Battery degradation estimation method

Abstract

This invention provides a battery degradation estimation method that can improve estimation accuracy while ensuring the performance of models based on physical and chemical degradation characteristics. [Solution] The battery degradation estimation method according to the present disclosure is a method for estimating the degradation of a battery that is mounted on a vehicle and supplies power to a motor, and comprises: a processing step of weighting pre-stored battery degradation performance data according to its similarity to supplementary information of the battery to be estimated; a generation step of generating a trained model that takes battery usage history information as input and outputs battery degradation level by using the weighted degradation performance data from the processing step as training data; and a degradation estimation step of estimating the degradation level of the battery to be estimated using the trained model based on the usage history information of the battery to be estimated.

Description

Technical Field

[0001] The present disclosure relates to a method for estimating battery degradation.

Background Art

[0002] Patent Document 1 discloses a degradation degree (SOH: State of Health) estimation model for a secondary battery based on physical and chemical characteristics of battery degradation, such as the degradation behavior of an electrode following the Arrhenius equation and the root law of aging over time.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The secondary battery degradation degree estimation model disclosed in Patent Document 1 takes as input the usage history information of the secondary battery, such as the elapsed time, the amount of electricity conducted, and the state of charge (SOC) of the secondary battery, and outputs the degradation degree of the secondary battery. The inventors have found that by additionally using parameters such as the exposure environment of the battery (secondary battery) and the driving operation of the vehicle on which the battery is mounted, the estimation accuracy of the degradation degree estimation model can be improved. However, when parameters are directly added to the secondary battery degradation degree estimation model disclosed in Patent Document 1, there is a risk that the performance of the model for estimating the degradation degree based on physical and chemical degradation characteristics cannot be guaranteed.

[0005] In view of the above circumstances, the present disclosure provides a battery degradation estimation method capable of improving the estimation accuracy while ensuring the performance of the model based on physical and chemical degradation characteristics.

Means for Solving the Problems

[0006] A battery degradation estimation method according to one aspect of this disclosure is: A method for estimating the degradation of a battery mounted on a vehicle and supplying power to a motor, A processing step that weights pre-stored battery degradation data according to its similarity to the supplementary information of the battery to be estimated, A generation step is performed to generate a trained model that takes battery usage history information as input and outputs battery degradation level by using the weighted degradation performance data from the processing step as training data for machine learning, The system includes a degradation estimation step of estimating the degree of degradation of the target battery using the trained model based on the usage history information of the target battery. [Effects of the Invention]

[0007] This disclosure provides a battery degradation estimation method that can improve estimation accuracy while ensuring the performance of a model based on physical and chemical degradation characteristics. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing the configuration of a battery degradation estimation system according to an embodiment of the present disclosure. [Figure 2] (a) A graph of the State of Health (SOH) against time for a battery according to an embodiment of this disclosure. (b) Time-series data of the SOH for a battery according to an embodiment of this disclosure. (c) An example of supplementary information for a battery according to an embodiment of this disclosure. (d) An example of supplementary information for a battery according to an embodiment of this disclosure. [Figure 3] (a) A flowchart of the input and output of the SOH estimation model according to the embodiment of this disclosure. (b) A comparison diagram of the SOH estimation model according to the embodiment of this disclosure with an existing SOH estimation model and an SOH estimation model that is a more complex version of the existing model. [Figure 4](a) A diagram showing weighting based on dimensionality reduction according to the similarity of the battery's accompanying information according to the embodiment of this disclosure. (b) A diagram showing weighting based on a rule-based system according to the similarity of the battery's accompanying information according to the embodiment of this disclosure. [Figure 5] This is a flowchart of a battery degradation estimation method according to an embodiment of the present disclosure. [Figure 6] This figure shows a method for updating the degradation estimation model according to the embodiment of this disclosure. [Modes for carrying out the invention]

[0009] The following describes specific embodiments of this disclosure in detail with reference to the drawings. However, this disclosure is not limited to the following embodiments. Also, for clarity, the following descriptions and drawings have been simplified as appropriate.

