Battery degradation prediction system, battery degradation prediction method, and battery degradation prediction program

Abstract

This allows for adjustment of adaptability and robustness in battery degradation prediction. [Solution] A battery degradation prediction system using a prediction model, comprising: a reliability calculation unit 12 that calculates the prediction reliability using the difference between the predicted SOH value of the target battery calculated using the prediction model and the measured SOH value; and a model adjustment unit 14 that adjusts the prediction model when the prediction reliability exceeds a predetermined threshold. This adjusts the model being used by providing a quantitative prediction reliability to the output.

Description

Technical Field

[0001] The present disclosure relates to a battery degradation prediction system, a battery degradation prediction method, and a battery degradation prediction program.

Background Art

[0002] In recent years, technologies for performing battery degradation diagnosis and degradation prediction by inputting the physical or chemical degradation characteristics of a battery and usage history such as elapsed time, amount of electricity charged, state of charge (SOC), and temperature have been used.

[0003] Patent Document 1 discloses constructing an objective function by combining a model that separates the aging part and the energized part of a battery with a calculation model such as a root rule, and using a solver or the like to create tables of the discharge coefficient ha(T, S) and the energization coefficient ac(T, S) with T being the temperature and S being the SOC, and performing battery degradation prediction based on the tables.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in related existing technologies, it is not possible to adjust the fitness and robustness of the prediction output based on the physical and chemical degradation characteristics of the battery. Therefore, there is a risk that the model may have low fitness for individuals even if it has robustness, or that the model may have strong fitness for specific individuals but weak robustness and low generalization performance.

[0006] Furthermore, there is a known technique that converts time-series data such as battery voltage or SOC, temperature, and current into intermediate data, allowing the input of data as latent features without explicitly revealing its physical and chemical degradation characteristics. However, this method cannot guarantee physical and chemical degradation characteristics, making it impossible to calculate a reasonable predictive confidence level or provide feedback. Therefore, the risk of creating models with poor individual fit or poor generalization performance cannot be eliminated.

[0007] Furthermore, techniques for correcting prediction formulas by evaluating the error between predicted and actual measured data are known. However, this correction method cannot reflect the adjustment of the contributions of physical and chemical degradation characteristics in the strength of the correction, and therefore cannot eliminate the risk of resulting in models with poor individual fit or poor generalization performance.

[0008] This disclosure provides a battery degradation prediction system, a battery degradation prediction method, and a battery degradation prediction program that enable adjustment of fitness and robustness. [Means for solving the problem]

[0009] The battery degradation prediction system according to this disclosure is a battery degradation prediction system using a prediction model, comprising: a reliability calculation unit that calculates prediction reliability using the difference between a predicted SOH value of a target battery calculated using the prediction model and an actual SOH value; and a model adjustment unit that adjusts the prediction model when the prediction reliability exceeds a predetermined threshold. This allows for adjustment of the model being used by providing a quantitative predictive confidence level to the output.

[0010] Furthermore, the battery degradation prediction method according to this disclosure is a battery degradation prediction method using a prediction model, comprising the steps of: calculating a prediction reliability using the difference between a predicted SOH value of a target battery calculated using the prediction model and an actual measured SOH value; and performing an adjustment of the prediction model when the prediction reliability exceeds a predetermined threshold. This allows for adjustment of the model being used by providing a quantitative predictive confidence level to the output.

[0011] Furthermore, the battery degradation prediction program according to this disclosure is a battery degradation prediction program using a prediction model, and includes the steps of: calculating a prediction reliability using the difference between a predicted SOH value of a target battery calculated using the prediction model and an actual measured SOH value; and performing an adjustment of the prediction model when the prediction reliability exceeds a predetermined threshold. This allows for adjustment of the model being used by providing a quantitative predictive confidence level to the output. [Effects of the Invention]

[0012] This disclosure provides a battery degradation prediction system, a battery prediction method, and a battery program that enable adjustment of fitness and robustness. [Brief explanation of the drawing]

[0013] [Figure 1] This is a block diagram showing the configuration of the degradation prediction system according to Embodiment 1. [Figure 2] This is a flowchart illustrating the operation of the degradation prediction system according to Embodiment 1. [Figure 3] This figure shows an example of a prediction model f for SOH at a future time t according to Embodiment 1. [Figure 4] This diagram shows the overall calculation flow of the prediction confidence score according to Embodiment 1. [Figure 5] This figure shows an example of a predictive reliability system according to Embodiment 1. [Figure 6] It is a diagram showing another example of the prediction reliability system according to Embodiment 1.

