SOHC Prediction Method and System for Electric Vehicles (EVs)

Abstract

The present invention provides a method and a system for predicting the SOHC of a battery in use. When predicting the SOHC of a battery in use, an SOHC prediction model is generated using cell data and first field data obtained from a test battery, and the second field data obtained from the battery in use is input into the SOHC prediction model to predict the real-time SOHC.

Description

Technical Field

[0001] The present invention relates to a method and a system for predicting the SOHC of a battery for an electric vehicle (EV) using field data. In particular, in a cloud BMS, the present invention relates to a method and a system for calculating an accurate SOHC (state of health capacity) value of a battery by using a machine learning prediction model for calculating the accurate SOHC value of the battery.

Background Art

[0002] Conventionally, when predicting the SOHC (state of health capacity) of a battery for an electric vehicle (EV) using field data, the battery capacity value calculated in an on-board BMS (Battery Management System) has been used. However, since the battery capacity value calculated in this way cannot be regarded as the true value, the SOHC value of the battery predicted using this value contains even more errors.

[0003] Therefore, a method for directly calculating SOHC from cell data is required.

[0004] Related prior inventions include Patent Document 1 that predicts the remaining life of a battery via an LSTM (Long Short Term Memory) model, and Patent Document 2 that predicts the dischargeable time of a secondary battery using a neural network.

[0005] However, the prior inventions as described above cannot present a method for generating an accurate prediction model as in the present invention. In addition, the limitation of only using on-board calculated values when generating the prediction model cannot be overcome.

[0006] The related prior art is as follows.

Prior Art Documents

Patent Documents

[0007]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0008] Therefore, an object of the present invention is to solve the above-described problems of the conventional technology, generate an accurate SOHC prediction model using cell data obtained in the battery cell test process, and provide a method and system for predicting the accurate SOHC of a field battery using the same.

Means for Solving the Problems

[0009] Therefore, the present invention provides a battery test device that acquires cell data from a plurality of battery cells and acquires first field data from other battery cells or in other charge / discharge cycles during the cell test process, calculates a first SOHC value from the cell data, configures a first learning dataset with the cell data as input values and the first SOHC value as label values to generate a first SOHC prediction model for predicting the SOHC value from the cell data, inputs the first field data into the first SOHC prediction model to calculate a second SOHC value, configures a second learning dataset from the first field data and the second SOHC value as a label value, and retrains the first SOHC prediction model with the second learning dataset to generate a second SOHC prediction model, and provides an SOHC prediction model generation device including the same.

[0010] Furthermore, the present invention provides an application BMS that includes an SOHC prediction model unit including an SOHC prediction model, receives input of field data from a field battery in operation in an application, inputs the data into the SOHC prediction model, and calculates a real-time SOHC value of the field battery.

[0011] At this time, in the cell test process, the SOHC prediction model acquires cell data from a plurality of battery cells, acquires first field data from other battery cells or in other charge / discharge cycles, calculates a first SOHC value from the cell data, constructs a first learning data set from the cell data and the first SOHC value as a label value, and generates a first SOHC prediction model for predicting an SOHC value from the cell data. Then, the first field data is input into the first SOHC prediction model to calculate a second SOHC value, a second learning data set is constructed from the first field data and the second SOHC value as a label value, and the first SOHC prediction model is re-learned with the second learning data set and then generated.

[0012] Furthermore, the present invention provides a SOHC prediction system for a field battery, comprising: a battery test device that acquires cell data from a plurality of battery cells in a cell test process and acquires first field data from other battery cells or in other charge-discharge cycles; a first SOHC prediction model generation unit that calculates a first SOHC value from the cell data, constructs a first learning data set from the cell data and the first SOHC value as a label value to generate a first SOHC prediction model for predicting the SOHC value from the cell data, inputs the first field data into the first SOHC prediction model to calculate a second SOHC value, constructs a second learning data set from the first field data and the second SOHC value as a label value, and relearns the first SOHC prediction model with the second learning data set to generate a second SOHC prediction model; and an application BMS that includes the second SOHC prediction model as a SOHC prediction model, receives an input of field data from a field battery during operation in an application, and inputs the field data into the SOHC prediction model to calculate a real-time SOHC value of the field battery.

