Method and system for detecting battery abnormalities through an artificial intelligence neural network battery model based on field data

Abstract

The present invention provides a method and system for detecting abnormal battery behavior using field data, which extracts real-time field data in a time series manner from a battery in operation, compares predicted values ​​of cell voltage and cell temperature from the extracted field data with the cell voltage and cell temperature values ​​of the real-time field data, and detects whether or not the battery is behaving abnormally based on the deviation.

Description

[Technical Field]

[0001] The present invention relates to a method and system for detecting battery abnormalities through a battery model. In particular, the present invention relates to a method and system for detecting battery abnormalities through an artificial intelligence neural network battery model using field data when constructing a battery model, rather than a conventional chemical / electrical equivalent circuit model. [Background technology]

[0002] Traditionally, modeling of conventional lithium-ion batteries has been achieved by utilizing chemical composition and electrical equivalent circuits. Modeling of such batteries is basically based on cell test data, including charge / discharge.

[0003] However, internal cell test data differs from the actual field environment, and even if internal cell tests are conducted under various conditions, there are practical experimental limitations such as the number of cell samples, module / rack unit, degree of deterioration, temperature conditions, test equipment, and time, so it may not be possible to obtain sufficient data.

[0004] In this regard, Patent Document 1 presents a system that receives measurement data of a battery's initial characteristics, trains an artificial intelligence neural network from the data, predicts long-term characteristic data from the data, and determines its reliability, while Patent Document 2 discloses a battery state estimation method that inputs physical quantity information of the battery into a battery learning model and obtains estimated information from the battery learning model.

[0005] However, such conventional techniques are still insufficient to realize a method for estimating the state of a battery based on actual field data and determining whether or not there is an abnormality in the battery.

[0006] The relevant prior art includes the following documents: [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Republic of Korea Publication Patent No. 10-2009-0020448 [Patent Document 2] Republic of Korea Publication Patent No. 10-2018-0057266 Summary of the Invention [Problem to be solved by the invention]

[0008] To solve the above-mentioned problems, the present invention aims to provide a method and system for detecting abnormalities in battery cells by utilizing field data.

[0009] There are also issues to be resolved when utilizing field data, namely, the inevitable limitations on the types of data that can be collected and the low accuracy of the data.

[0010] To solve this problem, we are trying to provide a battery anomaly detection method and system using artificial intelligence (AI) learning to model batteries using a wide variety of field data. [Means for solving the problem]

[0011] In order to solve the above-mentioned problems, the present invention provides a system for detecting abnormal battery behavior, which includes a field data calculation unit that receives real-time measurement values ​​from a battery in operation, calculates and outputs field data from the field data, a field data pre-processing unit that extracts and outputs field data for predicting cell voltage from the field data, an artificial intelligence neural network unit that receives the output of the field data pre-processing unit, predicts cell voltage, and outputs the predicted value, and a battery abnormal behavior detection unit that compares the predicted value output by the artificial intelligence neural network unit with the field data, and determines that the battery is behaving abnormally if the deviation is greater than or equal to a predetermined range.

[0012] The artificial intelligence neural network unit may include a cell voltage prediction model that learns, as learning data, field data corresponding values ​​for predicting the cell voltage calculated from a standard battery rather than a battery that calculates the field data, and that receives input of field data for predicting the cell voltage and outputs a predicted value of the cell voltage for the next period; and a cell temperature prediction model that learns, as learning data, field data corresponding values ​​for predicting the cell temperature calculated from a standard battery rather than a battery that calculates the field data, and that receives input of field data for predicting the cell temperature and outputs a predicted value of the cell temperature for the next period.The artificial intelligence neural network unit may further include a neural network learning unit that re-learns and updates the cell voltage prediction model and the cell temperature prediction model by adding field data for a predetermined period of a normal operating section of the battery as new learning data.

[0013] The present invention also provides a method for detecting abnormal battery behavior, including: a field data calculation process for measuring and calculating real-time battery status information data from a battery in operation; a field data preprocessing process including a field data preprocessing process for cell voltage prediction, which extracts data for predicting cell voltage from the calculated field data; a real-time prediction process including a cell voltage prediction process for inputting the data for predicting cell voltage into a cell voltage prediction model to calculate a predicted cell voltage value for the next period; and a battery abnormal behavior detection process including an abnormal cell voltage detection process for comparing the predicted cell voltage value with the cell voltage value of the field data and generating an abnormal cell voltage behavior detection signal if the deviation is greater than or equal to a predetermined range.

