Method, device, equipment, medium and product for predicting energy storage battery life

Abstract

The present application relates to the technical field of energy storage battery, and discloses a method, device, equipment, medium and product for predicting the service life of an energy storage battery, the method comprising: obtaining first operation data of a first energy storage battery and second operation data of a second energy storage battery; wherein the first energy storage battery is an energy storage battery of the same model as the second energy storage battery and with a usage time greater than a first preset time period; adjusting the first operation data so that a first error distribution corresponding to the adjusted third operation data and a second error distribution corresponding to the second operation data satisfy a preset similar distribution condition; and predicting the service life of the second energy storage battery according to the third operation data and the second operation data. The present application uses the data of an energy storage battery of the same model as the to-be-predicted energy storage battery to predict the service life of the to-be-predicted energy storage battery, thereby improving the accuracy of service life prediction for newly used energy storage batteries.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage batteries, in particular to a method and device for predicting the life of an energy storage battery, equipment, medium and products. BACKGROUND

[0002] As an important energy storage method, energy storage batteries play an increasingly important role in the development of the electric vehicle industry. With the development of the electric vehicle industry, the demand for energy storage batteries is increasing, and therefore the life requirements for energy storage batteries are becoming higher and higher, and the life prediction of energy storage batteries is becoming more and more important.

[0003] However, in the process of predicting the life of the energy storage battery, the capacity of the energy storage battery is the most important parameter, and the capacity of the energy storage battery is difficult to measure directly in most cases. In the related art, the capacity of the energy storage battery is estimated according to the detected operating state data of the energy storage battery, and the life of the energy storage battery is predicted according to the capacity of the energy storage battery. The above method is mainly applicable to energy storage batteries with sufficient operating state data, but for newly used energy storage batteries and energy storage batteries lacking data detection devices, sufficient operating state data cannot be collected, and even if a small amount of operating state data is collected, the estimation of the capacity of the energy storage battery is not accurate enough due to insufficient operating state data, resulting in inaccurate life prediction of the energy storage battery. SUMMARY

[0004] Therefore, the present application provides a method and device for predicting the life of an energy storage battery to solve the problem of inaccurate life prediction of an energy storage battery due to insufficient operating state data.

[0005] In a first aspect, the present application provides a method for predicting the life of an energy storage battery, comprising: obtaining first operating data of a first energy storage battery and second operating data of a second energy storage battery; wherein the first energy storage battery is an energy storage battery of the same model as the second energy storage battery with a use time greater than a first preset time period, the second energy storage battery is an energy storage battery with a use time less than a second preset time period, and the first preset time period is greater than the second preset time period; adjusting the first operating data to make the first error distribution corresponding to the adjusted third operating data satisfy a preset similar distribution condition with the second error distribution corresponding to the second operating data; and predicting the life of the second energy storage battery according to the third operating data and the second operating data.

[0006] The application obtains first operation data of a first energy storage battery of the same type as a newly used second energy storage battery, adjusts the first operation data so that a first error distribution corresponding to the adjusted third operation data meets a preset similar distribution condition with a second error distribution corresponding to second operation data, adjusts the first operation data into third operation data whose error distribution meets the preset similar distribution condition with the second error distribution corresponding to the second operation data, and predicts the life of the second energy storage battery in combination with the third operation data and the second operation data. The application processes the first operation data of the first energy storage battery of the same type as the newly used second energy storage battery, and uses it for predicting the life of the newly used second energy storage battery, thereby overcoming the influence of individual differences between different energy storage batteries. Compared with related technologies, only a small amount of operation data of the newly used energy storage battery is needed to realize the life prediction of the newly used energy storage battery, thereby improving the life prediction accuracy of the energy storage battery.

[0007] In an optional implementation, the first operation data is adjusted so that the first error distribution corresponding to the adjusted third operation data meets the preset similar distribution condition with the second error distribution corresponding to the second operation data, including: performing weighted processing on the first operation data so that the similarity between the first error distribution corresponding to the adjusted third operation data and the second error distribution corresponding to the second operation data reaches a preset similarity.

[0008] The application performs weighted processing on the first operation data so that the first error distribution corresponding to the adjusted third operation data is as similar as possible to the second error distribution corresponding to the second operation data, thereby overcoming the influence of individual differences between different energy storage batteries.

[0009] In an optional implementation, the first operation data is weighted processed so that the similarity between the first error distribution corresponding to the adjusted third operation data and the second error distribution corresponding to the second operation data reaches a preset similarity, including: inputting the first operation data and the second operation data into a trained weight neural network to output the third operation data, the input of the weight neural network being the first operation data and the second operation data, the output of the weight neural network being the third operation data, and the weight neural network being used to add a weight value to the first operation data based on the second error distribution corresponding to the second operation data so that the similarity between the first error distribution corresponding to the obtained third operation data and the second error distribution reaches a preset similarity.

[0010] In an optional implementation, the life of the second energy storage battery is predicted according to the third operation data and the second operation data, including: inputting the third operation data and the second operation data into the trained prediction neural network, predicting the life of the second energy storage battery to obtain the remaining life percentage; the input of the prediction neural network is the third operation data and the second operation data, and the output of the prediction neural network is the remaining life percentage of the second energy storage battery.

