Dynamic battery life prediction method

Abstract

The invention discloses a dynamic battery life prediction method, and relates to the technical field of battery life prediction, and the method comprises the steps: connecting a battery performance detection system, and calling the factory performance data of a battery; performing real-time performance detection to obtain real-time performance data of the battery; analyzing performance degradation based on factory and real-time data, and determining a degradation matrix and a degradation coefficient; building a life prediction module, wherein the life prediction module comprises a life prediction gating channel and a life prediction channel; inputting the recession coefficient into a gating channel to generate a gating coefficient; and generating a battery life prediction result through a life prediction channel in combination with the gating coefficient and the recession matrix. Therefore, the technical effects of improving the adaptability and responsiveness of the prediction working condition and improving the prediction accuracy are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery life prediction, and in particular to a dynamic battery life prediction method. Background Art

[0002] With the widespread application of battery technology, especially in the fields of new energy vehicles, energy storage systems and mobile electronic devices, in actual use, the state and performance of the battery will change with the usage conditions and time. The actual battery can be considered as a dynamic battery. Accurate assessment of battery life is of great significance for improving equipment reliability, reducing maintenance costs and extending equipment life.

[0003] At present, battery life prediction technology mainly relies on statistical methods based on historical data, physical models and some algorithms based on machine learning. These methods have shown certain advantages in the early evaluation of battery life, but they are still insufficient in dynamic application scenarios and cannot capture the dynamic characteristics of battery performance degradation in real time, resulting in insufficient accuracy and robustness of the prediction results. There are technical problems such as poor adaptability to working conditions and low prediction accuracy. Summary of the invention

[0004] The present invention provides a dynamic battery life prediction method to solve the technical problems of poor adaptability to working conditions and low prediction accuracy in the prior art, and to achieve the technical effect of improving the adaptability and responsiveness of predicted working conditions and enhancing the prediction accuracy.

[0005] The present invention provides a dynamic battery life prediction method, comprising: Connect to the battery performance testing system to retrieve the battery factory performance testing data of the dynamic battery.

[0006] According to the battery performance detection system, real-time performance detection is performed on the dynamic battery to obtain real-time battery performance detection data.

[0007] The dynamic battery is analyzed for performance degradation according to the battery factory performance detection data and the battery real-time performance detection data to determine a battery performance degradation matrix and a battery performance degradation coefficient.

[0008] A battery life prediction module is constructed, wherein the battery life prediction module includes a life prediction gating channel and a battery life prediction channel.

[0009] The battery performance degradation coefficient is input into the life prediction gating channel to obtain the life prediction gating coefficient.

[0010] Based on the life prediction gating coefficient and the battery performance degradation matrix, a battery life prediction result is obtained according to the battery life prediction channel.

[0011] In a feasible implementation, the dynamic battery is analyzed for performance degradation according to the battery factory performance detection data and the battery real-time performance detection data to determine a battery performance degradation matrix and a battery performance degradation coefficient, including: The battery factory performance test data is sorted to establish a battery factory performance matrix; the battery real-time performance test data is sorted to generate a battery real-time performance matrix; the battery real-time performance matrix is ​​identified for deviations according to the battery factory performance matrix to obtain the battery performance degradation matrix; the battery performance degradation matrix is ​​evaluated to generate the battery performance degradation coefficient.

[0012] In a feasible implementation, evaluating according to the battery performance degradation matrix to generate the battery performance degradation coefficient includes: According to the battery performance degradation matrix, a performance degradation depth evaluation is performed to obtain a battery performance degradation depth coefficient; according to the battery performance degradation matrix, a performance degradation breadth evaluation is performed to obtain a battery performance degradation breadth coefficient; a battery performance degradation evaluation weighted network is constructed, wherein the battery performance degradation evaluation weighted network includes a battery performance degradation depth weight and a battery performance degradation breadth weight; the battery performance degradation depth coefficient and the battery performance degradation breadth coefficient are input into the battery performance degradation evaluation weighted network to obtain the battery performance degradation coefficient.