[0010] <Configuration of the battery degradation estimation system> Figure 1 is a block diagram showing the configuration of the battery degradation estimation system according to an embodiment of this disclosure. The battery degradation estimation system S consists of a vehicle C and a battery degradation estimation device 5. Vehicle C includes a battery 1, a battery sensor 2, an operating unit 3, and an operating history acquisition unit 4. The battery sensor 2 includes a usage history acquisition unit 21, a usage history recording unit 22, an ancillary information acquisition unit 23, and an ancillary information recording unit 24. The battery degradation estimation device 5 includes a performance data storage unit 51, a model generation unit 52, and a degradation estimation unit 53. Battery 1 is the battery to be estimated.

[0011] Vehicle C is a car that can move by driving a motor (not shown) in the operating unit 3 using power supplied from the battery 1 to the operating unit 3. Vehicle C is, for example, an electric vehicle, but may also be a hybrid vehicle with an external charging function.

[0012] The battery 1 is connected to the usage history acquisition unit 21 and the attached information acquisition unit 23. The battery 1 is a power storage device mounted on the vehicle C and having a function of supplying power to a drive unit that drives the vehicle C. The battery 1 is a secondary battery such as a lithium-ion battery, a lead-acid battery, or a nickel-metal hydride battery, and is configured by housing a positive electrode active material layer, a negative electrode active material layer, a current collector, a separator, an electrolyte, etc. inside a sealing body. The battery 1 supplies power to the motor by connecting to the motor that drives the vehicle C. The size of the battery 1 and the type of secondary battery used for the battery 1 are appropriately determined according to the size and use of the vehicle C on which the battery 1 is mounted.

[0013] The battery sensor 2 is connected to the battery 1, the operation history acquisition unit 4, and the model generation unit 52. As shown in, for example, Fig. 2(a), the battery sensor 2 acquires the value of the SOH (State of Health) of the battery 1 at a predetermined timing, and transmits the acquired result to the model generation unit 52. Here, the SOH is the ratio of the fully charged capacity of the battery 1 at the time of deterioration to the fully charged capacity of the battery 1 at the initial stage, and

Equation

[0014] The usage history acquisition unit 21 is connected to the battery 1 and the usage history recording unit 22. The usage history acquisition unit 21 acquires the State of Health (SOH) value of the battery 1 at predetermined timings and transmits the acquired data to the usage history recording unit 22. The usage history acquisition unit 21 is equipped with sensors such as a current sensor or a voltage sensor and acquires current or voltage values ​​at predetermined timings. Based on the acquired current or voltage values, the usage history acquisition unit 21 calculates the SOH value of the battery 1 by determining, for example, the amount of current supplied from an empty state to a fully charged state of the battery 1. In order to calculate the SOH value of the battery 1, the usage history acquisition unit 21 may be composed of a CPU (Central Processing Unit), an MPU (Micro Processing Unit), working memory, and a non-volatile storage device that stores a control program.

[0015] The usage history recording unit 22 is connected to the usage history acquisition unit 21 and the degradation estimation unit 53. The usage history recording unit 22 records the State of Health (SOH) value of battery 1 received from the usage history acquisition unit 21 and transmits the recorded SOH value of battery 1 to the degradation estimation unit 53. However, the usage history recording unit 22 may also record information regarding parameters used to determine the SOH of battery 1, such as the energizing time and energizing amount of battery 1. In other words, the usage history information of battery 1 refers to parameters used to determine the SOH of battery 1, such as the SOH value of battery 1 at a predetermined timing, the energizing time, and the energizing amount.

[0016] In the usage history recording unit 22, the SOH value of the battery 1 is recorded in the form of time-series data as, for example, the battery ID assigned to each battery, the date when the SOH was recorded, and the recorded SOH value, as shown in FIG. 2(b). The usage history recording unit 22 includes a storage device that can hold various types of data, and does not necessarily have to be part of the battery sensor 2. It may be an external storage device or a cloud storage connected to the usage history acquisition unit 21 via a network. Also, the usage history recording unit 22 includes a communication interface that can communicate with the degradation estimation unit 53 via wired communication means, wireless communication means, or the like.

[0017] The additional information acquisition unit 23 is connected to the battery 1 and the additional information recording unit 24. The additional information acquisition unit 23 acquires the additional information of the battery 1 and transmits the acquired additional information of the battery 1 to the additional information recording unit 24.