[0011] In an optional implementation, the prediction neural network is used to fuse the third operation data and the second operation data to obtain target operation data, and is used to perform multi-scale feature extraction on the target operation data to obtain target extraction features, and is also used to perform cross-scale feature fusion on the target extraction features and output the remaining life percentage of the second energy storage battery.

[0012] In an optional implementation, after obtaining the first operation data corresponding to the first energy storage battery and the second operation data corresponding to the second energy storage battery, the prediction method of the life of the energy storage battery further includes: determining the first filling data quantity according to the data quantity and the feature distribution of the first operation data, and determining the second filling data quantity according to the data quantity and the feature distribution of the second operation data; identifying a first data position point of a missing value in the first operation data, and identifying a second data position point of a missing value in the second operation data; determining a first target data of the first filling data quantity in the first operation data according to a preset distance measurement function; determining a second target data of the second filling data quantity in the second operation data according to the preset distance measurement function; calculating the mean or median of the first target data to obtain a first calculation result, and filling the first calculation result into the first data position point; calculating the mean or median of the second target data to obtain a second calculation result, and filling the second calculation result into the second data position point.

[0013] The first operation data and the second operation data are filled in the application, the completeness of the first operation data and the second operation data is improved, the error when the first operation data and the second operation data are used subsequently is reduced, and the accuracy of the first operation data and the second operation data is improved.

[0014] In a second aspect, the present application provides a device for predicting the life of an energy storage battery, comprising: a data acquisition module configured to acquire first operation data of a first energy storage battery and second operation data of a second energy storage battery; wherein the first energy storage battery is an energy storage battery of the same model as the second energy storage battery and has been used for more than a first preset time period, the second energy storage battery is an energy storage battery that has been used for less than a second preset time period, and the first preset time period is greater than the second preset time period; a data adjustment module configured to adjust the first operation data so that a first error distribution corresponding to the adjusted third operation data and a second error distribution corresponding to the second operation data satisfy a preset similar distribution condition; and a life prediction module configured to predict the life of the second energy storage battery based on the third operation data and the second operation data.

[0015] In a third aspect, the present application provides a computer device, comprising: a memory and a processor, which are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the method for predicting the life of an energy storage battery according to the first aspect or any one of the corresponding embodiments thereof.

[0016] In a fourth aspect, the present application provides a computer readable storage medium, which stores computer instructions for causing a computer to perform the method for predicting the life of an energy storage battery according to the first aspect or any one of the corresponding embodiments thereof.

[0017] In a fifth aspect, the present application provides a computer program product, which comprises computer instructions for causing a computer to perform the method for predicting the life of an energy storage battery according to the first aspect or any one of the corresponding embodiments thereof. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the specific embodiments or the related art, the drawings needed in the specific embodiments or the related art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0019] Figure 1 is a flowchart of a method for predicting the life of an energy storage battery according to an embodiment of the present application.

[0020] Figure 2 is a flowchart of a data preprocessing process according to an embodiment of the present application.

[0021] Figure 3 is a flowchart of another method for predicting the life of an energy storage battery according to an embodiment of the present application.

[0022] Figure 4 This is a schematic diagram of the weighted CNN-BIGRU neural network structure according to an embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of the workflow of a weighted CNN-BIGRU neural network according to an embodiment of the present invention.

[0024] Figure 6 This is a flowchart illustrating a data filling method according to an embodiment of the present invention.

[0025] Figure 7 This is a flowchart illustrating the multi-scale CBiGRU experience transfer method for predicting the lifetime of energy storage batteries according to an embodiment of the present invention.

[0026] Figure 8 This is a schematic diagram of a one-dimensional convolutional multi-filter feature extraction process according to an embodiment of the present invention.

[0027] Figure 9 This is a schematic diagram of the local information selection module of the convolutional layer according to an embodiment of the present invention.

[0028] Figure 10 This is a schematic diagram of the fusion process of the BiGRU bidirectional fusion mechanism according to an embodiment of the present invention.

[0029] Figure 11 This is a schematic diagram of the convergence result of the predictive neural network according to an embodiment of the present invention.

[0030] Figure 12 This is a schematic diagram illustrating the process of predicting remaining lifespan using data from a new battery according to an embodiment of the present invention.

[0031] Figure 13 This is a structural block diagram of a device for predicting the lifespan of an energy storage battery according to an embodiment of the present invention.

[0032] Figure 14 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] As an important energy storage method, energy storage batteries play an increasingly important role in the wide application of renewable energy, grid stability and the development of electric vehicles. At present, with the increasing demand for energy storage batteries, an accurate energy storage battery prediction system becomes crucial, which plays a key role in battery management system, safe operation and life cycle management. For life cycle management, the energy storage battery prediction system is used to predict the life of the energy storage battery. However, the newly used energy storage battery group usually lacks sufficient historical operation data in the initial operation stage, so it is difficult to establish a prediction model with high precision.