[0013] In a feasible implementation, a battery life prediction module is built, including: Life prediction gating feature learning is performed based on a battery performance decay coefficient sample set and a life prediction gating coefficient sample set to construct the life prediction gating channel; sample retrieval is performed based on the dynamic battery to obtain a plurality of sample batteries; life prediction feature learning is performed based on the plurality of sample batteries to construct the battery life prediction channel; the life prediction gating channel and the battery life prediction channel are connected to generate the battery life prediction module.

[0014] In a feasible implementation, life prediction gating feature learning is performed according to a battery performance degradation coefficient sample set and a life prediction gating coefficient sample set to construct the life prediction gating channel, including: Taking the battery performance decay coefficient sample set as input information and the life prediction gating coefficient sample set as output information, supervised training is performed on the life prediction gated learning model, and a gated learning loss coefficient is obtained after each predetermined number of trainings; it is determined whether the gated learning loss coefficient is less than a gated learning loss threshold; if the gated learning loss coefficient is less than the gated learning loss threshold, the life prediction gating channel is generated.

[0015] In a feasible implementation, performing life prediction feature learning based on the plurality of sample batteries to construct the battery life prediction channel includes: Perform life record retrieval based on the multiple sample batteries to obtain multiple sample battery life record sets, wherein each sample battery life record set includes a historical battery performance degradation matrix set and a historical battery life set corresponding to each sample battery; perform supervised learning on battery life prediction learning models based on the multiple sample battery life record sets to obtain multiple battery life prediction models; perform incremental learning on the multiple battery life prediction models based on the battery life record sets of the dynamic battery to obtain multiple battery life predictors; connect the multiple battery life predictors as parallel nodes to generate the battery life prediction channel.

[0016] In a feasible implementation, based on the life prediction gating coefficient and the battery performance degradation matrix, according to the battery life prediction channel, a battery life prediction result is obtained, including: The battery life prediction channel is feature activated according to the life prediction gating coefficient to obtain multiple activated life predictors; the battery performance degradation matrix is ​​input into the multiple activated life predictors to obtain multiple predicted battery lives; and the multiple predicted battery lives are integrated to generate the battery life prediction result.

[0017] In a feasible implementation, based on the life prediction gating coefficient and the battery performance degradation matrix, obtaining a battery life prediction result according to the battery life prediction channel also includes: According to the dynamic battery, a battery life prediction expectation is set; whether the battery life prediction result meets the battery life prediction expectation is determined; if the battery life prediction result does not meet the battery life prediction expectation, a battery life warning signal is generated.

[0018] The present invention discloses a dynamic battery life prediction method, including: connecting a battery performance detection system to retrieve the factory performance detection data of the dynamic battery; performing real-time performance detection on the dynamic battery according to the battery performance detection system to obtain the real-time performance detection data of the battery; performing performance degradation analysis on the dynamic battery according to the factory performance detection data of the battery and the real-time performance detection data of the battery, and determining the battery performance degradation matrix and the battery performance degradation coefficient; building a battery life prediction module, wherein the battery life prediction module includes a life prediction gating channel and a battery life prediction channel; inputting the battery performance degradation coefficient into the life prediction gating channel to obtain the life prediction gating coefficient; based on the life prediction gating coefficient and the battery performance degradation matrix, according to the battery life prediction channel, obtaining the battery life prediction result. The dynamic battery life prediction method disclosed by the present invention solves the technical problems of poor adaptability to working conditions and low prediction accuracy, and achieves the technical effect of improving the adaptability and responsiveness of the predicted working conditions and improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The figure is a flow chart of a dynamic battery life prediction method of the present invention.