[0013] Furthermore, the present invention includes, in the cell test process, a cell data acquisition process for acquiring cell data from a plurality of battery cells, a first SOHC value calculation process for calculating a first SOHC value corresponding to the cell data, a first SOHC prediction model generation process for learning and generating a first SOHC prediction model that predicts an SOHC value from cell data with the cell data as an input value and the first SOHC value as a first label value, a first field data acquisition process for acquiring second cell data different from the first field data from a plurality of battery cells in the cell test process, a second SOHC prediction value calculation process for inputting the first field data into the generated first SOHC prediction model to generate a second SOHC prediction value corresponding to the first field data, a second SOHC prediction model generation process for relearning the first SOHC prediction model with the generated second SOHC prediction value as a label value and the first field data as an input value to generate a second SOHC prediction model, and an SOHC value prediction process for inputting field data generated while operating a predetermined battery cell into the second SOHC prediction model to predict an SOHC value during operation of the predetermined battery cell, and provides a method for predicting the SOHC of a field battery cell.

Effect of the Invention

[0014] According to an embodiment of the present invention, after generating a first SOHC prediction model using cell data acquired in the battery test process, the first SOHC prediction model is relearned with the SOHC value calculated using this as a label value to generate a second SOHC prediction model, and by applying this as the final SOHC prediction model, it is possible to provide a prediction model with higher accuracy than the conventional technology that uses only the on-board calculated capacity value.

[0015] The drawings attached to this specification illustrate desirable embodiments of the present invention and serve to further understand the technical idea of the present invention together with the content of the invention described above. Therefore, the present invention is not to be construed as limited only to the matters described in the drawings.

Brief Description of Drawings

[0016]

Figure 1

Figure 2

Figure 3

Modes for Carrying Out the Invention

[0017] When predicting the SOHC of a battery in use, the present invention provides a method and a system for generating a SOHC prediction model using cell data acquired from a test battery and first field data different therefrom, and inputting second field data acquired from the battery in use into the SOHC prediction model to predict real-time SOHC.

[0018] 1. Definition of Terms The terms used in the present invention shall follow the definitions of the terms described below unless otherwise specified.

[0019] The present invention constructs a model for predicting the SOHC of a battery in use from cell data and field data. In the present invention, SOHC (State of health capacity) means the ratio of capacity reduction due to deterioration compared to the initial capacity of the battery. Generally, it is calculated by the formula SOHC (%) = (current battery capacity / initial battery capacity) × 100, where the current battery capacity indicates the actual capacity of the battery that changes according to use and charge-discharge cycles, and the initial battery capacity indicates the capacity when the battery is first manufactured. The closer the SOHC value is to 100%, the better the health state of the battery can be judged. The lower the SOHC value, the shorter the battery life and the lower the performance.

[0020] The cell data in the present invention means battery cell data acquired from a test battery or a reference battery during the cell test process. The field data in the present invention means battery data acquired from an in-use battery when the battery is used in an application such as a vehicle.

[0021] The Cloud Battery Management System (Cloud BMS) in the present invention means a system that executes battery management functions on a cloud server, enabling users to monitor and manage the battery status in real time via a web browser at any time and from anywhere. The Cloud BMS can efficiently monitor and adjust battery systems centered at various locations, collect and analyze data, and perform predictive analysis. It also has the effect of sharing and associating data among various systems to optimize overall energy management.

[0022] 2. SOHC Prediction System for Field Battery According to the Present Invention The SOHC prediction system for field battery according to the present invention is a system that performs a real-time SOHC prediction method for field batteries. Based on FIGS. 1 and 2, the SOHC prediction system of the present invention will be described.

[0023] As shown in FIG. 1, the system of the present invention includes a SOHC prediction model generation unit 200, and an application BMS 400 that mounts the generated prediction model and receives second field data from a field battery in use in an application, and calculates a real-time SOHC prediction value of the field battery therefrom. The application BMS 400 can be composed of a cloud BMS. Further, the SOHC prediction model generation unit 200 generates a SOHC prediction model from cell data received from the battery test device 100 and first field data received from the application battery 300.

