Battery pack life prediction method, system, device

Abstract

The application discloses a storage battery pack life prediction method, system and device, belongs to the technical field of storage batteries, and relates to a storage battery pack life prediction method, system and device, which comprises the following steps: step 1, collecting working information of each battery to be managed in real time; step 2, obtaining transformation information of a predicted state transition value of each battery according to the working information of each battery; and step 3, obtaining a battery life score of the battery according to the transformation information.In the technical scheme, the transformation information of the state transition value is used to predict the battery life score, so that the battery life score of a single battery is predicted with sufficient accuracy in practice.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery pack, in particular, relates to a battery pack life prediction method, system and device. BACKGROUND

[0002] In large photovoltaic power stations, energy storage power stations and other places, there are a large number of battery packs, which are distributed in various areas of the station and used to store the generated power. These battery packs are constantly charged and discharged during operation, and the batteries in the battery pack will gradually age, so they need to be repaired or replaced. In the prior art, the batteries are maintained only after a fault alarm is issued when the battery fails. Therefore, the batteries cannot be managed in a timely manner.

[0003] Currently, the predicted life of the battery is directly predicted by fitting the historical use data of the battery according to the use frequency, internal resistance and other data of the battery. For a single battery, the accuracy of this battery life prediction method is low. SUMMARY

[0004] The summary part of the present application is used to introduce the concept in a simple form, which will be described in detail in the specific embodiment part. The summary part of the present application is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0005] As a first aspect of the present application, in order to solve the technical problems mentioned in the background part, the present application provides a battery pack life prediction method, comprising the following steps:

[0006] Step 1: Collecting the working information of each battery to be managed in real time;

[0007] Step 2: Obtaining the transformation information of the predicted state transition value of each battery according to the working information of each battery;

[0008] Step 3: Obtaining the battery life score of the battery according to the transformation information;

[0009] In the technical solution provided by the present application, the transformation information of the state transition value is used to predict the life score of the battery, so in practice, for a single battery, it has sufficient accuracy.

[0010] Further, step 1 comprises the following steps:

[0011] The working information includes the internal resistance, discharge direction and charge transfer amount of the battery.

[0012] The existing technology for evaluating the battery predicted state transition value is generally focused on the battery use frequency or the battery capacity, and it is difficult to describe the battery state change information. To solve the problem, the application provides the following technical solutions.

[0013] Further, step 2 includes the following steps:

[0014] Step 21: Obtain the factory data of each battery, including the battery model and the battery production batch;

[0015] Step 22: Collect the working information of each battery in the battery pack; and use the working information to construct a working distribution diagram of each battery.

[0016] The vertical coordinate of the working distribution diagram is the internal resistance, the positive direction of the horizontal coordinate is the discharge frequency, the reverse direction of the horizontal coordinate is the charge frequency, and each element in the working distribution diagram is the charge transfer amount.

[0017] Step 23: Real-time update the working distribution diagram of each battery, and obtain the battery transformation information according to the working distribution diagram.

[0018] In the technical solutions provided by the application, the working distribution diagram of the battery is continuously updated, and adaptive state analysis can be performed on each battery. The accuracy of the current life state prediction of the battery is improved.

[0019] The life of the battery is related to the voltage, the battery capacity, the charge frequency, and the like. However, in practice, the relationship is not linear, that is, the life of the battery is not linearly related to the charge frequency or the internal resistance of the battery. Using a simple linear model to describe the aging state of the battery will result in low accuracy. To solve the problem, the application provides the following technical solutions:

[0020] Further, step 23 includes the following steps:

[0021] Step 231: Construct a battery state transition model;

[0022] Step 232: Construct a state distribution model based on the battery state transition model:

[0023] Step 233: Obtain the battery transformation information according to the state distribution model.

[0024] In the existing scheme, when constructing the battery state transition model, the following two schemes are generally used to construct the aging model. The first scheme directly sets the change of the battery life as an aging model linearly related to the use degree. The second scheme sets the aging process of the battery as a multi-segment type to try to meet the change of the internal resistance of the battery and the influence on the battery life model.