[0014] In this case, the data for predicting the cell voltage may be time series values ​​of rack current, atmospheric temperature, fan on / off information, module SOC (state of charge), cell SOH (health status or state of deterioration), and cell voltage for each battery cell and battery module constituting the battery, and the data for predicting the cell temperature may be time series values ​​of cell temperature, atmospheric temperature, and fan on / off information for each battery cell and battery module constituting the battery. [Effects of the Invention]

[0015] According to the present invention, the temperature and voltage values ​​of the battery can be predicted in real time using field data of the battery, and these can be compared with the actual temperature and voltage values ​​to detect whether the battery is operating abnormally in real time.

[0016] The following drawings attached to this specification illustrate preferred embodiments of the present invention and, together with the above detailed description of the invention, serve to further understand the technical concept of the present invention. Therefore, the present invention should not be interpreted as being limited to the matters described in the drawings. [Brief explanation of the drawings]

[0017] [Figure 1] 1A to 1C are diagrams showing examples of field data measured and calculated in the present invention. [Figure 2] 2 is a diagram showing an example in which field data acquired during a certain section or a predetermined time period among the field data in FIG. 1 is preprocessed and extracted as learning data. FIG. [Figure 3] 1 is a diagram showing the steps of a method for detecting abnormal battery behavior according to the present invention; [Figure 4] 1 is a graph showing field data values ​​of cell temperature and predicted values ​​of cell temperature over time. [Figure 5] 1 is a block diagram of a system for detecting abnormal battery behavior according to the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0018] The present invention predicts the voltage and temperature at the next time point by learning a prediction model based on field data. The types of field data used in the present invention are as follows.

[0019] 1. Field Data

[0020] In the present invention, "field data" refers to the following data measured in real time from a battery in operation and the following data calculated from the measured data: The field data is time-series data of battery status information calculated at predetermined time intervals during battery operation.

[0021] The types of field data used in the present invention are as follows:

[0022] (1) Rack Data

[0023] Battery rack data. Rack data includes rack current, ambient temperature, and fan on / off information. These are usually measured and calculated by the rack battery management system (BMS).

[0024] The rack current is expressed as Rack.Current, the ambient temperature as Ambient.Temp, and the fan on / off information as Fan_ON / OFF flag value, and are measured at predetermined time intervals.

[0025] The rack current (Rack Current) and ambient temperature (Ambient Temp.) measured at a predetermined interval at time t can be expressed as Rack.Current(t), Ambient.Temp(t), and Fan.ON(t) as fan on / off information. The values ​​of Rack.Current(t) and Ambient.Temp(t) are measured values ​​of the appropriate measurement sensors, and Fan.ON(t) can have a value of "1" if the fan is running and a value of "0" if the fan is not running.

[0026] For example, when measured every second, the rack current, ambient temperature, and fan on / off information (if the fan is running) 3 seconds after activation can be written as Rack.Current(3), Ambient.Temp(3), and Fan.ON(3), respectively.

[0027] (2) Module Data

[0028] This is data for each of the many battery modules included in the battery rack. Module data includes the SOC value of each module. When there are i modules in the battery rack, each module is represented as Mk (k=1, 2, 3, ..., i), and the SOC value (SOCM) of each module is represented as SOC.Mk. For example, the SOC value of the second module is represented as SOC.M2. Module data is usually measured and calculated by the module BMS (Module Battery Management System).

[0029] Similar to rack data, the SOC.Mk value can be measured at predetermined time intervals, and the SOC.Mk value at time t is written as SOC.Mk(t).

[0030] (3) Cell data

[0031] This is data for each of the many battery cells contained in a battery module. Cell data includes a cell voltage value (Cell V) as a measured voltage value for each cell, and the cell voltage value is expressed as Cell.V.MkCj (k=1, 2, 3, ..., i, j=1, 2, 3, ..., m). For example, the cell voltage value of the third cell in the second module is expressed as Cell_V_M2C3. Cell data is usually measured and calculated by a cell battery management system (BMS) or, if there is no cell BMS, by a module battery management system (BMS).

[0032] Similarly, the Cell.V.MkCj value can be measured at predetermined time intervals, and the Cell.V.MkCj value at time t is written as Cell.V.MkCj(t).