[0024] In another embodiment, as shown in FIG. 2, the system of the present invention can be composed of a separate SOHC prediction device 500 instead of the SOHC prediction model unit 420 and the SOHC prediction unit 430 being mounted on the application BMS 400.

[0025] Hereinafter, each component will be described.

[0026] (1) Battery test device 100 The battery test device 100 is a device that performs various tests on a battery after the battery is manufactured. The present invention uses cell data including various state information of the test battery generated during the test of the battery from such a known battery test device 100 as data for generating a SOHC prediction model.

[0027] The battery test device 100 acquires cell data from the test battery and transmits it to the SOHC prediction model generation unit 200.

[0028] The types and characteristics of the cell data will be described later.

[0029] (2) SOHC prediction model generation unit 200 The SOHC prediction model generation unit 200 is a component that generates a SOHC prediction model using not only the cell data described above but also the first field data, and includes a computer algorithm for performing the process. In the present invention, a known neural network is used as the neural network that forms the basis of the SOHC prediction model, and the data and learning method for learning the SOHC prediction model are characteristic.

[0030] The SOHC prediction model generation unit 200 generates a SOHC prediction model by performing the processes of S10 to S60 described later, and provides the generated prediction model, that is, a computer-executed algorithm for calculating the SOHC prediction value of the battery, to the SOHC prediction model unit 420 online or offline.

[0031] The SOHC prediction model generation unit 200 obtains cell data and first SOHC data, which is the SOHC value corresponding to the cell data, from a battery test device or a reference data set already secured as a first label value, and constructs a first learning data set from the cell data and the first SOHC value as the first label value to generate a first SOHC prediction model for predicting the SOHC value from the cell data. The first SOHC value is the SOHC value calculated based on the cell data from the on-board BMS of the test battery. As the first SOHC prediction model, a predetermined artificial neural network model configured to predict the SOHC value from the cell data can be used. Taking one embodiment as an example, a support vector regression (SVR) model is applied, and machine learning is advanced to obtain a first parameter set including coefficients and intercepts that determine the regression function of the SVR model and model parameters in a first learning data set with the cell data as the input value and the first SOHC value as the label value to obtain the first SOHC prediction model. The first learning data set is {cell data, first SOHC}.

[0032] After that, the SOHC prediction model unit 420 acquires first field data from a predetermined field battery, inputs this into the first SOHC prediction model to calculate a second SOHC value as a second label value, configures a second learning data set {first field data, second SOHC value} with the first field data as an input value and the second SOHC value as a label value. The second SOHC value is the SOHC value calculated by inputting the first field data into the first SOHC prediction model.

[0033] In another embodiment, the first field data can be acquired from the battery test device 100. In this case, the battery test device 100 acquires, as the first field data, cell data obtained from another test battery or in a charge / discharge cycle different from that at the time of obtaining the cell data, rather than the test battery from which the cell data was obtained. Also in this case, the second learning data set is set to {first field data, second SOHC value}.

[0034] After that, by re - learning the first SOHC prediction model using the second learning data set to obtain a second parameter set, a second SOHC prediction model is generated and provided as the final SOHC prediction model. The provided second SOHC prediction model is installed in the SOHC prediction model unit 420 of the application BMS 400 described later.

[0035] (3) Application BMS 400 The application BMS 400 of the present invention is a battery management device that manages the application battery 300, and in addition to normal BMS components, may include a field data collection unit 410, an SOHC prediction model unit 420, an SOHC prediction unit 430, and a monitoring unit 440.

[0036] The application BMS 400 of the present invention collects field data such as battery state information from the applied application battery 300, and can be composed of an on-site battery management system (On-Site BMS) configured at the same location or facility as the application battery, or can be composed of a cloud BMS.

[0037] When the application BMS 400 is composed of a cloud BMS, the application BMS 400 is connected to a plurality of on-board BMSs via a predetermined wired / wireless communication network, and receives cell data and field data from the on-board BMSs.

[0038] The on-board BMS in the present invention means the BMS of the application battery 300 or the test battery for calculating the first SOHC value. The on-site battery management system (On-Site BMS) means that the application BMS 400 that receives field data from the on-board BMSs of a plurality of application batteries 300 is physically the same as the application battery 300 or forms one system adjacent thereto. When the application BMS 400 is configured to receive field data from the application battery 300 via a wired / wireless network at a distance, it is called a cloud BMS.