[0018] Here, examples of the additional information of the battery 1 include Characteristic information of the battery 1, such as the initial full charge capacity of the battery 1, the cathode material, the anode material, the materials constituting the electrolyte, and the manufacturing manufacturer; Characteristic information of the vehicle C on which the battery 1 is mounted, such as the type of the vehicle C, the driving performance, and the weight of the vehicle body; History information of the driving environment or vehicle state of the vehicle C on which the battery 1 is mounted, such as the driving area of the vehicle C, the SOC of the battery 1 during driving, the battery temperature, and the environmental temperature; History information of the driving operation and charging operation of the vehicle C on which the battery 1 is mounted, such as the number of sudden accelerations and decelerations, and the usage ratio of rapid charging and normal charging; and so on. Here, the characteristic information of the vehicle C is the vehicle information of the vehicle C. Also, the history information of the driving environment and vehicle state of the vehicle C and the history information of the driving operation and charging operation of the vehicle C are the driving history information of the vehicle C. Also, the additional information acquisition unit 23 acquires, as the additional information of the battery 1, information that can be directly acquired from the battery 1, such as the SOC of the battery 1 during driving, the battery temperature, and the usage ratio of rapid charging and normal charging.

[0019] The ancillary information acquisition unit 23 includes sensors such as a current sensor or a temperature sensor. Furthermore, in order to acquire ancillary information of the battery 1 based on data such as current values ​​or temperature acquired by the sensors, the ancillary information acquisition unit 23 may be composed of a CPU, an MPU, working memory, and a non-volatile storage device that stores a control program.

[0020] The ancillary information recording unit 24 is connected to the ancillary information acquisition unit 23, the operation history acquisition unit 4, and the model generation unit 52. The ancillary information recording unit 24 records the ancillary information of battery 1 transmitted from the ancillary information acquisition unit 23 and the operation history acquisition unit 4, and transmits the recorded ancillary information of battery 1 to the model generation unit 52.

[0021] The ancillary information recording unit 24 is a storage device capable of holding various types of data, and does not necessarily have to be part of the battery sensor 2. It may be an external storage device or cloud storage connected to the usage history acquisition unit 21 via a network. The ancillary information recording unit 24 also has a communication interface that allows it to communicate with the degradation estimation unit 53 via wired communication means or wireless communication means. The usage history recording unit 22 and the ancillary information recording unit 24 may be the same storage device. Furthermore, ancillary information of the battery 1, such as characteristic information of the battery 1 and characteristic information of the vehicle C on which the battery 1 is installed, may be pre-recorded in the ancillary information recording unit 24, or it may be recorded by being transmitted from the ancillary information acquisition unit 23 and the operation history acquisition unit 4.

[0022] The operating unit 3 is connected to the operation history acquisition unit 4. The operating unit 3 consists of equipment that allows vehicle C to function as a vehicle, such as a motor, brakes, steering wheel, safety devices, and a car navigation system.

[0023] The operation history acquisition unit 4 is connected to the operation unit 3 and the ancillary information recording unit 24. The operation history acquisition unit 4 acquires the operation history of the operation unit 3 and transmits the obtained data to the ancillary information recording unit 24 as ancillary information for the battery 1. The operation history acquisition unit 4 is equipped with sensors such as a speed sensor, a motor rotation speed sensor, and GPS (Global Positioning System), and acquires information that can be directly obtained from the operation unit 3, such as the driving area of ​​vehicle C and the number of sudden accelerations and decelerations, as ancillary information for the battery 1.

[0024] The battery degradation estimation device 5 is connected to the usage history recording unit 22 and the ancillary information recording unit 24. The battery degradation estimation device 5 estimates the State of Health (SOH) based on the usage history information and ancillary information of battery 1, and the actual value data and test data of numerous batteries stored in the actual data storage unit 51. Here, an SOH estimation model is used that enables SOH estimation based on the physical and chemical characteristics of battery degradation. The SOH estimation model will be described later.

[0025] The performance data storage unit 51 is connected to the model generation unit 52. The performance data storage unit 51 stores performance data of used batteries used in the past, and test data of batteries that have been tested, as performance degradation data. Here, the performance degradation data includes battery usage history information and related information. The performance degradation data is training data required by the model generation unit 52 to generate a SOH estimation model, and is transmitted to the model generation unit 52 as needed. The performance data storage unit 51 is equipped with a storage device that can hold various types of data, and does not necessarily have to be part of the battery degradation estimation device 5; it may be an external storage device or cloud storage connected to the model generation unit 52 via a network.