[0033] The cell data may include cell temperature, where the cell temperature of the jth cell in the kth module at time t is written as Cell.T.MkCj(t).

[0034] (4) Calculated value

[0035] In addition, the present invention uses the SOH value (SOH MkCj) as one of the calculated values. The SOH value of each cell is calculated by a cell battery management system (BMS), a module battery management system (BMS), or a rack battery management system (BMS), and the SOH of a specific cell is expressed as SOH.MkCj (k=1, 2, 3, ..., i, j=1, 2, 3, ..., m).

[0036] Similarly, the SOH.MkCj value can be measured at predetermined time intervals, and the SOH.MkCj value at time t is written as SOH.MkCj(t).

[0037] 2. How to detect abnormal battery behavior

[0038] The method for detecting abnormal battery behavior according to the present invention will be described with reference to FIG.

[0039] 2-1. Field data calculation and preprocessing process (S100)

[0040] The present invention is a procedure for measuring, calculating and pre-processing the field data described above.

[0041] (1) Field data calculation process (S110)

[0042] This is the process of measuring and calculating real-time battery status information data from an operational battery.

[0043] The field data is measured and calculated in real time from the battery in operation at predetermined time intervals. FIG. 1 shows an example of each field data measured and calculated in the present invention.

[0044] The measurement and calculation of field data may be performed in each cell battery management system (BMS), a module battery management system (BMS), or a rack battery management system (BMS), and transmitted to the field data calculation unit 100. The field data calculation unit may be configured to be integrated into the rack BMS.

[0045] Of the field data shown in FIG. 1 and described above, field data acquired during a certain section or a predetermined time period is preprocessed and extracted as learning data as shown in FIG.

[0046] (2) Field data preprocessing process for cell voltage prediction (S120)

[0047] This is a pre-processing process of field data for predicting cell voltages, which extracts data for predicting cell voltages from the calculated field data.

[0048] In the present invention, the cell voltage is predicted using a cell voltage prediction model that predicts the cell voltage from field data. As shown in Figure 2, the rack current, ambient temperature, SOC, SOH, and cell voltage of the previous cycle are input as data for cell voltage prediction.

[0049] For this purpose, in the pre-processing step, the rack current, ambient temperature, SOC, SOH, and cell voltage values ​​Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t) are periodically extracted from field data calculated in real time at a predetermined cycle. Figure 2 shows the data format for extracting the values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), and SOC.Mk(t) for each time cycle, as well as the corresponding values ​​of SOH_M1C2(t) and Cell_V_M1C2(t) for cells 1 and 2 of module 1.

[0050] (3) Field data preprocessing process for cell temperature prediction (S130)

[0051] This is a step of extracting data for predicting the calculated cell temperature from the calculated field data.

[0052] In addition, the present invention predicts the cell temperature using a cell temperature prediction model that predicts the cell temperature from field data, and predicts the cell temperature for the next cycle using the current cell temperature value Cell.T.MkCj(t), the ambient temperature Ambient.Temp(t), and Fan.ON(t) as data for predicting the cell temperature.

[0053] For this reason, in the pre-processing step, the values ​​of Cell.T.MkCj(t), Ambient.Temp(t), and Fan.ON(t) are periodically extracted from field data calculated in real time at a predetermined cycle.

[0054] 2-2. Prediction model generation process (S200)

[0055] (1) Cell voltage prediction model generation process (S210)

[0056] Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t) are input to a predetermined artificial intelligence neural network, and the output value is used as Cell.V.MkCj(t+1) to perform training and generate a cell voltage prediction model 210 for predicting cell voltage. The cell voltage prediction model is generated using a predetermined artificial intelligence neural network. Any known artificial intelligence neural network can be used as the artificial intelligence neural network, and the artificial intelligence neural network that has completed training using predetermined training data is stored in the cell voltage prediction model 210 of FIG. 5.

[0057] As learning data for the cell voltage prediction model 210, values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t) for a predetermined time period preprocessed through a field data preprocessing process (S120) for predicting the cell voltage are used as input values ​​for a neural network, and Cell.V.MkCj(t+1 When the values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t) for a predetermined time interval are input to the artificial intelligence neural network using Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t) as its output value, the artificial intelligence neural network is trained and trained to generate a cell voltage prediction model so that its output value becomes Cell.V.MkCj(t+1), which is the measured value of the cell voltage for the next period.