[0039] Such an application BMS 400 includes the following components in addition to a data path connected to a plurality of application batteries 300 or a communication device (not shown) connected to a wired / wireless network.

[0040] A. Field data collection unit 410 The field data collection unit 410 collects field data from the battery used in the application and delivers it to the SOHC prediction unit 430. The field data collected by the field data collection unit 410 is shown as second field data in FIGS. 1 and 2 in order to distinguish it from the first field data for relearning the SOHC prediction model.

[0041] B. SOHC prediction model unit 420 The SOHC prediction model unit 420 is composed of a memory device equipped with an SOHC prediction model generated according to the SOHC prediction model generation method according to the present invention described later. The SOHC prediction model is a computer-executed algorithm that constitutes an artificial neural network model learned to receive input of field data in real time and calculate the real-time SOHC prediction value of the application battery 300 corresponding to the real-time field data.

[0042] C. SOHC prediction unit 430 The SOHC prediction unit 430 reads the computer-executed algorithm that constitutes the SOHC prediction model, inputs the field data into the SOHC prediction model, and outputs the SOHC prediction value corresponding to the field data. It may be composed of a processor of an arithmetic device or a computer device equipped with a predetermined processor.

[0043] D. Monitoring unit 440 The monitoring unit 440 monitors the state of the application battery 300 based on the real-time SOHC prediction value calculated by the SOHC prediction unit 430.

[0044] The application BMS 400 of the present invention as described above can be mounted on an application device such as a vehicle, or can be composed of a cloud BMS (Cloud Battery Management System). When composed of a cloud BMS, the cloud BMS may include a communication module that receives field data from a remote location from the application battery 300.

[0045] On the other hand, as shown in FIG. 2, the SOHC prediction system according to another embodiment of the present invention may not configure the SOHC prediction model unit 420, the SOHC prediction unit 430, and the monitoring unit 440 in the application BMS 400, but may be composed of a separate SOHC prediction device 500.

[0046] 3. SOHC prediction method for cells during use of the present invention The present invention predicts and calculates the SOHC of cells during use according to the following procedure.

[0047] (1) Cell data acquisition process (S10) It is a process of acquiring cell data from a plurality of battery cells during the cell test process after the manufacture of the cells. The cell data may include at least one or more of the cumulative charge capacity during the charge cycle of the battery cell, the cumulative discharge capacity during the discharge cycle, the cumulative charge energy which is the energy required for charging during the charge cycle, the cumulative discharge energy which is the discharge energy during the discharge cycle, and the average temperature data of the battery cell.

[0048] Preferably, the cell data is similar in characteristics to the field data and can be limited to the cell data acquired in the slow charge section which is a relatively shaped section. The slow charge can set the charge rate to 0.33C-rate. The reason for limiting the cell data to the data acquired in the slow charge section is to ensure that the cell data acquired in the test environment is in a section similar to the field data which is the data in the actual use environment.

[0049] More preferably, the cell data can be limited to the cell data obtained in a predetermined partial charge interval. At this time, the partial charge interval may be an interval of 3.6 to 3.9V. The reason for limiting the cell data to a predetermined partial charge interval is that there is a difference in the charge start voltage for each cell in each cycle during the battery test process. Therefore, a common voltage interval is extracted, and a voltage interval highly correlated with SOHC is extracted. The inventor of the present invention confirmed through experiments the correlation between SOHC and cell voltage in units of 0.1V, and found that the correlation between SOHC and cell voltage is high in the interval of 3.6 to 3.8V.

[0050] (2) First SOHC label value calculation process (S20) This is a process of calculating the cell SOHC value corresponding to the cell data in the cell test process after the manufacture of the cell.

[0051] After the manufacture of the cell, the first SOHC value calculated in the BMS during the test process is the value calculated in the on-board BMS. The BMS at this time may be the on-board BMS used in the cell test process. As a method for calculating the first SOHC using the cell data in the on-board BMS, a known method is utilized.

