CN-121254089-B - Energy storage battery life prediction method, device and energy storage system

CN121254089BCN 121254089 BCN121254089 BCN 121254089BCN-121254089-B

Abstract

The application relates to the technical field of battery life prediction and discloses an energy storage battery life prediction method, an energy storage battery life prediction device and an energy storage system. The energy storage battery life prediction method comprises the steps of determining a target battery to be predicted and an initial prediction model pre-trained by a training total set, obtaining a distance matrix calculated by a capacity attenuation curve, wherein the capacity attenuation curve comprises the capacity attenuation curve of the training total set battery and the capacity attenuation curve of the target battery, clustering according to the distance matrix to obtain a target battery device containing the target battery, forming a training subset by using other capacity attenuation curves except the target battery in the target battery device, adjusting the initial prediction model by using the training subset to obtain a target prediction model, and predicting the battery life of the target battery by using the target prediction model. And the accuracy of the target battery life prediction is improved through the target prediction model adjusted by the training subset after screening and optimizing.

Inventors

CAI XIN
YANG FENGWEI
CAO BAOJIAN

Assignees

浙江晶科储能有限公司

Dates

Publication Date: 20260512
Application Date: 20251203

Claims (11)

1. A method for predicting the life of an energy storage battery, comprising: determining a target battery to be predicted and an initial prediction model pre-trained by using a training total set, wherein the training total set comprises attenuation characteristic data of a plurality of batteries determined by a preset aging test or a capacity attenuation curve obtained by calculation according to the attenuation characteristics of the batteries; Obtaining a distance matrix calculated by a capacity attenuation curve, wherein the capacity attenuation curve comprises a capacity attenuation curve of a training total battery and a capacity attenuation curve of a target battery, the training total battery comprises n capacity attenuation curves, the capacity attenuation curve of the target battery is recorded as an n+1th capacity attenuation curve, the n+1th capacity attenuation curves are combined in pairs, euclidean distances of the capacity attenuation curves of all the two pairs of combinations are calculated respectively, and the distance matrix of the n+1th capacity attenuation curves is obtained; Clustering to obtain a target battery device containing the target battery according to the distance matrix, wherein other capacity attenuation curves except the target battery in the target battery device jointly form a training subset; adjusting the initial prediction model by utilizing the training subset to obtain a target prediction model; predicting the service life of the target battery by using the target prediction model; Said adjusting said initial predictive model with said training subset comprising: Determining timing characteristics of the training subset; And adjusting the weight parameters and the bias parameters of the full-connection layer of the initial prediction model according to the time sequence characteristics.
2. The method according to claim 1, wherein clustering the target battery devices including the target battery according to the distance matrix comprises: determining the similarity between the capacity fading curve of the target battery and the capacity fading curve of each of the remaining batteries based on the euclidean distance between the capacity fading curve of the target battery and the capacity fading curve of each of the remaining batteries; and clustering according to the similarity to obtain the target battery device.
3. The method of claim 2, wherein the euclidean distance of two of the capacity fade curves is determined by: detecting the capacity of a battery corresponding to a plurality of preset periodic points in the periodic battery aging test process, wherein the capacity attenuation curve consists of the capacities corresponding to the preset periodic points; Taking the sum of squares of the differences of the same preset periodic points of the two capacity fading curves as the Euclidean distance of the two capacity fading curves.
4. The method for predicting the life of an energy storage battery according to claim 2, wherein the distance matrix is specifically: ; Wherein i and j represent the numbers of the batteries in the training set, 0<i is less than or equal to (n+1), 0<j is less than or equal to (n+1), N is an integer greater than 1, D represents a number distance matrix, D ij represents the Euclidean distance between the capacity fading curve of the battery with number i and the capacity fading curve of the battery with number j, x i (k) represents the capacity of the capacity fading curve of the battery with number i in the kth period, and x j (k) represents the capacity of the capacity fading curve of the battery with number j in the kth period, 0<k is less than or equal to N.
5. The method of claim 1, further comprising, prior to said obtaining a distance matrix calculated from a capacity fade curve: filtering the capacity attenuation curves of the target battery and the training total set battery; the distance matrix obtained by calculating the capacity fading curve is: and obtaining a distance matrix calculated by the capacity fading curve after filtering.
6. The method of claim 5, wherein filtering the capacity fade curves of the target battery and the training total battery comprises: And carrying out filtering treatment on the capacity fading curve according to a moving average filtering mode with a preset window size.
7. The energy storage battery life prediction method according to any one of claims 1 to 6, wherein the initial prediction model is trained by: determining attenuation characteristic data of a plurality of batteries according to a preset battery aging test, wherein the attenuation characteristic data comprises battery attenuation characteristics corresponding to a plurality of preset periodic points in the periodic battery aging test process; and training the initial prediction model by taking the attenuation characteristic data as a training total set.
8. The method of claim 7, wherein the battery decay characteristic comprises any one of a capacity retention rate, a capacity decay rate, an internal resistance increase rate, a charge-discharge peak position, a charge-discharge amplitude, a voltage plateau change rate, and a temperature change trend, and combinations thereof.
9. The method of claim 7, wherein determining attenuation characteristic data of a plurality of batteries according to a predetermined battery aging test comprises: Constructing an accelerated aging experimental matrix by a plurality of controlled conditions according to an orthogonal principle or a full factor principle, wherein the controlled conditions comprise any one or combination of ambient temperature, charge-discharge multiplying power, charge state range and circulation depth; And determining the attenuation characteristic data of a plurality of batteries by using the accelerated aging test matrix.
10. The device for predicting the service life of the energy storage battery is characterized by comprising a determining module, an acquiring module, a clustering module, an adjusting module and a predicting module; The determining module is used for determining a target battery to be predicted and an initial prediction model pre-trained by utilizing a training total set, wherein the training total set comprises attenuation characteristic data of a plurality of batteries determined by a preset aging test or a capacity attenuation curve obtained by calculation according to the attenuation characteristics of the batteries; The acquisition module is used for acquiring a distance matrix calculated by a capacity attenuation curve, wherein the capacity attenuation curve comprises a capacity attenuation curve of a training total set battery and a capacity attenuation curve of a target battery, the training total set comprises n capacity attenuation curves, the capacity attenuation curve of the target battery is recorded as an n+1th capacity attenuation curve, the n+1th capacity attenuation curves are combined in pairs, euclidean distances of the capacity attenuation curves of all the two pairs of combinations are calculated respectively, and the distance matrix of the n+1th capacity attenuation curves is obtained; the clustering module is used for clustering to obtain a target battery device containing the target battery according to the distance matrix, wherein the capacity attenuation curves of the target battery device except the target battery jointly form a training subset; the adjustment module is used for adjusting the initial prediction model by utilizing the training subset to obtain a target prediction model; Said adjusting said initial predictive model with said training subset comprising: Determining timing characteristics of the training subset; According to the time sequence characteristics, adjusting weight parameters and bias parameters of a full-connection layer of the initial prediction model; The prediction module is used for predicting the service life of the target battery by utilizing the target prediction model.
11. An energy storage system, comprising: The energy storage battery life prediction apparatus according to claim 10, or the life prediction apparatus of an energy storage battery that performs the energy storage battery life prediction method according to any one of claims 1 to 9.