[0058] The artificial intelligence neural network trained in this way is stored in the cell voltage prediction model 210, and receives inputs of the values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t) to generate Cell.V, which is the predicted value of the cell voltage for the next period. prdic Calculate .MkCj(t+1).

[0059] The values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), Cell.V.MkCj(t), and Cell.V.MkCj(t+1), which are learning data used in the process of generating the cell voltage prediction model, may be field data measured at a time different from the data used in the cell voltage prediction process (S310) described below, or may be sample data obtained before the battery is put into operation or through the operation of a sample battery.

[0060] In another embodiment, the learning data for the cell voltage prediction model 210 may be generated using Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), Cell.V.MkCj(t), and Cell.V.MkCj(t+1) for a predetermined time interval calculated using a standard battery in a laboratory, rather than field data.

[0061] In other words, the cell voltage prediction model is generated by learning data corresponding to data values ​​for predicting the cell voltage calculated from a standard battery, rather than from a battery in operation for calculating the field data, and then receives input data for predicting the cell voltage from the battery in operation and outputs a predicted value for the cell voltage for the next cycle.

[0062] (2) Cell temperature prediction model generation process (S220)

[0063] Furthermore, the present invention includes a cell temperature prediction model, and the cell temperature prediction model for predicting cell temperature also uses the preprocessed data as learning data to generate the cell temperature prediction model 220 in a similar manner to the cell voltage prediction model generation process. That is, the values ​​of Cell.T.MkCj(t), Ambient.Temp(t), and Fan.ON(t) are periodically input to a predetermined artificial intelligence neural network, and the output value is taught and learned as Cell.T.MkCj(t+1). That is, the values ​​of Cell.T.MkCj(t), Ambient.Temp(t), and Fan.ON(t) are periodically input as input data to the artificial intelligence neural network over a predetermined time period, and the output value is taught and learned so that Cell.T.MkCj(t+1), the measured cell temperature for the next period, is used to generate the cell temperature prediction model.

[0064] The artificial intelligence neural network trained in this way is stored in the cell temperature prediction model 220, and receives inputs of Cell.T.MkCj(t), Ambient.Temp(t), and Fan.ON(t) values ​​to generate Cell.T, which is the predicted value of the cell temperature for the next period. predic Calculate .MkCj(t+1).

[0065] The values ​​of Cell.T.MkCj(t), Ambient.Temp(t), Fan.ON(t), and Cell.T.MkCj(t+1), which are learning data used in the process of generating the cell temperature prediction model, may be field data measured at a time different from the data used in the cell temperature prediction process (S320) described below, or may be sample data obtained before the battery is put into operation or through the operation of a sample battery.

[0066] In another embodiment, the learning data may be generated using values ​​of Cell.T.MkCj(t), Ambient.Temp(t), Fan.ON(t), and Cell.T.MkCj(t+1) for a specified time interval calculated using a standard battery in a laboratory, rather than field data.

[0067] That is, the cell temperature prediction model 220 learns data corresponding to values ​​for predicting the cell temperature calculated from a standard battery, rather than from a battery in operation that calculates the field data, as learning data, and receives input data for predicting the cell temperature from a battery in operation and outputs a predicted value for the cell temperature for the next cycle.

[0068] (3) Prediction model update process (S230)

[0069] In another embodiment, the field data for a predetermined period of time during the normal operation of the battery in operation for which the field data is calculated is added as new learning data, and the cell voltage prediction model and the cell temperature prediction model are re-learned and updated.

[0070] In the process of generating the cell voltage prediction model, values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), Cell.V.MkCj(t), and Cell.V.MkCj(t+1) calculated for a predetermined time period while the battery is not behaving abnormally may be input as new learning data to update the cell voltage prediction model 210. In this case, actual field data is applied to update the learning model, so it is possible to obtain more accurate prediction values ​​for an actual battery deployed in the field.

[0071] In the process of generating the cell temperature prediction model, values ​​of Cell.T.MkCj(t), Ambient.Temp(t), Fan.ON(t), and Cell.T.MkCj(t+1) calculated for a predetermined time period while the battery is not behaving abnormally may be input as new learning data to update the cell temperature prediction model 220. In this case, actual field data is applied to update the learning model (S230), so it is possible to obtain more accurate prediction values ​​for actual batteries deployed in the field.