[0025] The above two methods are difficult to fit the aging state of each battery in practice. In fact, the aging life curves of the same model of batteries are inconsistent due to production batches or use environment factors. The fixed setting of the aging curve has limited accuracy in predicting the life state of a single battery in actual use. To solve this problem, the application provides the following technical solutions:

[0026] Step 231 includes the following steps:

[0027] Step 2311: input the current working feature map of each battery into the matching database, and match the closest aging curve from the database based on the pre-configured twin network model;

[0028] Step 2312: based on the matched aging curve, obtain the number of aging states of the corresponding battery;

[0029] Step 2313: construct a battery state transition model based on the number of aging states of the battery.

[0030] In the technical solutions provided by the application, by comparing each battery with the information in the database, the closest aging curve can be matched. Thus, the aging trend of the battery can be roughly understood, and the success rate will be higher when the battery state transition model is constructed based thereon.

[0031] Further, the battery state transition model in step 231 is:

[0032] T(t) = T1(t) + T2(t) + … T n (t) = a1F1(t) + a2F2(t) + … a n F n (t).

[0033] T1(t) = a1F1(t);

[0034] T2(t) = a2F2(t);

[0035] T n (t) = a n F n (t);

[0036] The critical condition is: T n-1 (t) = T n (t);

[0037] Wherein, T(t) represents the state transition value T of the battery at time t, T1(t), T2(t), … T n (t) are mathematical variables corresponding to the aging states in order;

[0038] a1, a2, … a n are fitting coefficients;

[0039] F1(t), F2(t), …, F n (t) is a pre-set initial function.

[0040] Further, the step 3 comprises the following steps:

[0041] Step 31: setting an evaluation value of each transformation information;

[0042] Step 32: the feature extraction model parses the current transformation information of the battery from the transformation information, and calculates the battery life score R according to the probability distribution of the current transformation information of the battery;

[0043] R = pT i + (1-p)T i+1 ; wherein R represents the life score of the battery, p represents the probability that the battery does not transition at the current time state i, (1-p) represents the probability that the battery transitions to the next time state. T i+1 represents the evaluation value of the predicted life of the battery at the next time state. The calculation method of the evaluation value of the predicted life has been disclosed in the foregoing.

[0044] As a second aspect of the present application, in order to solve the technical problems mentioned in the background section, the present application provides a battery pack life prediction system, comprising:

[0045] a terminal device for receiving working information of each battery in each battery pack; and

[0046] a first server for executing the battery pack life prediction method;

[0047] As a third aspect of the present application, in order to solve the technical problems mentioned in the background section, the present application provides a terminal device, comprising:

[0048] one or more processors;

[0049] a memory for storing one or more computer programs,

[0050] characterized in that the one or more processors execute the one or more computer programs to implement the steps of the battery pack life prediction method.

[0051] In summary, the technical solution provided by the present application can accurately evaluate the predicted service life of the battery. BRIEF DESCRIPTION OF DRAWINGS

[0052] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The illustrative embodiments of the application, and their description, are presented to provide the best explanation of the application and its best mode of practice.

[0053] In addition, and where appropriate, the same reference numerals have been used throughout the drawings and the specification to refer to the same or like elements. It should be understood that these drawings are illustrative only and are not necessarily drawn to scale.

[0054] In the drawings:

[0055] Figure 1 A flow chart of a battery pack life prediction method;

[0056] Figure 2 A schematic diagram of a battery operating profile.

[0057] Figure 3 A schematic diagram of a battery aging curve. DETAILED DESCRIPTION

[0058] Embodiments of the present application will be described in greater detail below with reference to the accompanying drawings. While certain embodiments of the application are shown in the drawings, it should be understood that the application can be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art. It should be appreciated that the drawings are not necessarily drawn to scale.

[0059] It should also be noted that, in the drawings, like or similar elements are referred to with the same reference numerals. It should be understood that the drawings are merely schematic and that the elements and the features are not necessarily drawn to scale.

[0060] The application will be described in greater detail below with reference to the accompanying drawings.