Mode for Carrying Out the Invention

[0014] Embodiment 1 Hereinafter, a battery degradation prediction system according to this embodiment will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration example of the degradation prediction system 1. The degradation prediction system 1 includes a SOH (State of Health) prediction unit 11, a reliability calculation unit 12, an adjustment determination unit 13, a model adjustment unit 14, and an additional prediction execution determination unit 15. Note that the battery is a secondary battery mounted on a vehicle, and typically, it will be described that a plurality of batteries are provided in the vehicle.

[0015] The SOH prediction unit 11 predicts a value obtained by evaluating the state of the battery in comparison with an ideal state. In other words, the SOH prediction unit 11 can predict the soundness and degradation state of the battery.

[0016] More specifically, for an arbitrary future time t, the SOH prediction unit 11 takes as input the SOH history x = [x(0),..., x(t')] of the target battery that satisfies t' < t. Then, the SOH prediction unit 11 predicts the future SOH y(t) at time t as y(t) = f(x|θ) using a future SOH prediction model f. Here, θ is a parameter of the model f.

[0017] The reliability calculation unit 12 calculates the prediction reliability C(t) of the output y(t) of the SOH prediction model output by the SOH prediction unit 11 by the calculation g(x, f) of the reliability calculation model g.

[0018] The adjustment determination unit 13 determines whether to adjust the SOH prediction model f used in the SOH prediction unit 11 based on the result of the prediction reliability C(t) calculated by the reliability calculation unit 12.

[0019] The model adjustment unit 14 adjusts the balance between the adaptability and robustness of the SOH model f by changing the parameter θ.

[0020] The additional prediction execution determination unit 15 determines whether to further execute a prediction using the new input x'.

[0021] Next, referring to FIG. 2, the operation flow of the deterioration prediction system 1 will be described.

[0022] The SOH prediction unit 11 predicts a value obtained by evaluating the state of the battery by comparing it with an ideal state (step S1).

[0023] The reliability calculation unit 12 calculates the prediction reliability C(t) of the output y(t) of the SOH prediction model by the calculation g(x,f) of the reliability calculation model g (step S2).

[0024] The adjustment determination unit 13 determines whether to perform model adjustment of the SOH prediction model f based on the result of the prediction reliability C(t) calculated by the reliability calculation unit 12 (step S3). If the adjustment determination unit 13 determines to perform adjustment (Yes in step S3), it proceeds to step S4. If the adjustment determination unit 13 determines not to perform adjustment (No in step S3), it proceeds to step S5.

[0025] The model adjustment unit 14 adjusts the SOH model f (step S4). Then, it proceeds to step S5.

[0026] The additional prediction execution determination unit 15 determines whether to execute an additional prediction using the new input x' (step S5). If it determines to perform an additional prediction (Yes in step S5), it returns to step S1. If it determines not to perform an additional prediction (No in step S5), the process ends.

[0027] Next, a detailed operation example in each step shown in FIG. 2 will be described. First, regarding the SOH prediction unit 11 calculating the future SOH y(t) at time t shown in step S1, it will be described.

[0028] The SOH prediction unit 11 predicts the future SOH of the target battery using a secondary battery SOH prediction model based on the physical and chemical characteristics of battery degradation. FIG. 3 is a diagram showing an example of the SOH prediction model f at a future time t.

[0029] As described above, the SOH prediction unit 11 takes as input the SOH history x = [x(0),..., x(t')] of the target battery that satisfies t' < t for any future time t. Then, the SOH prediction unit 11 uses the prediction model f to calculate the future SOH y(t) at time t by y(t) = f(x|θ). Here, θ is a parameter of the model f. At this time, the SOH prediction unit 11 can include battery accessory information, which is information on the accessory status of the battery that affects SOH degradation, as information other than the time series of the actual values of the SOH itself in the SOH history x.

[0030] Here, the battery accessory information includes, but is not limited to, the characteristics of the battery, the characteristics of the vehicle on which the battery was mounted, the history of the driving environment and vehicle state of the vehicle on which the battery was mounted, and the history of the driving and charging operations of the vehicle on which the battery was mounted.