[0072] 2-3. Real-time prediction process (S300)

[0073] The real-time prediction process is a procedure for predicting the cell voltage and the cell temperature using the learned cell voltage prediction model and the learned cell temperature prediction model.

[0074] (1) Cell voltage prediction process (S310)

[0075] The cell voltage prediction process is performed by inputting the values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t) preprocessed and extracted in the preprocessing process (S120) of field data for predicting the cell voltage into the generated or updated cell voltage prediction model 210 to obtain a cell voltage prediction value of the next period, Cell.V. prdic This is the procedure for calculating .MkCj(t+1).

[0076] (2) Cell temperature prediction process (S320)

[0077] In the cell temperature prediction process, the values ​​of Cell.T.MkCj(t), Ambient.Temp(t), and Fan.ON(t) extracted in the pre-processing process (S220) are input to the generated or updated cell temperature prediction model 220 to obtain Cell.T, which is the predicted value of the cell temperature for the next period. predic Calculate .MkCj(t+1).

[0078] 2-4. Battery Abnormal Behavior Detection Process (S400)

[0079] (1) Cell voltage abnormality detection process (S410)

[0080] The abnormal behavior detection process of the cell voltage is performed by using the predicted value Cell.V output from the cell voltage prediction model 210 through the cell voltage prediction process (S310). prdicThis is the process of comparing the value of .MkCj(t+1) with the value of Cell.V.MkCj(t+1) in the field data calculated from the battery to detect when a difference of a predetermined reference value or more occurs. If a difference of a predetermined reference value or more occurs, an abnormal cell voltage behavior detection signal is generated.

[0081] (2) Cell temperature abnormality behavior detection process (S420)

[0082] This cell temperature abnormality behavior detection process is performed by using Cell.T, which is a predicted value output by the cell temperature prediction model 220 through the cell temperature prediction process (S320). predic This is the process of comparing the value of .MkCj(t+1) with the value of Cell.T.MkCj(t+1) in the field data calculated from the battery to detect when a difference of a predetermined standard value or more occurs. If a difference of a predetermined standard value or more occurs, an abnormal cell temperature behavior detection signal is generated.

[0083] To explain an example of detection based on Figure 4, the field data Cell.T.MkCj(t+1) is shown as "test data" and the predicted value Cell.T predic .MkCj(t+1) is shown as "simulation", and if there is no deviation between the field data and the predicted value that exceeds a predetermined reference value, as in Figure 4(a), it is determined to be normal behavior, and if there is a deviation that exceeds a predetermined reference value, as in Figure 4(b), it is determined to be abnormal battery behavior and detected.

[0084] 2-5. Diagnosis and alarm generation process (S500)

[0085] The diagnosis and alarm generation process is carried out based on the Cell.V predicted in the battery abnormal behavior detection process (S400). prdic .MkCj(t+1), Cell.T predicIf the deviation between the .MkCj(t+1) value and the Cell.V.MkCj(t+1) and Cell.T.MkCj(t+1) values ​​from the measured field data exceeds the reference value, a diagnostic procedure is carried out or an alarm is issued externally.

[0086] In the present invention, the diagnostic procedure is not limited to any particular procedure, and may output an alarm signal or a control signal to shut off the charging or discharging operation of the battery when a deviation equal to or greater than the reference value occurs a predetermined number of times within a predetermined time period.

[0087] 3. Battery Abnormal Behavior Detection System

[0088] A system for detecting abnormal battery behavior according to the present invention will be described with reference to FIG.

[0089] 3-1. Field data calculation unit 100

[0090] The field data calculation unit 100 receives real-time measurement values ​​from a battery in operation, calculates and outputs field data values ​​from the received data. Measurement and calculation of field data may be performed in each cell BMS (Cell Battery Management System), module BMS (Module Battery Management System), or rack BMS (Rack Battery Management System), and sent to the field data calculation unit 100. The field data calculation unit may be configured as an integrated part of the rack BMS. The measurement values ​​received from the battery are values ​​received from each sensor disposed in the battery, and include typical battery status information measurement values. The calculated field data is the field data of the present invention described above.

[0091] The field data calculation unit transmits the field data to the field data preprocessing unit 200 .