[0061] Embodiments of the present application will be described in greater detail below with reference to the accompanying drawings. While certain embodiments of the application are shown in the drawings, it should be understood that the application can be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art. It should be appreciated that the drawings are not necessarily drawn to scale. Figure 1 A battery pack life prediction method, comprising the following steps:

[0062] Step 1: Collecting the operating information of each battery to be managed in real time.

[0063] The operating information includes the internal resistance, discharge direction and charge transfer amount of the battery.

[0064] The operating information collected by the present application includes the internal resistance, discharge direction and charge transfer amount of the battery. The internal resistance of the battery will change with the increase of the number of uses, and in practice, a minimum resistance resolution needs to be set. The discharge direction is whether the battery is in a charging state or a discharging state.

[0065] Thus, for each use of the battery, the current internal resistance of the battery and the discharge direction, as well as the amount of charge transfer, can be obtained. The amount of charge transfer is the amount of electricity discharged by the battery, or the amount of charge.

[0066] Step 2: processing the working information of each battery to obtain transformation information representing the expected state transition value of each battery.

[0067] Step 2 includes the following steps:

[0068] Step 21: obtaining the factory data of each battery, including the model of the battery and the production batch of the battery.

[0069] The battery will continue to age as the working time increases, and the characteristic information of aging is actually the change of internal resistance. The factory voltage, factory battery capacity, and factory internal resistance of each model of battery are different, so directly using the working information of the battery may not be able to represent the health status of the battery. Moreover, because the manufacturing process and the manufacturer of the battery are different, the internal resistance of the same model of battery is not necessarily the same. Therefore, the factory data of the battery needs to be obtained, and the battery is classified according to the factory data. The classification basis is the model and production batch of the battery, and the same model and production batch of the battery are classified as the same type of battery.

[0070] Step 22: collecting the working information of each battery in the battery pack; using the working information to construct a working distribution graph of each battery;

[0071] The vertical coordinate of the working distribution graph is the internal resistance, the positive direction of the horizontal coordinate is the discharge times, the reverse direction of the horizontal coordinate is the charge times, and each element in the working distribution graph is the amount of charge transfer.

[0072] As shown in Figure 2 For each battery, a working distribution graph can be generated. The working distribution graph can distinguish the use of the battery. Each element in the working distribution graph indicates a use record of the battery.

[0073] For example, the position of element (1, 1) indicates that the internal resistance of the battery is 1 during the first discharge, and the discharge amount is the specific value of the position.

[0074] In practice, therefore, the battery can be matched according to the working distribution graph. If two batteries of the same production model and the same production batch are similar on the working distribution graph, the aging process of the two batteries will be very close.

[0075] Step 23: real-time updating the working distribution graph of each battery, and obtaining the transformation information of the battery according to the working distribution graph.

[0076] In the scheme, the update work distribution diagram is actually matching the work distribution diagram of each battery currently needing to be monitored with the work distribution diagram of each battery in the historical database, to obtain the life status of each battery currently needing to be monitored.

[0077] Step 23 comprises the following steps:

[0078] Step 231: constructing a battery state transition model.

[0079] The battery state transition model is actually used to analyze the probability of the battery transitioning from the current state to the next state. To accurately construct the battery state transition model, the number of states of the battery needs to be accurately obtained. Therefore, the present application provides the following scheme:

[0080] Step 231 comprises the following steps:

[0081] Step 2311: inputting the current work feature map of each battery into the matching database, and matching the closest aging curve from the database based on the pre-configured twin network model.

[0082] Specifically, a large number of work feature maps of batteries and the aging curves of the batteries are saved in the database. Therefore, if the closest work feature map can be matched from the database, the corresponding aging curve can be obtained.

[0083] In a more specific embodiment, the following method is used:

[0084] The twin network model is constructed as follows: two convolutional neural networks (CNNs) with the same structure and weight sharing are constructed as input layers. The input layer accepts a pair of images as input.

[0085] A similarity calculation layer is constructed: used to calculate the similarity or distance between two feature vectors. Common similarity measurement methods include Euclidean distance, cosine similarity, etc.

[0086] In the present application, cosine similarity is used to evaluate the similarity between two pictures.