[0035] At present, in the process of predicting the life of the energy storage battery, the capacity of the energy storage battery is the most important parameter, and the capacity of the energy storage battery is difficult to be directly measured in most cases. In the related technology, the capacity of the energy storage battery is estimated according to the detected operation state data of the energy storage battery, and the life of the energy storage battery is predicted according to the capacity of the energy storage battery. The traditional energy storage battery life prediction method mainly relies on statistical model and machine learning algorithm, such as support vector machine, multilayer perception neural network and the like. Such method can effectively utilize the historical operation data of the battery, and provide a scientific basis for the health management of the battery. However, this method is mainly suitable for energy storage battery groups with sufficient historical data, and for newly used energy storage batteries and energy storage batteries lacking data detection devices, sufficient operation state data cannot be collected. Even if a small amount of operation state data is collected, the estimation of the capacity of the energy storage battery is not accurate enough due to insufficient operation state data, resulting in inaccurate life prediction of the energy storage battery.

[0036] The embodiment of the present application provides a kind of prediction method of energy storage battery life, by the data of the same energy storage battery of the energy storage battery model to be predicted is used for the life prediction of newly used energy storage battery, to reach the effect of improving the life prediction accuracy of newly used energy storage battery.

[0037] According to the embodiment of the present application, a kind of prediction method of energy storage battery life is provided, it needs to be explained, the steps shown in the flow chart of the drawing can be executed in computer system such as a group of computer executable instructions, and although logical order is shown in flow chart, in some cases, the steps shown or described can be executed in different order from here.

[0038] A kind of prediction method of energy storage battery life is provided in the embodiment, it can be used in computer equipment, Figure 1 The flow chart of the prediction method of energy storage battery life according to the embodiment of the present application is as shown in Figure 1 The flow chart includes the following steps:

[0039] In step S101, first operation data of a first energy storage battery and second operation data of a second energy storage battery are acquired; the first energy storage battery is an energy storage battery of the same type as the second energy storage battery and having a use time greater than a first preset time period, and the second energy storage battery is an energy storage battery having a use time less than a second preset time period, the first preset time period being greater than the second preset time period.

[0040] The first energy storage battery can be an energy storage battery of the same type as the second energy storage battery in the same energy storage battery group, and the second energy storage battery can be a newly used energy storage battery and / or an energy storage battery lacking a data detection device. The time difference between the first preset time period and the second preset time period is greater than a first preset value, which can be set according to actual conditions. In the embodiment of the present application, the first preset value is as large as possible, that is, the time difference between the first preset time period and the second preset time period is as large as possible, that is, the use time of the first energy storage battery is much greater than the use time of the second energy storage battery, and thus the data quantity of the first operation data is much greater than the data quantity of the second operation data.

[0041] In the embodiment of the present application, the first energy storage battery and the second energy storage battery can be lithium batteries, and the first operation data and the second operation data can be data represented in a matrix form.

[0042] In some optional embodiments, the first operation data and the second operation data are data of different energy storage batteries and have the same attribute. The first operation data can be data collected according to an actual operation process or data obtained according to a public data set, and the second operation data is data collected according to an actual operation process. Exemplarily, the first operation data can be one or more of battery voltage, current, temperature, sampling time, and state of charge of the first energy storage battery, and the second operation data can be one or more of battery voltage, current, temperature, sampling time, and state of charge of the second energy storage battery. As shown in Table 1, some operation data are illustrated.

[0043] Table 1, some operation data.

[0044]

[0045] In some optional embodiments, after the first operation data of the first energy storage battery and the second operation data of the second energy storage battery are acquired, the method for predicting the service life of the energy storage battery further includes: pre-processing the first operation data and the second operation data.

[0046] As Figure 2As shown, it is a data preprocessing process flow diagram, including outlier removal processing based on mean algorithm in data; using neighbor algorithm to fill in missing values; normalization processing, dimension processing, and data reconstruction.

[0047] wherein, Figure 2 The data in the first and second running data. The mean algorithm can be K-means algorithm, and the plurality of data elements in the data are respectively subjected to outlier processing and divided into two clusters. The cluster with less number is determined as the outlier, and the outlier in the data is removed. The neighbor algorithm can be KNN (K-Nearest Neighbors, K-Nearest Neighbors) algorithm, which is used to identify the missing value position and fill in the missing value position according to the neighbor data.

[0048] The data is normalized according to the following formula:

[0049]

[0050] wherein, is the normalized value; is the original value; is the data mean; is the standard deviation.

[0051] In some optional embodiments, the data is subjected to dimension elimination processing by using 0-1 standardization, and other normalization methods can also be used.

[0052] In the embodiments of the application, the K-means algorithm can process the outlier problem in high-dimensional big data, and has the characteristics of high processing precision and fast calculation speed. KNN can effectively fill in the small amount of missing values in the data.

[0053] Step S102, adjusting the first running data, so that the first error distribution corresponding to the adjusted third running data and the second error distribution corresponding to the second running data satisfy the preset similar distribution condition.

[0054] wherein, the first running data is adjusted, including: weighting processing the first running data.