[0031] For example, the characteristics of the battery include the full charge capacity of the battery, the materials of the positive electrode, negative electrode, and electrolyte, and the manufacturer. The characteristics of the vehicle on which the battery was mounted include the vehicle model, performance such as average fuel consumption, and the weight of the vehicle body. The history of the driving environment and vehicle state of the vehicle on which the battery was mounted includes the driving area, the SOC during driving, and the frequency distribution of the battery temperature. The history of the driving and charging operations of the vehicle on which the battery was mounted includes the number of sudden accelerations and decelerations and the usage ratio of rapid charging and normal charging.

[0032] In addition, the SOH prediction model f can use an SOH prediction model such as the degradation behavior of the electrode according to the Arrhenius equation or the root law of degradation over time. This is a known technique that enables SOH prediction based on the physical and chemical characteristics of battery degradation.

[0033] The parameter θ used in model f may be one that has been calculated in advance, or one that has been newly calculated using the SOH history x.

[0034] Next, we will explain an example of the calculation of the predicted reliability by the reliability calculation unit 12, as shown in step S2. Here, we will explain both cases in the degradation prediction system 1 where the SOH history x=[x(0),…,x(t')] can be obtained for multiple batteries and where it cannot.

[0035] First, we will explain the case in the degradation prediction system 1 where SOH history cannot be obtained for multiple batteries, that is, the case where SOH history can only be obtained for a single battery.

[0036] First, the confidence calculation unit 12 calculates the input to the confidence system. Figure 4 shows an example of the overall calculation flow of the prediction confidence. As shown in Figure 4, the confidence calculation unit 12 considers the past SOH history x = [x(0), …, x(t')] as sequential inputs in order from the oldest time, and calculates the prediction output y_0(t), …, y_t'(t) of the prediction system for each time point using x up to a certain time point, such as [x(0)], [x(0, x(1)], [x(0), x(1), x(2)], …, [x(0), …, x(t')], and the difference y_0(t) - x(t), …, y_t'(t) - x(t) of the model's prediction output for each time point.

[0037] Next, the confidence calculation unit 12 calculates the prediction confidence using the confidence system. Figure 5 shows an example of the prediction confidence system. As shown in Figure 5, the confidence calculation unit 12 uses the discrepancies in the model's prediction output at each time point y_0(t) - x(t), ..., y_t'(t) - x(t) to calculate, using regression or other methods, how much of a prediction discrepancy can be expected between the model creation time and the future time to be predicted, and provides this as the prediction confidence C(t) for the final output y(t) of the future SOH prediction.

[0038] Next, we will explain the case where the SOH history x = [x(0), …, x(t')] can be obtained for multiple batteries in the degradation prediction system 1.

[0039] First, the reliability calculation unit 12 calculates the input to the reliability system. Specifically, as described above, the reliability calculation unit 12 calculates the deviation of the predicted output of each time model for reliability for one battery, and then performs this calculation for each of the multiple batteries.

[0040] In other words, the reliability calculation unit 12 calculates the predicted output y_i = [y_0(t),…, y_t'(t)] for multiple batteries and the difference in the model's predicted output at each time point e_i = [y_0(t) - x(t),…, y_t'(t) - x(t)] for battery i. Here, if there is SOH history data for n batteries, i = 0,…,n-1.

[0041] Next, the confidence calculation unit 12 uses the n calculated prediction outputs y_i = [y_0(t),…, y_t'(t)] of the prediction system and the difference in the model's prediction output at each time point e_i = [y_0(t) - x(t),…, y_t'(t) - x(t)] to calculate the prediction confidence. Here, Figure 6 shows another example of the prediction confidence system. As shown in Figures 4 and 6, the confidence calculation unit 12 takes n pairs of (y_i, e_i) as input and calculates how much prediction deviation is expected between the model creation time and the future time to be predicted using methods such as regression. The confidence calculation unit 12 then provides the prediction confidence C_i(t) for the final output y_i(t) of the future SOH prediction.

[0042] As a result, the reliability calculation unit 12 can calculate the prediction reliability using the difference between the predicted SOH value of the target battery calculated using the prediction model and the measured SOH value.

[0043] Next, we will explain an example of the model adjustment execution determination of the adjustment determination unit 13 shown in step S3.