[0092] 3-2. Field data preprocessing unit 200

[0093] The field data pre-processing unit 200 includes a cell voltage prediction field data pre-processing unit 210 that extracts field data for predicting the cell voltage from the field data calculated in the field data calculation unit, and a cell temperature prediction field data pre-processing unit 220 that extracts field data for predicting the cell temperature. The field data pre-processing unit 200 performs a field data pre-processing step (S120) for predicting the cell voltage and a field data pre-processing step (S130) for predicting the cell temperature.

[0094] The field data preprocessing unit 200 transmits the training data and the field data to the artificial intelligence neural network unit 300 .

[0095] 5, the learning data may be calculated over a predetermined period of time from a standard battery whose normal quality has been verified in a laboratory, and the calculated data may correspond to each item constituting the field data, and provided to the artificial intelligence neural network unit 300. That is, the cell voltage prediction model 320 constituting the artificial intelligence neural network unit may be trained using field data corresponding values ​​for predicting the cell voltage calculated from the standard battery as learning data, and the cell temperature prediction model 330 may be trained using field data corresponding values ​​for predicting the cell temperature calculated from the standard battery as learning data.

[0096] In another embodiment, the learning data may be configured to be provided to the artificial intelligence neural network unit 300 by extracting field data for a certain period of time during battery operation.

[0097] On the other hand, regardless of which of the two types of data is used as the learning data, the field data pre-processing unit 200 may provide the field data of the battery in the normal operating section as learning data to train the already trained artificial intelligence neural networks 320 and 330.

[0098] 3-3. Artificial Intelligence Neural Network Department 300

[0099] (1) Neural network learning unit 310

[0100] The neural network learning unit 310 receives the above-mentioned learning data from the field data preprocessing unit 200 and learns the artificial intelligence neural network. The artificial intelligence neural network is trained using a cell voltage prediction model 320 and a cell temperature prediction model 330, which are learned and stored in the neural network learning unit to predict the cell voltage and cell temperature, respectively.

[0101] Furthermore, the neural network learning unit 310 may be controlled to add field data during a predetermined period of the battery's normal operating zone as new learning data at a predetermined period or under the control of the control unit 500, and re-learn the artificial intelligence neural network to update the prediction models 320 and 330.

[0102] For example, while the prediction models 320 and 330 are trained and operated using training data provided from a standard battery and generate predicted values ​​in response to input field data, field data from sections in which abnormal behavior is not detected by the battery abnormal behavior detection unit 400 (described later), i.e., normal sections, is added to the training data to retrain the prediction models 320 and 330. This allows the prediction models 320 and 330 to be trained to reflect field data generated from the operating state of an actual field battery, thereby improving the accuracy of predictions made by the prediction models 320 and 330.

[0103] The artificial intelligence neural networks trained using the training data or updated by adding field data are stored in the cell voltage prediction model 320 and the cell temperature prediction model 330, respectively.

[0104] (2) Cell voltage prediction model 320

[0105] The cell voltage prediction model 320 is a trained artificial intelligence neural network that receives field data for predicting the cell voltage from the field data preprocessing unit 200 and predicts the cell voltage for the next period. The field data for predicting the cell voltage may be values ​​of Rack.Current(t), Ambient.Temp(t), Fan.ON(t), SOC.Mk(t), SOH.MkCj(t), and Cell.V.MkCj(t), and receives these inputs and calculates a cell voltage prediction value, Cell.V. prdic Calculate .MkCj(t+1).

[0106] (3) Cell temperature prediction model 330

[0107] The cell temperature prediction model 330 is a trained artificial intelligence neural network that receives field data for predicting the cell temperature from the field data preprocessing unit 200 and predicts the cell temperature for the next period. The field data for predicting the cell temperature may be values ​​of Cell.T.MkCj(t), Ambient.Temp(t), and Fan.ON(t), and receives these inputs to generate Cell.T, which is a predicted value of the cell temperature. predic Calculate .MkCj(t+1).

[0108] 3-4. Battery Abnormal Behavior Detector 400

[0109] The battery abnormal behavior detection unit 400 compares the predicted values ​​of the cell voltage and cell temperature calculated by the cell voltage prediction model 320 and the cell temperature prediction model 330 of the artificial intelligence neural network unit 300 with the cell voltage and cell temperature of the field data to detect abnormal behavior of the battery, and transmits an abnormal behavior detection signal to the control unit. The abnormal behavior detection signal may include at least one or both of a cell voltage abnormal behavior signal and a cell temperature abnormal behavior signal, which will be described later.