[0087] In the data, at least 3 work feature maps of each type of battery are saved. Then, for the battery to be predicted, the work feature map of the battery is sequentially matched with the 3 work feature maps of the corresponding type of battery, and the work feature map with the best matching result is selected.

[0088] The calculation method of the similarity of two pictures is as follows:

[0089] S1: pre-processing picture 1 and picture 2. The pre-processing mainly unifies the calculation method of the numerical value.

[0090] S2: calculating the similarity of picture 1 and picture 2. Figure 1and Figure 2 Input into the input layer, get f(x1) and f(x2), f(x1) and f(x2) are respectively Figure 1 and Figure 2 The characteristics of.

[0091] In this application, mainly extract the texture features in the input layer extraction work schematic. Further facilitate the comparison of picture 1 and picture 2 in the use process of the similarity. The specific texture feature extraction method can adopt SIFT algorithm.

[0092] S3: input f(x1) and f(x2) into the similarity calculation layer to calculate the cosine similarity of the two pictures.

[0093] If the similarity of picture 1 and picture 2 is greater than the preset threshold, it means that the two are matched. Further, the corresponding curve can be obtained.

[0094] Step 2312: based on the matched aging curve, the number of corresponding battery aging state is obtained;

[0095] Step 2313: based on the number of battery aging state, the battery state transition model is constructed.

[0096] The battery state transition model in step 231 is:

[0097] T(t) = T1(t) + T2(t) + … T n (t) = a1F1(t) + a2F2(t) + … a n F n (t).

[0098] T1(t) = a1F1(t);

[0099] T2(t) = a2F2(t);

[0100] T n (t) = a n F n (t);

[0101] The critical condition is: T n-1 (t) = T n (t);

[0102] Wherein, T(t) represents the state transition value T of the battery at t time, T1(t), T2(t), … T n (t) are respectively the mathematical variables of the corresponding order of aging state;

[0103] a1, a2, … a n are fitting coefficients;

[0104] F1(t), F2(t),..., F n (t) is a pre-set initial function.

[0105] The aging of the battery is a slow process, but as the internal resistance of the battery increases, the heat generation of the battery in operation will become more and more serious, and the aging rate of the battery will gradually increase accordingly, so during the aging process of the battery, it cannot be simply treated as a linear model. It is a gradually changing process. Therefore, in this application, according to the pre-matched aging curve, the battery state transition model is set for each battery according to its expected aging curve, so that the accuracy of the battery state transition model of each battery will be higher.

[0106] Where T(t) is actually the evaluation value of the expected life of the current battery. The specific evaluation method can refer to the battery life evaluation method in Chinese patent CN112731164B. Or directly evaluate the number of times the battery is used. That is, the value of T(t) is a fixed value that does not contain the fixed value or reference value of the battery transition to the next state.

[0107] Step 232: Construct a state distribution model based on the battery state transition model.

[0108] Reference Figure 3 In step 231, the battery aging model has been provided and fitted, and the battery aging model is actually a mathematical representation of the battery aging curve. After obtaining the battery aging model, the battery aging curve can be divided into several time states to construct a state distribution model. The specific construction method is not further described here.

[0109] Step 233: Obtain the transformation information of the battery according to the state distribution model.

[0110] In step 232, the battery aging model is divided into individual time states. The transformation information of the battery is the probability of the current state of the battery transitioning to the next time state.

[0111] Step 3: Obtain the battery life score of the battery according to the transformation information of the battery;

[0112] Step 3 includes the following steps:

[0113] Step 31: Set the evaluation value of each transformation information;

[0114] Step 32: The feature extraction model parses the current transformation information of the battery from the transformation information, and calculates the battery life score R according to the probability distribution of the current transformation information of the battery;

[0115] R=pTi + (1-p)T i+1 ; wherein R represents the life fraction of the battery, p represents the probability that the battery does not transition to the next state at the current time state i, (1-p) represents the probability that the battery transitions to the next state. T i+1 represents the evaluation value of the predicted life of the battery at the next state. The calculation method of the evaluation value of the predicted life has been disclosed above.