[0055] Further, the first running data is weighted, including: inputting the first running data and the second running data into the trained weight neural network, weighting the first running data, and outputting the third running data.

[0056] The weight neural network is used to add a weight value to the first running data based on a second error distribution corresponding to the second running data, so that a similarity between a first error distribution corresponding to the third running data and the second error distribution reaches a preset similarity.

[0057] In some optional embodiments, the preset similarity condition is that the similarity between the first error distribution corresponding to the third running data and the second error distribution reaches a preset similarity. The preset similarity can be set according to actual conditions. For example, the preset similarity can be 99%.

[0058] In the embodiments of the present application, the type of error distribution can be set according to actual conditions. For example, the type of error distribution can be normal distribution, uniform distribution, Poisson distribution, binomial distribution, etc.

[0059] In some optional embodiments, the weight neural network gives different weights to the first running data to change the error distribution of the first running data. The weight neural network can be composed of a BP neural network (Back Propagation Neural Network).

[0060] In step S103, the life of the second energy storage battery is predicted according to the third running data and the second running data.

[0061] In step S103, the life of the second energy storage battery is predicted according to the third running data and the second running data.

[0062] In step S103, the life of the second energy storage battery is predicted according to the third running data and the second running data.

[0063] For example, the remaining life percentage of the second energy storage battery can be 80%.

[0064] The method for predicting the service life of the energy storage battery provided in the embodiment comprises the following steps: obtaining first operation data of a first energy storage battery of the same type as a second energy storage battery newly put into use; adjusting the first operation data, so that a first error distribution corresponding to the adjusted third operation data meets a preset similar distribution condition with a second error distribution corresponding to second operation data; adjusting the first operation data into third operation data whose error distribution meets the preset similar distribution condition with the second error distribution corresponding to the second operation data; and combining the third operation data and the second operation data to predict the service life of the second energy storage battery. The first operation data of the first energy storage battery of the same type as the second energy storage battery newly put into use is processed to predict the service life of the second energy storage battery newly put into use, so that the influence of individual differences between different energy storage batteries is overcome. Compared with the related art, only a small amount of operation data of the energy storage battery newly put into use is required to predict the service life of the energy storage battery newly put into use, and the service life prediction accuracy of the energy storage battery is improved.

[0065] In the embodiment, a method for predicting the service life of an energy storage battery is provided, which can be used for a computer device, Figure 3 FIG. 6 is a flowchart of another method for predicting the service life of an energy storage battery according to an embodiment of the present application, as shown in the figure, the flowchart comprises the following steps: Figure 3

[0066] In step S301, first operation data of a first energy storage battery and second operation data of a second energy storage battery are obtained; the first energy storage battery is an energy storage battery of the same type as the second energy storage battery and with a use time greater than a first preset time period, the second energy storage battery is an energy storage battery with a use time less than a second preset time period, and the first preset time period is greater than the second preset time period. For details, refer to step S101 of the embodiment shown in FIG. 1, which will not be described here again. Figure 1

[0067] In step S302, the first operation data is adjusted, so that a first error distribution corresponding to the adjusted third operation data meets a preset similar distribution condition with a second error distribution corresponding to the second operation data.

[0068] Specifically, step S302 comprises the following steps:

[0069] In step S3021, the first operation data is weighted, so that a similarity between the first error distribution corresponding to the adjusted third operation data and the second error distribution corresponding to the second operation data reaches a preset similarity.

[0070] In some optional embodiments, step S3021 comprises the following steps:

[0071] ​​Step a1, inputting the first operation data and the second operation data into the trained weight neural network to output third operation data, the input of the weight neural network being the first operation data and the second operation data, the output of the weight neural network being the third operation data, the weight neural network being used to add a weight value in the first operation data based on a second error distribution corresponding to the second operation data, so that a similarity between a first error distribution corresponding to the obtained third operation data and the second error distribution reaches a preset similarity.

[0072] In some optional embodiments, the method for predicting the service life of the energy storage battery further includes a process of training the weight neural network, and the process of training the weight neural network includes: inputting historical energy storage battery operation data and newly put-into-use energy storage battery operation data into the weight neural network, training the weight neural network, and outputting weighted energy storage battery operation data.

[0073] The first operation data and the second operation data are distributed with differences, and direct migration of the first operation data may cause negative migration. In order to reduce the domain difference, the weight neural network models the dependent relationship between the first operation data and the corresponding weight, so that the weighted third operation data is as similar as possible to the second operation data distribution. In the training process, when the number of training times reaches a set value or an evaluation function of the first operation data and the second operation data reaches a target value, the prediction network is output.

[0074] Step S303, predicting the service life of the second energy storage battery according to the third operation data and the second operation data.

[0075] Specifically, the above step S303 includes:

[0076] Step S3031, inputting the third operation data and the second operation data into the trained prediction neural network to predict the service life of the second energy storage battery and obtain a remaining service life percentage; the input of the prediction neural network being the third operation data and the second operation data, and the output of the prediction neural network being the remaining service life percentage of the second energy storage battery.