[0020] Figure 2 The figure is a flow chart of a dynamic battery life prediction method of the present invention. DETAILED DESCRIPTION

[0021] The technical solution provided in the embodiments of the present invention is to solve the technical problems of poor adaptability to working conditions and low prediction accuracy in the prior art. The overall idea adopted is as follows: First, connect the battery performance detection system and retrieve the battery factory performance detection data of the dynamic battery; then, perform real-time performance detection on the dynamic battery according to the battery performance detection system to obtain the battery real-time performance detection data; then, perform performance degradation analysis on the dynamic battery according to the battery factory performance detection data and the battery real-time performance detection data to determine the battery performance degradation matrix and the battery performance degradation coefficient; then, build a battery life prediction module, wherein the battery life prediction module includes a life prediction gating channel and a battery life prediction channel; then, input the battery performance degradation coefficient into the life prediction gating channel to obtain the life prediction gating coefficient; finally, based on the life prediction gating coefficient and the battery performance degradation matrix, obtain the battery life prediction result according to the battery life prediction channel.

[0022] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0023] Example Figure 1 The present invention is a flowchart of a dynamic battery life prediction method, wherein the method comprises: Connect to the battery performance testing system to retrieve the battery factory performance testing data of the dynamic battery.

[0024] Specifically, first, a communication connection is established with a battery performance testing system through a data interface and a corresponding communication protocol to access the battery performance testing data, wherein the battery performance testing system is a system used to perform performance testing and verification on a target dynamic battery during the target battery production phase, and the battery performance testing system stores monitoring records of dynamic batteries. By interacting with the battery performance testing system, the battery factory performance testing data of the dynamic battery can be obtained.

[0025] Optionally, the battery factory performance test data of the dynamic battery includes parameters in multiple dimensions such as battery capacity, voltage (including discharge voltage and charge voltage), internal resistance, maximum charge and discharge rate, and discharge curve.

[0026] The above steps, by retrieving factory data, provide baseline data for subsequent comparison with the current performance status of the dynamic battery, which helps to quickly identify the performance change trend of the battery during use and achieve more accurate life prediction.

[0027] According to the battery performance detection system, real-time performance detection is performed on the dynamic battery to obtain real-time battery performance detection data.

[0028] Specifically, the same battery performance testing equipment and testing system used to obtain the battery factory performance test data is used to test the current performance of the dynamic battery, obtain the corresponding battery real-time performance test data, and avoid detection errors caused by equipment differences, thereby ensuring the consistency, accuracy and comparability of the obtained battery real-time performance test data, that is, battery tests performed at different time points have the same standards and measurement accuracy.

[0029] The dynamic battery is analyzed for performance degradation according to the battery factory performance detection data and the battery real-time performance detection data to determine a battery performance degradation matrix and a battery performance degradation coefficient.

[0030] Specifically, by comparing the battery factory performance test data and the battery real-time performance test data, the quantitative data of the performance change of the target dynamic battery from the factory to the current moment can be obtained, so as to analyze and obtain the corresponding battery performance degradation matrix and battery performance degradation coefficient.

[0031] Specifically, the battery performance degradation matrix is ​​a structured representation of the change data of multiple performance parameters of the target dynamic battery from leaving the factory to the current moment in matrix form, wherein, for example, the columns of the matrix represent different indicator items; the rows of the matrix correspond to different analysis moments (including usage time or number of cycles), such as the 100th cycle after leaving the factory and the 200th cycle after leaving the factory; the elements in the matrix are the changes of the corresponding indicator items at the corresponding moments.

[0032] Optionally, the degradation matrix is ​​updated over time through continuous updating of the real-time monitoring system and historical detection data, thereby providing the latest evaluation data for the battery degradation trend.

[0033] Specifically, based on the degradation matrix, various battery performances are quantitatively analyzed to obtain a specific performance degradation coefficient, which is used to quantitatively represent the performance degradation characteristics of the target dynamic battery.

[0034] The above process, through comparative analysis of the battery factory performance data and real-time performance data, constructs a battery performance degradation matrix and calculates the battery performance degradation coefficient, which can accurately track the battery degradation process and achieve quantitative evaluation.