[0026] In the performance data storage unit 51, the battery's associated information included in the degradation performance data is recorded by linking the battery ID with characteristic information of the battery and vehicle C, such as the manufacturer, positive electrode material, negative electrode material, and the vehicle in which the battery is installed, as shown in Figure 2(c). In addition, the battery's associated information included in the degradation performance data is recorded by linking the battery ID with historical information of vehicle C, such as the number of times the vehicle has made sudden acceleration and the percentage of time rapid charging is used, as shown in Figure 2(d).

[0027] The model generation unit 52 is connected to the ancillary information recording unit 24, the actual data storage unit 51, and the degradation estimation unit 53. The model generation unit 52 generates a State of Health (SOH) estimation model based on the ancillary information and actual degradation data of the battery 1, and transmits the generated model to the degradation estimation unit 53. The model generation unit 52 is composed of, for example, a CPU, an MPU, working memory, and a non-volatile storage device that stores a control program. Furthermore, the model generation unit 52 does not necessarily have to be part of the battery degradation estimation device 5; it may be located in an external cloud computing environment.

[0028] Here, the SOH estimation model f generated by the model generation unit 52 is, for example, a square root law model relating to the capacity retention rate y of the battery 1.

number

number

[0029] Figure 3(a) is a flowchart of the input and output of the SOH estimation model according to the embodiment of this disclosure. In the SOH estimation model f, the parameter θ is estimated by executing a parameter learning process g to learn the parameter θ, using the degradation performance data z as training data. The SOH estimation model f generated by the estimation of the parameter θ is a trained model that takes the usage history information x_t of battery 1 as input and outputs the SOH value y_t of battery 1. Here, the inventors have found that by weighting the performance data of each battery included in the degradation performance data according to its similarity to the supplementary information x_history of battery 1, degradation performance data z' that matches the supplementary information x_history of battery 1 can be generated. By using the degradation performance data z' as training data when performing machine learning, it is possible to improve the estimation accuracy of the parameter θ and improve the performance of the SOH estimation model f.

[0030] Figure 3(b) shows a comparison of the SOH estimation model according to the embodiment of this disclosure with an existing SOH estimation model and an SOH estimation model that is a more complex version of the existing model. The SOH estimation model f according to the embodiment of this disclosure has been improved by changing the model structure and parameter learning process g, and by using actual degradation data z' as the training data for the parameter learning process g. This makes it possible to improve the accuracy of SOH estimation while ensuring the performance of the SOH estimation model f itself, which is based on the physical and chemical characteristics of battery degradation.

[0031] Furthermore, the weighting W of the degradation performance data z is performed by a process that, for example, increases the weight of the performance data of batteries that have ancillary information that is highly similar to the ancillary information of battery 1. Figure 4(a) is a diagram of the weighting based on dimensionality reduction according to the similarity of the ancillary information of batteries according to the embodiment of this disclosure. The weighting of the degradation performance data z is performed, for example, by performing dimensionality reduction on the set of ancillary information to calculate feature vectors between batteries for the ancillary information of battery 1, which is the target of estimation, and the ancillary information of batteries included in the degradation performance data. The similarity between batteries is calculated according to the length of the feature vectors, and by performing weighting according to the similarity, degradation performance data z' is generated in which batteries that have ancillary information similar to the ancillary information of battery 1 are given a larger weight.

[0032] However, the weighting of the degradation performance data z may be performed by rule-based weighting, as shown in Figure 4(b). In Figure 4(b), for example, vehicles equipped with batteries are compared, and batteries installed in vehicles of the same type as vehicle C, which is equipped with battery 1, are given a greater weight. By performing similar weighting based on each accompanying information and summing the weight values ​​of each battery, for example, degradation performance data z' is generated in which batteries with accompanying information similar to that of battery 1 are given a greater weight.