Description

Energy storage battery life prediction method, device and energy storage system Technical Field The present application relates to the field of battery life prediction technologies, and in particular, to a method and an apparatus for predicting the life of an energy storage battery, and an energy storage system. Background The service life of the energy storage battery is closely related to the battery performance of the energy storage battery, and if the service life exceeds the service life of the battery, electricity utilization can be possibly influenced, or damage to electric equipment is caused, so that a safety problem is seriously brought. Therefore, the life prediction of the energy storage battery is beneficial to improving the stability and reliability of electricity utilization. However, the aging mechanism of the energy storage battery is complex, and the working condition of the battery changes obviously along with the time, so that the current prediction mode cannot accurately reflect the degradation rule of the battery. If the prediction accuracy needs to be improved, a large amount of historical data and frequent data updating are relied on, and under the condition that the battery types are complex and various, the required data are more multiplied, and the training and updating difficulty of the prediction model is higher. Disclosure of Invention The embodiment of the application aims to provide a method, a device and an energy storage system for predicting the service life of an energy storage battery, which are used for screening a training subset which is more in line with the attenuation rule of a target battery through clustering calculation of raw data in a training total set, optimizing an initial prediction model through the training subset to obtain a target prediction model, and further improving the accuracy of the target prediction model on the service life prediction of the target battery. In order to solve the technical problems, the embodiment of the application provides an energy storage battery life prediction method, which comprises the steps of determining a target battery to be predicted and an initial prediction model pre-trained by a training total set, obtaining a distance matrix obtained by calculation of a capacity attenuation curve, wherein the capacity attenuation curve comprises the capacity attenuation curve of the training total set battery and the capacity attenuation curve of the target battery, clustering to obtain a target battery device comprising the target battery according to the distance matrix, forming a training subset by the aid of the other capacity attenuation curves except the target battery in the target battery device, adjusting the initial prediction model by the aid of the training subset to obtain a target prediction model, and predicting the battery life of the target battery by the aid of the target prediction model. The embodiment of the application also provides an energy storage battery life prediction device which comprises a determination module, an acquisition module, a clustering module, an adjustment module and a prediction module, wherein the determination module is used for determining a target battery to be predicted and an initial prediction model pre-trained by using a training total set, the acquisition module is used for acquiring a distance matrix calculated by a capacity attenuation curve, the capacity attenuation curve comprises a capacity attenuation curve of the training total set battery and a capacity attenuation curve of the target battery, the clustering module is used for clustering to obtain a target battery device containing the target battery according to the distance matrix, other capacity attenuation curves except the target battery in the target battery device form a training subset together, the adjustment module is used for adjusting the initial prediction model by using the training subset to obtain a target prediction model, and the prediction module is used for predicting the battery life of the target battery by using the target prediction model. The embodiment of the application also provides an energy storage system, which comprises the energy storage battery life prediction device or the energy storage battery life prediction device for executing the energy storage battery life prediction method. Compared with the related art, the method and the device for predicting the capacity of the battery have the advantages that after the target battery to be predicted and the initial prediction model pre-trained by the training total set are determined, the capacity attenuation curve of the battery in the training total set and the capacity attenuation curve of the target battery are obtained, and the distance matrix is obtained through calculation according to the capacity attenuation curve of the battery. The method comprises the steps of clustering according to a distance matrix to obtain a target batter