[0035] In some embodiments, Figure 2 As shown, the dynamic battery is analyzed for performance degradation according to the battery factory performance detection data and the battery real-time performance detection data to determine the battery performance degradation matrix and the battery performance degradation coefficient, including: The battery factory performance test data is sorted to establish a battery factory performance matrix; the battery real-time performance test data is sorted to generate a battery real-time performance matrix; the battery real-time performance matrix is ​​identified for deviations according to the battery factory performance matrix to obtain the battery performance degradation matrix; the battery performance degradation matrix is ​​evaluated to generate the battery performance degradation coefficient.

[0036] Specifically, first, the factory performance test data of the battery is sorted out, and a battery factory performance matrix is ​​established based on a predetermined performance matrix format. Through this process, the structuring of various performance parameters of the battery when it leaves the factory is achieved, that is, they are summarized according to dimensions such as time and number of cycles. Exemplarily, the battery factory performance matrix is ​​a two-dimensional data structure, in which the horizontal axis represents the performance indicators of the battery (such as capacity, internal resistance, etc.), and the vertical axis represents different batteries or different performance measurement points, and each cell in the matrix stores the corresponding performance value.

[0037] Specifically, based on the above-mentioned method principle for data sorting of battery factory performance test data, the real-time performance test data of the battery is sorted according to a predetermined performance matrix format to generate a battery real-time performance matrix, which is consistent with the factory performance matrix format for subsequent comparative analysis.

[0038] Furthermore, the factory performance matrix is ​​compared with the real-time performance matrix, and the numerical differences of the corresponding indicators are calculated to identify the deviations in battery performance. These deviations reflect the degradation of battery performance over time, usage and other factors, and ultimately a battery performance degradation matrix is ​​generated. The matrix reflects the degradation of various performance properties of the battery (such as capacity, internal resistance, charge and discharge efficiency, etc.) during use. The deviation identification results may include absolute deviation and relative deviation, that is, the numerical change and the numerical change rate.

[0039] Specifically, after obtaining the battery performance degradation matrix, the overall degradation degree of the battery is evaluated according to the battery performance degradation matrix, and the corresponding degradation coefficient is calculated. The degradation coefficient is a comprehensive quantitative indicator reflecting the degradation degree of the battery health state. In other words, the degradation coefficient is the combined result of multiple degradation indicators. By accurately calculating the battery's performance degradation matrix and degradation coefficient, the degradation trend of the target dynamic battery can be accurately obtained, thereby providing a basis for subsequent life prediction.

[0040] In some implementations, performing evaluation according to the battery performance degradation matrix to generate the battery performance degradation coefficient includes: According to the battery performance degradation matrix, a performance degradation depth evaluation is performed to obtain a battery performance degradation depth coefficient; according to the battery performance degradation matrix, a performance degradation breadth evaluation is performed to obtain a battery performance degradation breadth coefficient; a battery performance degradation evaluation weighted network is constructed, wherein the battery performance degradation evaluation weighted network includes a battery performance degradation depth weight and a battery performance degradation breadth weight; the battery performance degradation depth coefficient and the battery performance degradation breadth coefficient are input into the battery performance degradation evaluation weighted network to obtain the battery performance degradation coefficient.

[0041] Specifically, the depth of performance degradation refers to the degree of degradation of battery performance during use relative to its performance when it leaves the factory. The depth of performance degradation is quantitatively measured through the deviation of performance indicators (such as capacity, voltage, etc.); the breadth of performance degradation refers to the proportion of multiple performance indicators of the battery that have declined within a certain period of time, which reflects the universality of battery degradation.

[0042] Optionally, based on multiple battery performance degradation matrices of multiple battery samples from the same batch of the target dynamic battery, performance degradation depth evaluation and performance degradation breadth evaluation are performed to ensure the representativeness and stability of the evaluation results obtained and avoid evaluation deviations caused by individual special samples.