[0052] (3) First SOHC prediction model generation process (S30) Using the acquired cell data as input data and the corresponding first SOHC value as the label value, a first SOHC prediction model for predicting the SOHC value from the cell data is generated. The first SOHC prediction model is generated by machine learning a model based on a known neural network using the cell data and the corresponding first SOHC label value.

[0053] (4) First field data acquisition process (S40) On the one hand, the present invention obtains first field data from an application battery 300 in actual operation, which is different from the test battery that obtains the cell data in the cell test process. Similar to the cell data, the first field data is also composed of data including at least one or more of the cumulative charge capacity during the charge cycle of the battery cell, the cumulative discharge capacity during the discharge cycle, the cumulative charge energy which is the energy required for charging during the charge cycle, the cumulative discharge energy which is the discharge energy during the discharge cycle, and the average temperature data of the battery cell. However, the first field data is different in that it is battery data actually obtained from a battery in actual operation in the application.

[0054] On the other hand, the first field data may be data obtained from a battery different from the battery that obtains the second field data described later, or may be data obtained in a cycle different from the charge / discharge cycle that obtains the second field data.

[0055] (5) Second SOHC Label Value Calculation Process (S50) This is a process of inputting the first field data into the first SOHC prediction model to calculate the second SOHC label value corresponding to the first field data.

[0056] (6) Second SOHC Prediction Model Generation Process (S60) This is a process of re - learning the first SOHC prediction model to generate a second SOHC prediction model. The second SOHC prediction model uses the first field data input to the first SOHC prediction model as input data, and the second SOHC calculation value calculated from the first SOCH prediction model as the label value. Taking {the first field data, the second SOHC calculation value} as learning data, this is a process of re - learning the first SOHC prediction model to generate a second SOHC prediction model.

[0057] (7) Second Field Data Collection Process (S70) This is the process of collecting second field data from an in-operation field battery in an application that attempts to predict SOHC. Similar to the cell data, the second field data also includes at least one or more of the cumulative charge capacity during the charge cycle of the battery cell, the cumulative discharge capacity during the discharge cycle, the cumulative charge energy which is the energy required for charging during the charge cycle, the cumulative discharge energy which is the discharge energy during the discharge cycle, and the average temperature data of the battery cell.

[0058] (8) SOHC Prediction Process of Field Battery (S80) This is the process of inputting the collected field data (second field data) into the second SOHC prediction model to calculate the real-time SOHC value of the field battery (application battery).

[0059] The embodiments disclosed based on the accompanying drawings have been described above. Those with ordinary knowledge in the technical field to which the present invention belongs should be able to understand that the present invention can be implemented in forms different from the disclosed embodiments without changing the technical idea and essential features of the present invention. The disclosed embodiments are merely exemplary and should not be construed as restrictive.

[0060] The names of the respective parts shown in the drawings of the present invention are as follows.

Explanation of Reference Signs

[0061] 100 Battery Test Device 200 SOHC Prediction Model Generation Unit 300 Application Battery 400 Application BMS 410 Field Data Collection Unit 420 SOHC Prediction Model Unit 430 SOHC Prediction Unit 440 Monitoring Unit 500 SOHC Prediction Device

Claims

1. In a cell test process, a battery test device that acquires cell data from a plurality of test battery cells, calculates a first SOHC value from the cell data, constructs a first learning data set with the cell data as an input value and the first SOHC value as a first label value, and generates a first SOHC prediction model for predicting the SOHC value from the cell data, an SOHC prediction model generation unit that generates a first SOHC prediction model by training a predetermined artificial neural network using {cell data, first SOHC value} as first learning data, and retrains the first SOHC prediction model using {first field data, second SOHC value} as second learning data to generate a second SOHC prediction model, comprising: The first field data is battery data acquired from an application battery in operation, cell data acquired from other test battery cells other than the test battery cell from which the cell data was acquired, or cell data acquired in a charge / discharge cycle different from the cell data from the test battery cell from which the cell data was acquired, The first SOHC value is calculated from the cell data, The second SOHC value is a value calculated by inputting the first field data into the first SOHC prediction model. An SOHC prediction model generation device.