[0116] In the technical solutions provided in the present application, the probability of the battery transitioning to the subsequent state can be introduced when calculating the battery life fraction R. Therefore, the use of the battery can be accurately measured.

[0117] Embodiment 2: A battery pack life prediction system, comprising:

[0118] a terminal device configured to receive working information of each battery in each battery pack; and

[0119] a first server configured to execute the battery pack life prediction method;

[0120] Embodiment 3: A terminal device, comprising:

[0121] one or more processors;

[0122] a memory configured to store one or more computer programs,

[0123] characterized in that the one or more processors execute the one or more computer programs to implement the steps of the battery pack life prediction method.

[0124] The above description is merely some preferred embodiments of the present application and a description of the principles of the applied technology. Those skilled in the art should understand that the scope of the application involved in the embodiments of the present application is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or equivalent features without departing from the above inventive concept. For example, the above features can be replaced with technical features disclosed in the embodiments of the present application (but not limited to) having similar functions to form technical solutions.

Claims

1. A method for predicting the lifespan of a battery pack, characterized in that: Includes the following steps: Step 1: Collect real-time operating information of each battery to be managed; Step 2: Obtain the transformation information of the expected state transition values ​​of each battery based on the working information of each battery; Step 2 includes the following steps: Step 21: Obtain the factory data for each battery, including the battery model and production batch. Step 22: Collect the operating information of each battery in the battery pack; use the operating information to construct an operating distribution diagram of each battery; In the working distribution diagram, the vertical axis represents the internal resistance, the positive direction of the horizontal axis represents the number of discharges, the complex direction of the horizontal axis represents the number of chargings, and each element in the working distribution diagram represents the amount of charge transferred. Step 23: Update the working distribution map of each battery in real time, and obtain the battery transformation information based on the working distribution map; Step 23 includes the following steps: Step 231: Construct a battery state transition model; Step 232: Construct a state distribution model based on the battery state transition model; Step 233: Obtain the battery transformation information based on the state distribution model; Step 231 includes the following steps: Step 2311: Input the current working feature map of each battery into the matching database. The matching database is based on a pre-configured Siamese network model and matches the closest aging curve from the database. The Siamese network model is constructed as follows: Construct two convolutional neural networks (CNNs) with the same structure and shared weights as input layers. Step 2312: Based on the matched aging curve, obtain the number of aging states of the corresponding battery; Step 2313: Construct a battery state transition model based on the number of battery aging states; The battery state transition model in step 231 is as follows: T(t)=T1(t)+T2(t)+……T n (t)=a1F1(t)+a2F2(t)++……a n F n (t); T1(t) = a1F1(t); T2(t) = a2F2(t); T n (t)=a n F n (t); The critical condition is: T n-1 (t) = T n (t); Where T(t) represents the state transition value T of the battery at time t, T1(t), T2(t), ..., T n (t) represents the mathematical variables for the aging states in the corresponding order; a1、a2……a n These are the fitting coefficients; F1(t), F2(t), ..., F n (t) is a pre-set initial function; Step 3: Obtain the battery life score based on the transformation information; Wherein, T(t) is actually the evaluation value of the current battery's expected lifespan; The battery transformation information is the probability that the battery's current state will transition to the next time state.

2. The battery pack life prediction method according to claim 1, characterized in that: Step 1 includes the following steps: The operating information includes the battery's internal resistance, discharge direction, and charge transfer amount.

3. The battery pack life prediction method according to claim 1, characterized in that: Step 3 includes the following steps: Step 31: Set the evaluation values ​​for each transformation information; Step 32: The feature extraction model parses the current transformation information of the battery from the transformation information and calculates the battery life score R based on the probability distribution of the current transformation information of the battery. R=pT i +(1-p)T i+1 Where R represents the battery's lifespan fraction, p represents the probability that the battery will not transition in the current time i, and (1-p) represents the probability that the battery will transition to the next time i; T i+1 This represents the estimated lifespan of the battery in the next time period.

4. A battery pack life prediction system, characterized in that, include: Terminal equipment is used to receive operating information from each battery in each battery pack. as well as The first server executes the method according to any one of claims 1 to 3.

5. A terminal device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 3.

Images