[0043] Specifically, the battery performance degradation weighted network is a model that combines the degradation depth and the degradation breadth. The battery performance degradation weighted network comprehensively evaluates the battery degradation state by assigning different weights to the two. Among them, the battery performance degradation evaluation weighted network includes a degradation depth weight defined according to the degree of influence of the degradation depth on the battery health and a degradation breadth weight defined according to the degree of influence of the degradation breadth on the battery health. Optionally, the degradation depth weight and the degradation breadth weight can be dynamically adjusted according to the usage scenarios and working environments of different batteries to ensure that the obtained battery performance degradation coefficient can better reflect the scenario-based degradation characteristics of the battery under different scenarios.

[0044] A battery life prediction module is constructed, wherein the battery life prediction module includes a life prediction gating channel and a battery life prediction channel.

[0045] Specifically, the life prediction gating channel is the calling decision channel of the battery life prediction channel, that is, according to different dynamic battery attenuation conditions, the battery life prediction channel is adaptively called to perform life prediction, thereby improving the prediction efficiency of battery life while ensuring the prediction accuracy.

[0046] In some embodiments, a battery life prediction module is constructed, including: Life prediction gating feature learning is performed based on a battery performance decay coefficient sample set and a life prediction gating coefficient sample set to construct the life prediction gating channel; sample retrieval is performed based on the dynamic battery to obtain a plurality of sample batteries; life prediction feature learning is performed based on the plurality of sample batteries to construct the battery life prediction channel; the life prediction gating channel and the battery life prediction channel are connected to generate the battery life prediction module.

[0047] Specifically, the life prediction gating channel is used to determine the number of battery life predictors required to accurately predict the input information based on the real-time operating status or data of the battery. For example, the more difference dimensions and the more complex the difference distribution of the target dynamic battery, the more battery life predictors are needed to ensure the prediction accuracy.

[0048] Specifically, the battery life prediction channel is used to predict the performance degradation of the battery in the current or future period according to the battery performance degradation matrix and the corresponding life prediction gating coefficient of the battery, and then estimate the remaining service life of the battery. Preferably, the battery life prediction channel is a statistical model established based on the battery historical data, such as a polynomial regression model or a neural network model. Through the battery life prediction channel, a complex mapping relationship between the difference data and the remaining life of the target dynamic battery can be established.

[0049] Specifically, multiple sample batteries are retrieved from a battery library or historical data set. The sample batteries are the learning objects of the battery life prediction channel, and the multiple sample batteries are all batteries of the same or similar models as the target dynamic battery; the battery performance data of the sample batteries are representative, that is, they include battery performance data under different usage conditions, degradation states and environmental factors.

[0050] Specifically, the life prediction gating channel and the battery life prediction channel are connected to form a battery life prediction module, wherein the output of the gating channel will be passed as input to the life prediction channel to ensure whether to call the life predictor based on the current state of the battery; through the combination of the gating channel and the life prediction channel, the appropriate predictor can be automatically selected and called according to different battery states, ensuring the accuracy and efficiency of battery life prediction.

[0051] In some implementations, life prediction gating feature learning is performed based on the battery performance degradation coefficient sample set and the life prediction gating coefficient sample set to construct the life prediction gating channel, including: Taking the battery performance decay coefficient sample set as input information and the life prediction gating coefficient sample set as output information, supervised training is performed on the life prediction gated learning model, and a gated learning loss coefficient is obtained after each predetermined number of trainings; it is determined whether the gated learning loss coefficient is less than a gated learning loss threshold; if the gated learning loss coefficient is less than the gated learning loss threshold, the life prediction gating channel is generated.

