC, which will cause relatively large errors during data fitting, and only peak features can be extracted, not It reflects all the information contained in the ultrasonic signal; the method of extracting features through spectrum integration is to integrate the peaks in the spectrum to the frequency, and the obtained value is used as the feature value extracted from the ultrasonic signal. After experimental verification, this method can be compared with SOC Establish a linear relationship with high precision, but the defect is that only one eigenvalue can be extracted, which is difficult to reflect all the information contained in the ultrasonic signal; the feature extraction through the time domain envelope is equivalent to a data compression method, similar to filtering , the envelope data of the time-domain waveform is used as the eigenvalue, but the accuracy of the eigenvalue extracted by this method is not high. After a large number of experiments, it has been proved that the envelope information cannot reflect the complete information of the ultrasonic signal waveform, and cannot distinguish the ultrasonic signal. Whether the change in amplitude is caused by SOC or SOH

Images