[0033] The degradation estimation unit 53 is connected to the usage history recording unit 22 and the model generation unit 52. The degradation estimation unit 53 uses the SOH estimation model generated by the model generation unit 52 to estimate the SOH of battery 1 based on the usage history information of battery 1. The degradation estimation unit 53 is composed of, for example, a CPU, MPU, working memory, and a non-volatile storage device that stores a control program. Furthermore, the degradation estimation unit 53 does not necessarily have to be part of the battery degradation estimation device 5, and may be built in an external cloud computing environment. Also, the model generation unit 52 and the degradation estimation unit 53 may be composed of the same CPU, MPU, working memory, and a non-volatile storage device that stores a control program.

[0034] As described above, the battery degradation estimation system according to the embodiment of this disclosure generates new degradation data based on accumulated degradation data, giving greater weight to batteries that are highly similar to the supplementary information of the battery to be estimated. In addition, by using the generated degradation data as training data for machine learning, it becomes possible to generate a SOH estimation model that better matches the battery to be estimated. This provides a battery degradation estimation method that can improve estimation accuracy while ensuring the performance of the model based on physical and chemical degradation characteristics.

[0035] <Method for estimating battery degradation> Next, with reference to Figure 5, a battery degradation estimation method according to an embodiment of this disclosure will be described. Figure 5 is a flowchart of the battery degradation estimation method according to an embodiment of this disclosure. First, the model generation unit 52 acquires supplementary information and degradation history data for battery 1, which is the battery to be estimated, and the degradation estimation unit 53 acquires usage history information for battery 1 (step S1). However, the acquisition of usage history information for battery 1 by the degradation estimation unit 53 only needs to be performed before the degradation estimation unit 53 executes step S5, which will be described later.

[0036] Next, it is decided whether or not to perform weighting on the degradation performance data z (step S2). In step S2, for example, if the number of degradation performance data z is excessive for the performance of the model generation unit 52, the model generation unit 52 does not perform weighting. However, depending on the performance of the model generation unit 52, weighting may be performed on only some of the degradation performance data z. The threshold number of degradation performance data z for determining whether or not to perform weighting is appropriately determined according to the performance of the model generation unit 52.

[0037] Furthermore, when comparing the supplementary information of battery 1 with the supplementary information of batteries included in the degradation performance data z, if the number of batteries with supplementary information similar to that of battery 1 is small, weighting based on that supplementary information will not be performed in the rule-based weighting. For example, if the number of battery data points for vehicles similar to vehicle C in which battery 1 is installed is extremely small, weighting based on the vehicle in which it is installed will not be performed. However, the threshold for the number of batteries with supplementary information similar to that of battery 1 will be appropriately determined according to the number of battery data points included in the degradation performance data z.

[0038] If weighting is performed on the degradation performance data z (Yes in step S2), the model generation unit 52 performs weighting on the degradation performance data z (step S3). The weighting is performed as described above, and degradation performance data z' is generated in which batteries with similar supplementary information to battery 1 are given greater weight. On the other hand, if weighting is not performed on the degradation performance data z (No in step S2), step S3 is not performed, and the degradation performance data z becomes the training data as is.

[0039] Next, machine learning is performed using the actual degradation data z' or z as training data to generate a trained model that takes battery usage history information as input and outputs the battery's degradation level, which is then used as a degradation estimation model (SOH estimation model) (Step S4). Here, if the actual degradation data z' that matches the supplementary information of battery 1 is used as training data, a SOH estimation model that is more suitable for estimating the SOH of battery 1 is generated.

[0040] In step S4, machine learning is performed to minimize the error between the SOH value output by the SOH estimation model f and the actual SOH value contained in the degradation performance data z', thereby generating the SOH estimation model f. Here, the optimization method for the parameter θ used to generate the SOH estimation model f can be, for example, the least squares method. Alternatively, the parameter θ may be optimized using, for example, artificial intelligence (AI). When optimizing the parameter θ, the weight values ​​assigned to each degradation performance data point in step S3 are used directly as the weights for each degradation performance data point.

[0041] Next, using the degradation estimation model generated in step S4, the degradation estimation unit 53 estimates the State of Health (SOH) of battery 1 (step S5). Finally, using the measured SOH value of battery 1, it is decided whether or not to update the degradation estimation model generated in step S4 (step S6). If the degradation estimation model is not updated (No in step S6), the process is terminated. Alternatively, the measured SOH value data of battery 1 may be saved to the degradation degree data, and the number of samples in the degradation degree data may be increased.