[0052] Specifically, first, a battery performance degradation coefficient sample set is input as a sample, and the corresponding life prediction gating coefficient sample set is used as the desired output result, and supervised training of the life prediction gated learning model is performed, wherein the life prediction gated learning model is constructed based on a neural network or a decision tree, and the life prediction gating coefficient sample set is a sample set obtained based on the prediction performance evaluation, that is, the minimum life prediction gating coefficient corresponding to different battery performance degradation coefficients when the prediction accuracy requirement is met.

[0053] Specifically, through supervised learning, the input battery performance degradation coefficient sample set is mapped to the corresponding life prediction gating coefficient sample set. During the training process, the life prediction gating learning model predicts the life prediction gating coefficient according to the input sample, and adjusts the model parameters according to the error with the true target value. Preferably, the mean square error (MSE) is used as the loss function, and the gated learning loss coefficient is calculated after each training for a certain number of times.

[0054] Specifically, the gated learning loss threshold is used to determine whether the model is sufficiently trained and whether it can provide gating coefficient predictions with sufficient accuracy. After each training for a certain number of times (such as 100 times or 1000 times), check whether the current gated learning loss coefficient is less than the preset loss threshold. If the loss coefficient is less than the threshold, it can be considered that the life prediction gated learning model has converged and can effectively predict the gating coefficient. The training process is completed and the life prediction gating channel is constructed.

[0055] Through the above steps, different life predictors can be intelligently judged and called according to the battery's performance degradation status and historical data, thereby providing accurate battery remaining life estimation capabilities.

[0056] In some implementations, performing life prediction feature learning based on the plurality of sample batteries to construct the battery life prediction channel includes: Perform life record retrieval based on the multiple sample batteries to obtain multiple sample battery life record sets, wherein each sample battery life record set includes a historical battery performance degradation matrix set and a historical battery life set corresponding to each sample battery; perform supervised learning on battery life prediction learning models based on the multiple sample battery life record sets to obtain multiple battery life prediction models; perform incremental learning on the multiple battery life prediction models based on the battery life record sets of the dynamic battery to obtain multiple battery life predictors; connect the multiple battery life predictors as parallel nodes to generate the battery life prediction channel.

[0057] Specifically, the battery life prediction learning model is built based on machine learning models, such as support vector regression, long short-term memory, etc., and multiple battery life prediction models are built based on different model types and supervised training is performed through different sample data to ensure diversity.

[0058] Specifically, in the sample battery life record set, the historical battery performance degradation matrix set is a set of matrix data that describes the changes in battery performance over time or usage cycles. Each matrix in the historical battery performance degradation matrix set contains battery capacity, internal resistance, voltage and other degradation indicators; the historical battery life set is a set of records that describe the battery life, in which indicators such as the number of battery life usage cycles, percentage of effective capacity decrease, and number of charge and discharge times are represented.

[0059] Specifically, after supervised training of multiple battery life prediction models through a sample battery life record set, enhanced training of the multiple battery life prediction models is performed through the battery life record set of a target dynamic battery, thereby improving the prediction performance on the target dynamic battery and improving the prediction accuracy. The multiple battery life prediction models after enhanced training are the required multiple battery life predictors.

[0060] Furthermore, through the ensemble learning method, multiple independently trained battery life predictors are integrated as parallel nodes to form a parallel node network as a battery life prediction channel. Each node in the channel has independent prediction capabilities, and multiple nodes work together to make the final life prediction.

[0061] The above method steps, by training multiple battery life prediction models and integrating them into parallel nodes, can predict battery life from multiple angles, improve the accuracy and robustness of the prediction, and combine incremental learning to enable the model to adapt to changes in battery use in real time, thereby being suitable for life prediction tasks of different types of batteries.

[0062] The battery performance decay coefficient is input into the life prediction gating channel to obtain the life prediction gating coefficient.

[0063] Specifically, the battery performance degradation coefficient of the target dynamic battery obtained above is input into the constructed life prediction gating channel to obtain the corresponding life prediction gating coefficient. The life prediction gating coefficient gives the scale of nodes in the battery life prediction channel that needs to be called for the current battery life prediction task (i.e., the number of battery life predictors).