[0044] The adjustment determination unit 13 decides whether or not to perform model adjustment based on the predicted confidence level C(t) output by the confidence level calculation unit 12.

[0045] For example, in the process of step S2 described above, the adjustment determination unit 13 determines whether to perform an adjustment to the prediction model if C_i(t) exceeds a pre-set threshold for y(t), i.e., if the result is bad, regardless of whether there is one battery or multiple batteries for which the SOH history is acquired.

[0046] For example, in the process of step S2 described above, if the SOH history can be obtained for multiple batteries, the adjustment determination unit 13 can determine to make adjustments to the battery prediction model if the C_i(t) is relatively large compared to other batteries, i.e., relatively bad. In other words, the adjustment determination unit 13 can determine whether the target battery exceeds the C_i(t) value of other batteries, using the C_i(t) value of other batteries as a threshold.

[0047] Based on the above, the adjustment determination unit 13 can change the calculation of whether the prediction reliability exceeds a predetermined threshold depending on whether the measured SOH of the battery was obtained from multiple batteries.

[0048] Next, we will explain the model adjustment performed by the model adjustment unit 14 as shown in step S4.

[0049] The model adjustment unit 14 adjusts the balance between the adaptability and robustness of model f by changing the parameter θ. Here, the model adjustment unit 14 adjusts the trade-off between the adaptability and robustness of the prediction model using physical domain knowledge of battery degradation, such as degradation modes and root laws. Here, the current time is t_{now}, the future time for which we want to predict SOH is t_{future}, and the following feedback of prediction confidence C(t) can be given as an example.

[0050] The model adjustment unit 14 modifies the weights of the training data during parameter learning according to the battery degradation mode (initial degradation, intermediate degradation, late degradation) determined by t_{now} for each battery. For example, if the model adjustment unit 14 wants to improve fitness, it increases the weight of battery data with a degradation mode close to the target (closer t_{now}) during parameter learning. Alternatively, if the model adjustment unit 14 wants to improve robustness, it decreases the weight of battery data with a degradation mode close to the target (closer t_{now}).

[0051] Furthermore, the model adjustment unit 14 can improve robustness during parameter learning by fitting the prediction model to a general degradation curve that does not depend on individual batteries, such as a root rule, for the range of t_{future} in which the prediction confidence C(t) is large.

[0052] This allows the model adjustment unit 14 to adjust the prediction model when the adjustment determination unit 13 determines that the prediction confidence level exceeds a predetermined threshold.

[0053] Next, we will explain an example of the execution determination of an additional prediction by the additional prediction execution determination unit 15, as shown in step S5.

[0054] The additional prediction execution determination unit 15 determines, for example, that an additional prediction should be executed if there is a new input x', as shown below.

[0055] The additional prediction execution determination unit 15 determines whether to perform an additional prediction if the latest time data is added to the original input x = [x(0), …, x(t')], resulting in x' = [x(0), …, x(t'), x(t'+1)]. This assumes a situation where data accumulates over time.

[0056] Furthermore, the additional prediction execution determination unit 15 determines whether to perform additional predictions if a new feature z is added to the original input x = [x(0), …, x(t')], resulting in x' = [x(0), z(0), …, x(t'), z(t')]. This is intended to account for cases where information from another database is added.

[0057] Furthermore, the additional prediction execution determination unit 15 determines whether to perform an additional prediction if the original input x = [x(0), …, x(t')] is combined with the SOH history of a new battery to become x' = [x, x_add]. This is intended to handle cases such as when data for a later battery is added.

[0058] Furthermore, the additional prediction execution determination unit 15 may also determine that additional predictions should be executed even if the original input x = [x(0), …, x(t')] is simply set to x' = x. This allows the adjusted output to be obtained using the adjusted model, while the input remains unchanged.

[0059] In this way, the additional prediction execution determination unit 15 can determine whether or not to perform additional predictions if the prediction confidence level does not exceed a predetermined threshold in the determination of the adjustment determination unit 13, or if the prediction confidence level exceeds a predetermined threshold in the determination of the adjustment determination unit 13 and the model adjustment unit 14 has adjusted the prediction model.

[0060] Based on the above, the degradation prediction system 1 can calculate the prediction confidence level based on the SOH history and adjust the model according to this prediction confidence level. Furthermore, the degradation prediction system 1 can adjust the balance between the adaptability and robustness of the model by changing the parameters used in the model.

