Power battery screening method based on typical feature vectors

Abstract

The invention discloses a power battery screening method for typical feature vectors. The method comprises the following steps: acquiring charging and discharging voltage signals of a plurality of power batteries with known states, and extracting voltage signal characteristics to form characteristic vectors; estimating a probability density function of each feature according to the values of the same type of samples in each feature variable, and generating a plurality of new samples by using a sampling method; aggregating the new samples by using a clustering method to form a plurality of representative typical feature vectors; and calculating cosine similarity between the feature vector of the to-be-detected power battery and each typical feature vector of each type, and identifying the health state of the to-be-detected power battery. By means of the screening method, the problem of overfitting of a traditional classification model under the condition of small samples and deviation caused by data imbalance are avoided, the practicability of power battery screening is improved, meanwhile, the typical characteristic conditions of the power batteries under different faults can be known more clearly, and a foundation is laid for better defining the states of the power batteries.

Description

Technical field: [0001] The invention relates to the technical field of energy storage batteries, in particular to a power battery screening method based on typical feature vectors. Background technique: [0002] With the rapid development of my country's electric vehicle industry, the power battery used to provide energy has become the core key component of the electric vehicle industry. The production of power battery cells, group technology, monitoring and analysis, operation and maintenance, and recycling. The research has received widespread attention. As the mileage of electric vehicles increases, the actual capacity of the power battery will continue to decline. Generally, when the actual capacity drops to about 80% of the rated value, a new power battery needs to be replaced. If the decommissioned power battery is directly recycled as raw materials, it will cause a great waste of resources. Therefore, the decommissioned battery is used for other non-strict service ap...

AI Technical Summary

Problems solved by technology

Among them, the experimental analysis method is the most accurate, but the time and labor costs are too high; while the model estimation method requires a more accurate modeling process and estimation model, resulting in a weak robustness of the method and difficulty in calculating many difficult-to-evaluate physical parameters; The data-driven method mainly relies on historical data to mine the characteristics of power batteries in different operating states, and group the operating states of the power batteries or return the estimated health value. This type of method is highly automated, simple and easy to implement, and does not require too much prior knowledge and Accurate mathematical models, but this type of method has high data requirements, that is, the accuracy of the model method depends entirely on the typicality and comprehensiveness of the data, and at the same time, mining methods and modeling methods also have a certain impact

Images