[0064] Based on the life prediction gating coefficient and the battery performance degradation matrix, a battery life prediction result is obtained according to the battery life prediction channel.

[0065] In some embodiments, based on the life prediction gating coefficient and the battery performance degradation matrix, obtaining a battery life prediction result according to the battery life prediction channel includes: The battery life prediction channel is feature activated according to the life prediction gating coefficient to obtain multiple activated life predictors; the battery performance degradation matrix is ​​input into the multiple activated life predictors to obtain multiple predicted battery lives; and the multiple predicted battery lives are integrated to generate the battery life prediction result.

[0066] Specifically, first, a corresponding number of life predictors in a battery life prediction channel are activated according to a life prediction gating coefficient, wherein the activation may be random activation and the life prediction gating coefficient is a positive integer; then, the battery performance degradation matrix is ​​input into the corresponding multiple activated life predictors for parallel prediction, and each activated life predictor calculates a predicted battery life using an inherent feature combination and weight; finally, a weighted average, a voting mechanism or a confidence-based comprehensive strategy is used to fuse multiple predicted battery lives given by multiple activated life predictors into a battery life prediction result, so as to improve the prediction accuracy and reliability.

[0067] In some embodiments, based on the life prediction gating coefficient and the battery performance degradation matrix, obtaining a battery life prediction result according to the battery life prediction channel further includes: According to the dynamic battery, a battery life prediction expectation is set; whether the battery life prediction result meets the battery life prediction expectation is determined; if the battery life prediction result does not meet the battery life prediction expectation, a battery life warning signal is generated.

[0068] Specifically, the battery life prediction expectation is the battery life value required for the target dynamic battery application scenario, exemplarily including the number of cycles and calendar life; by comparing the battery life prediction expectation with the battery life prediction result, it can be determined whether the target dynamic battery meets the application requirements; if the battery life prediction result does not meet the set expectations, a battery life warning signal is generated. The warning signal is used to notify the user, providing an opportunity for early intervention to perform maintenance or replace the battery, thereby reducing the risk of battery failure and unexpected downtime.

[0069] In summary, the dynamic battery life prediction method provided by the present invention has the following technical effects: By connecting to the battery performance detection system, the factory performance detection data of the dynamic battery is retrieved; according to the battery performance detection system, the dynamic battery is tested in real time to obtain the real-time performance detection data of the battery; according to the factory performance detection data and the real-time performance detection data of the battery, the dynamic battery is analyzed for performance degradation, and the battery performance degradation matrix and the battery performance degradation coefficient are determined; a battery life prediction module is built, wherein the battery life prediction module includes a life prediction gating channel and a battery life prediction channel; the battery performance degradation coefficient is input into the life prediction gating channel to obtain the life prediction gating coefficient; based on the life prediction gating coefficient and the battery performance degradation matrix, the battery life prediction result is obtained according to the battery life prediction channel. Thus, the technical effect of improving the adaptability and responsiveness of the predicted working conditions and improving the prediction accuracy is achieved.

[0070] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A dynamic battery life prediction method, characterized in that: The method comprises: Connect to the battery performance testing system to retrieve the factory performance testing data of the dynamic battery; According to the battery performance detection system, real-time performance detection is performed on the dynamic battery to obtain real-time battery performance detection data; Performing performance degradation analysis on the dynamic battery according to the battery factory performance detection data and the battery real-time performance detection data, and determining a battery performance degradation matrix and a battery performance degradation coefficient; Building a battery life prediction module, wherein the battery life prediction module includes a life prediction gating channel and a battery life prediction channel; Inputting the battery performance decay coefficient into the life prediction gating channel to obtain a life prediction gating coefficient; Based on the life prediction gating coefficient and the battery performance degradation matrix, a battery life prediction result is obtained according to the battery life prediction channel.