[0077] In some optional embodiments, the method for predicting the service life of the energy storage battery further includes a process of training the prediction neural network, and the process of training the prediction neural network includes: inputting historical weighted operation data and newly put-into-use energy storage battery operation data into the prediction neural network, training the prediction neural network, and outputting a remaining service life percentage.

[0078] In some optional embodiments, the prediction neural network is configured to fuse the third operation data and the second operation data to obtain target operation data, and configured to perform multi-scale feature extraction on the target operation data to obtain target extraction features, and configured to perform cross-scale feature fusion on the target extraction features and output the remaining life percentage of the second energy storage battery.

[0079] In the embodiment of the present application, the multi-scale CNN network is used to perform multi-scale feature extraction on the target operation data to obtain target extraction features, and the BiGRU bidirectional fusion mechanism is used to perform cross-scale feature fusion on the target extraction features and output the remaining life percentage of the second energy storage battery.

[0080] In some optional embodiments, as shown in FIG. 6, a weighted CNN-BIGRU neural network structure diagram is shown. The first operation data and the second operation data are input into the weighted neural network. The weighted neural network includes a plurality of hidden layers. After the formaldehyde treatment of the plurality of hidden layers, the target operation data is obtained. The target operation data is input into the prediction neural network. The prediction neural network includes a plurality of multi-scale CNN-BIGRU layers. After the feature extraction and fusion of the target operation data by the plurality of multi-scale CNN-BIGRU layers, the remaining life percentage is output. Figure 4 In some optional embodiments, as shown in FIG. 7, a weighted CNN-BIGRU neural network workflow diagram is shown. The first operation data and the second operation data are input into the weighted neural network. The weighted neural network gives the initial weight to the first operation data. It is judged whether the training number of the weighted neural network reaches the preset training number. If it reaches, the target operation data is input into the final prediction neural network for prediction. If it does not reach, the distribution fitting is performed on the first operation data and the second operation data. It is judged whether the prediction neural network obtains the optimal solution. If the optimal solution is obtained, the target operation data is input into the final prediction neural network for prediction. If the optimal solution is not obtained, the weight parameter is adjusted, the distribution difference is reduced, and it is judged again whether the preset number is reached until the prediction neural network obtains the optimal solution.

[0081] Figure 5 The prediction method of the energy storage battery life provided in the embodiment is used to perform weighted processing on the first operation data, so that the first error distribution corresponding to the adjusted third operation data and the second error distribution corresponding to the second operation data are as similar as possible, thereby overcoming the influence of individual differences between different energy storage batteries. The improvement of the multi-scale CNN network local information selection mechanism and the BiGRU bidirectional fusion mechanism can extract information from the operation state data and effectively fuse multi-scale features, thereby significantly improving the accuracy of the remaining life of the energy storage battery.

[0082] The prediction method of the energy storage battery life provided in the embodiment is used to perform weighted processing on the first operation data, so that the first error distribution corresponding to the adjusted third operation data and the second error distribution corresponding to the second operation data are as similar as possible, thereby overcoming the influence of individual differences between different energy storage batteries. The improvement of the multi-scale CNN network local information selection mechanism and the BiGRU bidirectional fusion mechanism can extract information from the operation state data and effectively fuse multi-scale features, thereby significantly improving the accuracy of the remaining life of the energy storage battery.

[0083] ​The embodiment provides a method for predicting the service life of an energy storage battery, and further comprises a data filling process, and can be used for a computer device, Figure 6 is a flowchart of a data filling method according to an embodiment of the present application, after first running data corresponding to a first energy storage battery and second running data corresponding to a second energy storage battery are obtained, the data filling process comprises the following steps:

[0084] In step S601, the number of first filling data is determined according to the data amount and feature distribution of the first running data, and the number of second filling data is determined according to the data amount and feature distribution of the second running data.

[0085] The selection of the number of first filling data and the number of second filling data is related to the size of the data set and the distribution of the features, and the number of first filling data and the number of second filling data can be 3 exemplarily.

[0086] In step S602, a first data position point of a missing value in the first running data is identified, and a second data position point of a missing value in the second running data is identified.

[0087] In step S603, a first target data of the number of first filling data is determined in the first running data according to a preset distance measurement function, and a second target data of the number of second filling data is determined in the second running data according to the preset distance measurement function.

[0088] In step S604, the mean or median of the first target data is calculated to obtain a first calculation result, and the first calculation result is filled into the first data position point, and the mean or median of the second target data is calculated to obtain a second calculation result, and the second calculation result is filled into the second data position point.

[0089] The embodiment provides a method for predicting the service life of an energy storage battery by multi-scale CBiGRU experience migration, which can be used for a computer device, Figure 7 is a flowchart of a method for predicting the service life of an energy storage battery by multi-scale CBiGRU experience migration according to an embodiment of the present application, as shown in Figure 7 The flowchart comprises the following steps:

[0090] In step S701, outlier values are removed, missing values are filled, and normalization is performed.

[0091] The outlier values in the data matrix are removed based on the K-means algorithm, the missing values are filled by using the KNN algorithm, and normalization is performed.

[0092] In step S702, the model is pre-trained according to historical data.