[0042] On the other hand, if the degradation estimation model is updated (Yes in step S6), the measured SOH value of battery 1 is used, and the process returns to step S3 to re-weight the degradation data and update the degradation estimation model. Figure 6 shows a method for updating the degradation estimation model according to the embodiment of this disclosure. The degradation estimation model is updated, for example, by providing feedback to the degradation estimation model so that the absolute value of the error y(t)-s(t) between the estimated value y(t) and the precise SOH measurement value s(t) becomes smaller.

[0043] Here, feedback to the degradation estimation model may always be provided, or it may only be provided when the absolute value of the error y(t)-s(t) exceeds a threshold. Also, for example, if you want to avoid overestimating the estimate when estimating SOH, you may set a positive threshold and provide feedback when the value of the error y(t)-s(t) is greater than the positive threshold. On the other hand, if you want to avoid underestimating the estimate, you may set a negative threshold and provide feedback when the value of the error y(t)-s(t) is less than the negative threshold. Furthermore, the threshold may be other indicators besides the value of SOH itself, such as the ratio of the error y(t)-s(t) to s(t).

[0044] As a method for performing feedback, for example, steps S3 to S5 are repeatedly executed so that the error index (e.g., |ys|, etc.) calculated from the estimated output y=[y(t_1), ..., y(t_n)] corresponding to each time point, for s=[s(t_1), ..., s(t_n)] obtained as measured values ​​at one or more time points, is reduced to a predetermined value.

[0045] The feedback process may involve, for example, in step S3, changing the similarity during weighting, or changing the rule-based classification categories during rule-based weighting. Furthermore, in step S4, it may involve changing the optimization method for the parameter θ used to generate the degradation estimation model, or changing the initial values ​​or random numbers provided to the optimization method. Finally, in step S5, it may involve reducing or adding input variables to the degradation estimation model, or changing the combination of model architectures. The efficiency of the feedback process may be improved by having artificial intelligence perform these operations.

[0046] As described above, the battery degradation estimation method according to the embodiment of this disclosure generates a State of Health (SOH) estimation model using actual degradation data, weighted more heavily for batteries with high similarity to the supplementary information of the battery to be estimated, as training data. Furthermore, the SOH estimated value estimated using the generated SOH estimation model is compared with the measured SOH value, and feedback is provided to update the degradation estimation model. This makes it possible to provide a battery degradation estimation method that can improve estimation accuracy while ensuring the performance of the model based on physical and chemical degradation characteristics. [Explanation of symbols]

[0047] 1 Battery, 2 Battery sensor, 21 Usage history acquisition unit, 22 Usage history recording unit, 23 Ancillary information acquisition unit, 24 Ancillary information recording unit, 3 Operation unit, 4 Operation history acquisition unit, 5 Battery degradation estimation device, 51 Actual data storage unit, 52 Model generation unit, 53 Degradation estimation unit, C Vehicle, S Battery degradation estimation system

Claims

1. A method for estimating the degradation of a battery mounted on a vehicle and supplying power to a motor, A processing step that weights pre-stored battery degradation data according to its similarity to the supplementary information of the battery to be estimated, A generation step is performed to generate a trained model that takes battery usage history information as input and outputs battery degradation level by using the weighted degradation performance data from the processing step as training data for machine learning, The system includes a degradation estimation step of estimating the degree of degradation of the target battery using the trained model based on the usage history information of the target battery. Methods for estimating battery degradation.

2. The supplementary information of the estimated target battery includes both or either the vehicle information and the driving history information of the vehicle. The method for estimating battery degradation according to claim 1.

3. In the processing step described above, a process is performed to increase the weighting of the degradation performance data that has a high similarity to the supplementary information of the battery to be estimated. A method for estimating battery degradation according to claim 1 or 2.

4. In the processing step, it is determined whether or not to perform weighting of the degradation performance data according to the number of degradation performance data that have a high similarity to the supplementary information of the battery to be estimated. A method for estimating battery degradation according to claim 1 or 2.

5. A measurement step for measuring the degree of degradation of the battery to be estimated, The system further comprises an update step, which updates the trained model by adding the actual degradation data of the battery to be estimated, measured in the measurement step, to the training data. The method for estimating battery degradation according to claim 1.

Images