2. The method according to claim 1, characterized in that The dynamic battery is subjected to performance degradation analysis according to the battery factory performance detection data and the battery real-time performance detection data to determine a battery performance degradation matrix and a battery performance degradation coefficient, including: Arrange data based on the battery factory performance test data to establish a battery factory performance matrix; Arranging the battery real-time performance detection data to generate a battery real-time performance matrix; Perform deviation identification on the battery real-time performance matrix according to the battery factory performance matrix to obtain the battery performance degradation matrix; An evaluation is performed according to the battery performance degradation matrix to generate the battery performance degradation coefficient.

3. The method according to claim 2, characterized in that Evaluating according to the battery performance degradation matrix to generate the battery performance degradation coefficient includes: Performing a performance degradation depth evaluation according to the battery performance degradation matrix to obtain a battery performance degradation depth coefficient; Performing a performance degradation breadth evaluation according to the battery performance degradation matrix to obtain a battery performance degradation breadth coefficient; Constructing a battery performance degradation evaluation weighted network, wherein the battery performance degradation evaluation weighted network includes a battery performance degradation depth weight and a battery performance degradation breadth weight; The battery performance degradation depth coefficient and the battery performance degradation breadth coefficient are input into the battery performance degradation evaluation weighted network to obtain the battery performance degradation coefficient.

4. The method according to claim 1, characterized in that Build a battery life prediction module, including: Perform life prediction gating feature learning based on the battery performance decay coefficient sample set and the life prediction gating coefficient sample set to construct the life prediction gating channel; Perform sample retrieval according to the dynamic battery to obtain multiple sample batteries; Performing life prediction feature learning based on the plurality of sample batteries to construct the battery life prediction channel; The life prediction gating channel and the battery life prediction channel are connected to generate the battery life prediction module.

5. The method according to claim 4, characterized in that Life prediction gating feature learning is performed according to the battery performance decay coefficient sample set and the life prediction gating coefficient sample set to construct the life prediction gating channel, including: Using the battery performance decay coefficient sample set as input information and the life prediction gating coefficient sample set as output information, supervised training is performed on the life prediction gating learning model, and a gated learning loss coefficient is obtained after each predetermined number of trainings; Determining whether the gated learning loss coefficient is less than a gated learning loss threshold; If the gated learning loss coefficient is less than the gated learning loss threshold, the life prediction gated channel is generated.

6. The method according to claim 4, characterized in that Performing life prediction feature learning according to the plurality of sample batteries to construct the battery life prediction channel includes: Performing life record retrieval according to the plurality of sample batteries to obtain a plurality of sample battery life record sets, wherein each sample battery life record set includes a historical battery performance degradation matrix set and a historical battery life set corresponding to each sample battery; According to the multiple sample battery life record sets, supervised learning is performed on the battery life prediction learning models respectively to obtain multiple battery life prediction models; According to the battery life record set of the dynamic battery, incremental learning is performed on the multiple battery life prediction models respectively to obtain multiple battery life predictors; The plurality of battery life predictors are connected as parallel nodes to generate the battery life prediction channel.

7. The method according to claim 1, characterized in that Based on the life prediction gating coefficient and the battery performance degradation matrix, and according to the battery life prediction channel, a battery life prediction result is obtained, including: Performing feature activation on the battery life prediction channel according to the life prediction gating coefficient to obtain a plurality of activated life predictors; Inputting the battery performance degradation matrix into the multiple activated life predictors to obtain multiple predicted battery lifespans; The multiple predicted battery lifespans are integrated to generate the battery life prediction result.

8. The method according to claim 1, characterized in that Based on the life prediction gating coefficient and the battery performance degradation matrix, and according to the battery life prediction channel, a battery life prediction result is obtained, further comprising: According to the dynamic battery, setting a battery life prediction expectation; Determining whether the battery life prediction result meets the battery life prediction expectation; If the battery life prediction result does not meet the battery life prediction expectation, a battery life warning signal is generated.

Images