2. An application BMS comprising an SOHC prediction model unit including an SOHC prediction model, receiving an input of field data from an application battery in operation in an application, inputting the data into the SOHC prediction model, and calculating a real-time SOHC value of the application battery.

3. The SOHC prediction model generates a first SOHC prediction model through machine learning using {cell data, first SOHC value} as first learning data, is generated by retraining the first SOHC prediction model using {first field data, second SOHC value} as second learning data, The cell data is battery data acquired from a test battery, The first field data is battery data acquired from an application battery in operation, The first SOHC value is calculated from the cell data, The second SOHC value is a value calculated by inputting the first field data into the first SOHC prediction model. The application BMS according to claim 2.

4. A field data collection unit that receives first field data and second field data from the application battery via a predetermined communication network; The SOHC prediction model receives an input of second field data from the application battery and calculates a real-time SOHC value of the application battery. The application BMS according to claim 3.

5. A battery test device that acquires cell data from a plurality of battery cells in the cell test process; An application BMS that receives an input of field data from an application battery in operation in an application, inputs it into an SOHC prediction model, and calculates a real-time SOHC value of the application battery; Comprising; The application BMS is An SOHC prediction model generation unit that calculates a first SOHC value from the cell data, configures a first learning data set with the cell data as an input value and the first SOHC value as a first label value, and predicts the SOHC value from the cell data to generate a first SOHC prediction model, learns a predetermined artificial neural network using {cell data, first SOHC value} as first learning data to generate a first SOHC prediction model, and relearns the first SOHC prediction model using {first field data, second SOHC value} as second learning data to generate a second SOHC prediction model; An SOHC prediction model unit that includes the second SOHC prediction model as an SOHC prediction model; Comprising; The first SOHC value is calculated from the cell data; The second SOHC value is a value calculated by inputting the first field data into the first SOHC prediction model. An SOHC prediction system for an application battery.

6. The cell data or the first field data is The SOHC prediction system of the application battery according to claim 5, including at least one or more of the cumulative charge capacity during the charging cycle of the battery cell, the cumulative discharge capacity during the discharging cycle, the cumulative charge energy which is the energy required for charging during the charging cycle, the cumulative discharge energy which is the discharge energy during the discharging cycle, and the average temperature data.

7. The first field data is The SOHC prediction system of the application battery according to claim 6, which is data obtained in the slow charging section and a predetermined partial charging section.

8. During the cell test process, a cell data acquisition process for acquiring cell data from a plurality of battery cells, A first SOHC value calculation process for calculating a first SOHC value corresponding to the cell data, A first SOHC prediction model generation process for machine learning a learning data set with the cell data as the input value and the first SOHC value as the first label value to generate a first SOHC prediction model for predicting the SOHC value from the cell data, A first field data acquisition process for acquiring first field data from a plurality of battery cells, A second SOHC prediction value calculation process for inputting the first field data into the generated first SOHC prediction model to generate a second SOHC prediction value corresponding to the first field data, A second SOHC prediction model generation process for re-learning the first SOHC prediction model with the generated second SOHC prediction value as the label value and the first field data as the input value to generate a second SOHC prediction model, An SOHC value prediction process for inputting the field data generated while operating a predetermined battery cell into the second SOHC prediction model to predict the SOHC value during the operation of the predetermined battery cell, The SOHC prediction method for the application battery cell, including.

9. The cell data or the first field data is The SOHC prediction method for the application battery cell according to claim 8, including at least one or more of the cumulative charge capacity during the charging cycle of the battery cell, the cumulative discharge capacity during the discharging cycle, the cumulative charge energy which is the energy required for charging during the charging cycle, the cumulative discharge energy which is the discharge energy during the discharging cycle, and the average temperature data.

10. The method for predicting the SOHC of an application battery cell according to claim 9, wherein the first field data is obtained from a battery cell different from the cell data, or is obtained in a charge / discharge cycle different from the cell data.

11. The first field data is The method for predicting the SOHC of an application battery cell according to claim 10, wherein the data is obtained in a slow charge section and a predetermined partial charge section.

12. The method for predicting the SOHC of an application battery cell according to claim 11, wherein the first SOHC prediction model and the second SOHC prediction model are a regression model or an artificial neural network model generated by supervised learning.

Images