[0093] The migration model is pre-trained according to historical operation data of the present operation battery of the energy storage battery, wherein the migration model comprises a weighted neural network and a prediction neural network, and the prediction neural network is composed of a multi-scale CNN network and a BiGRU bidirectional fusion mechanism.

[0094] In some optional embodiments, the multi-scale CNN network mainly extracts feature information from local to global through a series of filters. The multi-scale CNN network of the embodiment of the present application adopts multiple multi-scale filters to perform convolution operation at the same time, so as to increase the extraction ability of multi-dimensional features. As shown in FIG. 3, it is a one-dimensional convolution multi-filter feature extraction process diagram. If the size of a convolution kernel is (3, 6) and four filters are used at the same time, the dimension of the output will be (3, 6, 4). In the embodiment of the present application, the CNN model uses four channels, and the size of the convolution kernel is 2, 3, 4 and 5 respectively. As shown in FIG. 4, it is a convolution layer local information selection module diagram. The channels in the figure are represented as Figure 8 Figure 9 As shown in FIG. 5, it is a convolution layer local information selection module diagram. The channels in the figure are represented as

[0095] In some optional embodiments, as shown in FIG. 6, it is a BiGRU bidirectional fusion mechanism fusion process diagram, which comprises an input layer, a hidden layer and an output layer. The input of the input layer is a plurality of target extraction features, and the output of the output layer is a plurality of residual life percentages. The hidden layer comprises two directions of GRU (Gated Recurrent Unit), which processes the forward and reverse information of the sequence respectively, splices the different forward and reverse information, and obtains a plurality of target fusion features. Figure 10

[0096] ​​​The BiGRU bidirectional fusion mechanism of the embodiment of the application realizes more comprehensive understanding and modeling of multi-resolution features in sequence data by simultaneously considering forward and reverse information of the sequence. The forward GRU processes data from the starting position to the ending position of the sequence, and the reverse GRU processes data from the ending position to the starting position of the sequence. By fusing the outputs of the GRU in the two directions, richer sequence representation can be obtained. The BiGRU bidirectional fusion mechanism can capture long-term dependencies and context information in the sequence, and improve the performance and accuracy of the migration model. Through the BiGRU bidirectional fusion mechanism, the features of adjacent scales are fused, so that the output information can represent more scale features. In the embodiment of the application, after adjusting the parameters, the hidden layer size is set to 256, the vector dimension is set to 100, the activation function is set to ReLU, the regularization technique is set to Dropout, the Dropout parameter is set to 0.2, and the batch size is set to 128. It should be noted that the parameters in the embodiment of the application are the optimized parameters corresponding to the data set, and in the training process of other data sets, the hyperparameters need to be re-optimized.

[0097] In some optional embodiments, as shown in FIG. 6, which is a schematic diagram of the convergence result of the prediction neural network, the horizontal coordinate is the round, and the vertical coordinate is the loss size. The more the training loss and the validation loss coincide, the better the convergence effect of the migration model. Figure 11

[0098] In some optional embodiments, when the migration model converges, the migration model can be further evaluated. The embodiment of the application uses the mean absolute percentage error (MAPE) as the main evaluation index, and selects the mean absolute error (MAE) as the auxiliary evaluation index. In addition to the MAPE and MAE of the embodiment of the application, performance indicators that are more close to the project requirements can also be selected for evaluation according to the task requirements. The calculation formulas of the MAPE and the MAE in the embodiment of the application are as follows:

[0099]

[0100]

[0101] In the formula, the percentage error is the average absolute error is the actual data is the predicted data is the mean value is the sample number of data is

[0102] ​​In this embodiment of the invention, the trained transfer model is comprehensively evaluated, including but not limited to assessing metrics such as accuracy, stability, and generalization ability. Through testing and analysis, the model is properly stored in the device in a specific format and storage method to facilitate subsequent model transfer. This allows for rapid retrieval and use in new battery applications, providing strong support for the efficient execution of related tasks.

[0103] Step S703: Use the new battery data to predict battery life.

[0104] The process involves inputting new battery data and historical battery data into a migration model to predict battery life. The migration model takes new battery data and historical battery data as inputs and outputs the percentage of remaining battery life.

[0105] In some alternative implementations, such as Figure 12 The diagram illustrates the process of predicting remaining battery life using new battery data, selecting four historical battery data points to determine the predicted values. First, the source model is imported by learning from the source task data; at this stage, the source model parameters are frozen. Re-encoding the input and output layers allows for the transfer of experience from the source task data, resulting in a transfer model. In this process, the transfer model can quickly adapt to the new task environment, reducing its dependence on a large amount of target task data. It should be noted that if there are changes in the number of historical battery data points and output values, the parameters of the input and output layers can be adjusted. Alternatively, the intermediate layers can be left unfrozen, allowing for fine-tuning of the entire transfer model. It is important to note that the learning rate should not be too high during this process. Finally, the new battery data is input into the transfer model for life prediction.