[0061] It should be noted that the present invention is not limited to the embodiments described above, and can be modified as appropriate without departing from the spirit of the invention. In other words, the above description has been omitted and simplified as appropriate for the sake of clarity, and those skilled in the art can easily change, add, and modify each element of the embodiments within the scope of the present invention.

[0062] Embodiments of the present disclosure may be implemented in hardware or in dedicated circuitry, software, logic, or any combination thereof. Some embodiments may be implemented in hardware, while others may be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device.

[0063] This disclosure also provides at least one computer program product tangibly stored on a non-temporary computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions contained in a program module, and is executed on a device on a target real or virtual processor to perform the processes or methods of this disclosure. The program module includes routines, programs, libraries, objects, classes, components, data structures, etc., that perform a specific task or implement a specific abstract data type. The functionality of the program module may be combined or divided among the program module as desired in various embodiments. The machine-executable instructions of the program module can be executed on a local or distributed device. On a distributed device, the program module can reside on both local and remote storage media.

[0064] Program code for performing the methods of this disclosure may be written in any combination of one or more programming languages. These program codes are provided to a processor or controller of a general-purpose computer, a dedicated computer, or other programmable data processing device. When the program code is executed by the processor or controller, the functions / operations in the flowchart and / or block diagrams it implements are performed. The program code may run entirely on a machine, partially on a machine, partially as a standalone software package, partially on a machine, partially on a remote machine, or entirely on a remote machine or server.

[0065] Programs can be stored and supplied to a computer using various types of non-temporary computer-readable media. Non-temporary computer-readable media include various types of tangible recording media. Examples of non-temporary computer-readable media include magnetic recording media, magneto-optical recording media, optical disc media, and semiconductor memory. Magnetic recording media include, for example, flexible disks, magnetic tapes, and hard disk drives. Magneto-optical recording media include, for example, magneto-optical disks. Optical disc media include, for example, Blu-ray discs, CD (Compact Disc)-ROM (Read Only Memory), CD-R (Recordable), and CD-RW (ReWritable). Semiconductor memory includes, for example, solid-state drives, mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (random access memory). Programs may also be supplied to a computer using various types of temporary computer-readable media. Examples of temporary computer-readable media include electrical signals, optical signals, and electromagnetic waves. Temporary computer-readable media can supply programs to a computer via wired communication channels such as electric wires and optical fibers, or via wireless communication channels. [Explanation of Symbols]

[0066] 1. Degradation prediction system 11 SOH Prediction Department 12. Reliability Calculation Unit 13 Adjustment judgment section 14 Model Adjustment Section 15 Additional prediction execution determination unit

Claims

1. A battery degradation prediction system using a predictive model, A reliability calculation unit calculates the prediction reliability using the difference between the predicted SOH value of the target battery calculated using a prediction model and the measured SOH value. The system includes a model adjustment unit that adjusts the prediction model when the prediction confidence exceeds a predetermined threshold. Battery degradation prediction system.

2. The system further includes an additional prediction execution determination unit that determines whether or not to perform additional predictions if the prediction confidence does not exceed a predetermined threshold, or if the prediction confidence exceeds a predetermined threshold and the prediction model has been adjusted. A battery degradation prediction system according to claim 1.

3. The system includes an adjustment determination unit that calculates whether the prediction reliability exceeds a predetermined threshold, The adjustment determination unit is, The calculation of whether the prediction reliability exceeds a predetermined threshold can be modified depending on whether the measured SOH of the battery was obtained from multiple batteries. A battery degradation prediction system according to claim 1 or claim 2.

4. A method for predicting battery degradation using a predictive model, The steps include: calculating the prediction confidence using the difference between the predicted SOH value of the target battery calculated using a prediction model and the measured SOH value; The steps include: performing adjustments to the prediction model when the prediction confidence exceeds a predetermined threshold; Methods for predicting battery degradation.

5. A battery degradation prediction program using a predictive model, The steps include: calculating the prediction confidence using the difference between the predicted SOH value of the target battery calculated using a prediction model and the measured SOH value; The steps include: performing adjustments to the prediction model when the prediction confidence exceeds a predetermined threshold; Battery degradation prediction program.

Images