[0106] In this embodiment of the invention, during the entire prediction process of the transfer model, an initial knowledge framework is first built by learning from the source task data. At this stage, the transfer model parameters are frozen. When faced with a new target task, the transfer model uses data transferred from the source task, combined with a small amount of data from the target task, for fine-tuning. In this process, the transfer model can quickly adapt to the new task environment, reducing its dependence on a large amount of target task data. Simultaneously, through continuous iteration and optimization, the transfer model gradually improves its prediction accuracy for the target task. Throughout the prediction process, the transfer model fully leverages the advantages of knowledge transfer, effectively improving the model's generalization ability and efficiency, and enabling accurate prediction of the remaining lifespan of new energy storage batteries.

[0107] In the embodiments of the present application, the source task data and the historical battery data are operation data of the energy storage battery after being used for a period of time, i.e., first operation data, and the target task data and the new battery data are operation data of the energy storage battery newly used, i.e., second operation data.

[0108] In the embodiments, a device for predicting the life of the energy storage battery is also provided, which is used to implement the above embodiments and preferred embodiments, and will not be described again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, implementation of hardware, or a combination of software and hardware, is also possible and contemplated.

[0109] The embodiments provide a device for predicting the life of the energy storage battery, as shown in Figure 13 The device comprises:

[0110] The data acquisition module 1301 is configured to acquire first operation data of a first energy storage battery and second operation data of a second energy storage battery; the first energy storage battery is an energy storage battery of the same type as the second energy storage battery and used for more than a first preset time period, and the second energy storage battery is an energy storage battery used for less than a second preset time period, and the first preset time period is greater than the second preset time period.

[0111] The data adjustment module 1302 is configured to adjust the first operation data, so that a first error distribution corresponding to the adjusted third operation data and a second error distribution corresponding to the second operation data satisfy a preset similar distribution condition.

[0112] The life prediction module 1303 is configured to predict the life of the second energy storage battery according to the third operation data and the second operation data.

[0113] In some optional embodiments, the data adjustment module 1302 comprises:

[0114] The weighting processing unit is configured to perform weighting processing on the first operation data, so that the similarity between the first error distribution corresponding to the adjusted third operation data and the second error distribution corresponding to the second operation data reaches a preset similarity.

[0115] Specifically, the weighting processing unit comprises:

[0116] The weight neural network running subunit is configured to input the first running data and the second running data into the trained weight neural network, and output third running data, wherein the input of the weight neural network is the first running data and the second running data, the output of the weight neural network is the third running data, and the weight neural network is configured to add a weight value to the first running data based on the second error distribution corresponding to the second running data, so that the similarity between the first error distribution corresponding to the obtained third running data and the second error distribution reaches a preset similarity.

[0117] In some optional embodiments, the life prediction module 1303 comprises:

[0118] The prediction neural network running unit is configured to input the third running data and the second running data into the trained prediction neural network, and predict the life of the second energy storage battery to obtain a remaining life percentage, wherein the input of the prediction neural network is the third running data and the second running data, and the output of the prediction neural network is the remaining life percentage of the second energy storage battery.

[0119] In some optional embodiments, the energy storage battery life prediction device further comprises:

[0120] The filling data quantity determination module is configured to determine a first filling data quantity according to the data quantity and the feature distribution of the first running data, and determine a second filling data quantity according to the data quantity and the feature distribution of the second running data.

[0121] The data position point identification module is configured to identify a first data position point of a missing value in the first running data, and identify a second data position point of a missing value in the second running data.

[0122] The target data determination module is configured to determine, in the first running data, first target data of the first filling data quantity according to a preset distance measurement function, and determine, in the second running data, second target data of the second filling data quantity according to the preset distance measurement function.

[0123] The data filling module is configured to calculate a mean value or a median value of the first target data to obtain a first calculation result, and fill the first calculation result into the first data position point; and calculate a mean value or a median value of the second target data to obtain a second calculation result, and fill the second calculation result into the second data position point.

[0124] Further function descriptions of the above-mentioned various modules and units are the same as those of the above-mentioned corresponding embodiments, and will not be described here again.

[0125] In this embodiment, the energy storage battery life prediction device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0126] This invention also provides a computer device having the above-described features. Figure 13 The device shown is for predicting the lifespan of energy storage batteries.

[0127] Please see Figure 14 , Figure 14 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 14 As shown, the computer device includes one or more processors 1410, memory 1420, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 14 Take the 1410 processor as an example.

[0128] Processor 1410 may be a central processing unit, a network processor, or a combination thereof. Processor 1410 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0129] The memory 1420 stores instructions executable by at least one processor 1410 to cause the at least one processor 1410 to perform the method shown in the above embodiments.

[0130] The memory 1420 can include a program storage area and a data storage area. The program storage area can store the operating system, application programs, and / or data needed to execute the application programs. The data storage area can store data created by the computer device, etc. Additionally, the memory 1420 can include a volatile and / or non-volatile memory. In some embodiments, the memory 1420 can include a memory that is remote from the processor 1410, which can be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranet, local area networks, mobile communication networks, and combinations thereof.

[0131] The memory 1420 can include a volatile memory, such as random access memory, and / or can include a non-volatile memory, such as flash memory, hard disk, or solid state disk. The memory 1420 can also include a combination of the above-mentioned types of memory.

[0132] The computer device also includes a communication interface 1430 for communicating with other devices or communication networks.

[0133] The embodiments of the present application also provide a computer readable storage medium, the method according to the embodiments of the present application can be implemented in hardware, firmware, or recorded in a storage medium, or be implemented through computer code stored in a remote storage medium or non-transitory machine readable storage medium and stored in a local storage medium, so that the method described herein can be processed by such software on a storage medium using a general purpose computer, a special purpose processor, or programmable or special purpose hardware. The storage medium can be a disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid state disk, etc. Further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that the computer, processor, microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, when the software or computer code is accessed and executed by the computer, processor, or hardware, the method shown in the above embodiments is implemented.

[0134] Part of the present application can be applied as a computer program product, for example, computer program instructions, when executed by a computer, through the operation of the computer, can invoke or provide the method and / or technical solutions according to the present application. Those skilled in the art should understand that the form of computer program instructions in computer readable medium includes but is not limited to source files, executable files, installation package files and the like, and accordingly, the way of computer program instructions executed by computer includes but is not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Here, the computer readable medium can be any available computer readable storage medium or communication medium accessible to the computer.

[0135] Although the embodiments of the present application are described in conjunction with the drawings, various modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the present application, and such modifications and changes fall within the scope defined by the appended claims.

Claims

1. A method for predicting the lifespan of an energy storage battery, characterized in that, The method includes: Acquire first operating data of the first energy storage battery and second operating data of the second energy storage battery; wherein, the first energy storage battery is an energy storage battery of the same model as the second energy storage battery with a usage time greater than a first preset time period, and the second energy storage battery is an energy storage battery with a usage time less than a second preset time period, and the first preset time period is greater than the second preset time period. The first running data and the second running data are input into the trained weighted neural network, and the third running data is output. The input of the weighted neural network is the first running data and the second running data, and the output of the weighted neural network is the third running data. The weighted neural network is used to add weight values ​​to the first running data based on the second error distribution corresponding to the second running data, so that the similarity between the first error distribution corresponding to the third running data and the second error distribution reaches a preset similarity. The lifespan of the second energy storage battery is predicted based on the third operating data and the second operating data.

2. The method according to claim 1, characterized in that, The step of predicting the lifespan of the second energy storage battery based on the third operating data and the second operating data includes: The third operating data and the second operating data are input into the trained predictive neural network to predict the lifespan of the second energy storage battery and obtain the remaining lifespan percentage. The input of the predictive neural network is the third operating data and the second operating data, and the output of the predictive neural network is the remaining lifespan percentage of the second energy storage battery.

3. The method according to claim 2, characterized in that, The predictive neural network is used to fuse the third operating data and the second operating data to obtain target operating data, and to perform multi-scale feature extraction on the target operating data to obtain target extracted features. It is also used to perform cross-scale feature fusion on the target extracted features and output the remaining lifespan percentage of the second energy storage battery.

4. The method according to claim 1, characterized in that, After acquiring the first operating data corresponding to the first energy storage battery and the second operating data corresponding to the second energy storage battery, the method further includes: Based on the data volume and characteristic distribution of the first running data, determine the first amount of fill data; based on the data volume and characteristic distribution of the second running data, determine the second amount of fill data. Identify the first data location point of the missing value in the first running data, and identify the second data location point of the missing value in the second running data; Based on a preset distance metric function, a first target data for the first number of fill data is determined in the first running data; based on a preset distance metric function, a second target data for the second number of fill data is determined in the second running data. Calculate the mean or median of the first target data to obtain a first calculation result, and fill the first calculation result into the first data position point; calculate the mean or median of the second target data to obtain a second calculation result, and fill the second calculation result into the second data position point.

5. A device for predicting the lifespan of an energy storage battery, characterized in that, The device includes: The data acquisition module is used to acquire first operating data of the first energy storage battery and second operating data of the second energy storage battery; wherein, the first energy storage battery is an energy storage battery of the same model as the second energy storage battery with a usage time greater than a first preset time period, and the second energy storage battery is an energy storage battery with a usage time less than a second preset time period, and the first preset time period is greater than the second preset time period. The data adjustment module is used to input the first running data and the second running data into the trained weighted neural network and output the third running data. The input of the weighted neural network is the first running data and the second running data, and the output of the weighted neural network is the third running data. The weighted neural network is used to add weight values ​​to the first running data based on the second error distribution corresponding to the second running data, so that the similarity between the first error distribution corresponding to the third running data and the second error distribution reaches a preset similarity. The lifespan prediction module is used to predict the lifespan of the second energy storage battery based on the third operating data and the second operating data.

6. A computer device, characterized in that, include: A memory and a processor are communicatively connected, the memory storing computer instructions, and the processor executing the computer instructions to perform the method for predicting the lifespan of an energy storage battery as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to execute the method for predicting the lifespan of an energy storage battery as described in any one of claims 1 to 4.

8. A computer program product, characterized in that, Includes computer instructions for causing a computer to execute the method for predicting the lifespan of an energy storage battery as described in any one of claims 1 to 4.

Images