[0110] The battery abnormal behavior detection unit 400 receives the field data output from the field data calculation unit and the predicted value output from the artificial intelligence neural network unit 300 .

[0111] (1) Cell voltage abnormal behavior determination unit 410

[0112] The cell voltage abnormal behavior determination unit 410 determines the above-mentioned Cell.V prdic The value of .MkCj(t+1) is compared with the value of Cell.V.MkCj(t+1) of the field data, and if the deviation is greater than a predetermined range, it is determined that the cell voltage is behaving abnormally, and an abnormal cell voltage behavior signal is sent to the control unit.

[0113] (2) Cell temperature abnormal behavior determination unit 420

[0114] The cell temperature abnormal behavior determination unit 420 determines the above-mentioned Cell.T predic The value of .MkCj(t+1) is compared with the value of Cell.T.MkCj(t+1) of the field data, and if the deviation is greater than a predetermined range, it is determined that the cell temperature is behaving abnormally, and an abnormal cell temperature behavior signal is sent to the control unit.

[0115] 3-5. Control unit 500

[0116] The control unit 500 receives the abnormal behavior detection signal from the abnormal behavior detection unit 400 and performs the above-mentioned diagnosis and alarm generation process (S500).

[0117] The control unit may be connected to each of the field data calculation unit, the field data preprocessing unit, the prediction model, and the abnormal behavior detection unit and control each of these components, or in another embodiment, all of the components of the field data calculation unit, the field data preprocessing unit, the prediction model, and the abnormal behavior detection unit may be physically integrated into one control unit. In such a case, the integrated control unit may be directly connected to the battery's RACK BMS so as to be able to communicate with it, or may be integrated into the RACK BMS. [Explanation of symbols]

[0118] The symbols and names used in the drawings of the present invention are as follows:

[0119] 100...Field data calculation unit 200...Field data preprocessing unit 210...cell voltage prediction field data preprocessing unit 220...Cell temperature prediction field data preprocessing unit 300…Artificial Intelligence Neural Network Department 310...Neural Network Learning Unit 320...Cell voltage prediction model 330...Cell temperature prediction model 400...Battery abnormal behavior detection unit 410...Cell voltage abnormal behavior determination unit 420...Cell temperature abnormal behavior determination unit 500...Control unit

Claims

1. 1. A system for detecting abnormal battery behavior, comprising: a field data calculation unit that receives real-time measurements from the battery in operation, calculates field data from the measurements, and outputs the field data; a field data preprocessing unit that extracts and outputs first field data for predicting cell voltages from the field data; receiving, as input, the output of the first field data of the field data pre-processing unit, and using, as learning data, a field data corresponding value for predicting the cell voltage calculated from a standard battery, rather than a battery for calculating the field data; generating a cell voltage prediction model; an artificial intelligence neural network unit that predicts the cell voltage of the next cycle based on the cell voltage prediction model and outputs the predicted value; a battery abnormal behavior detection unit that compares the predicted value output by the artificial intelligence neural network unit with the first field data, and determines that the battery is behaving abnormally when the difference is equal to or greater than a predetermined range; A battery abnormal behavior detection system comprising:

2. The field data preprocessing unit Further extracting and outputting second field data for predicting the cell temperature; The artificial intelligence neural network unit receiving the output of the second field data from the field data preprocessing unit as an input, further predicting the cell temperature for the next period, and outputting the predicted value; The battery abnormal behavior detection unit 2. The system for detecting abnormal battery behavior as described in claim 1, characterized in that the predicted cell voltage value output by the artificial intelligence neural network unit is compared with the cell voltage value in the first field data, and if the difference is greater than or equal to a predetermined range, the system determines that the battery is behaving abnormally, or if the predicted cell temperature value output by the artificial intelligence neural network unit is compared with the cell temperature value in the second field data, and if the difference is greater than or equal to a predetermined range.

3. The artificial intelligence neural network unit a cell voltage prediction model that receives the first field data and outputs a predicted value of the cell voltage; and a neural network learning unit that adds field data during a predetermined period of a normal operating section of the battery as new learning data and re-learns and updates the cell voltage prediction model.

4. The field data preprocessing unit: Further extracting and outputting second field data for predicting the cell temperature; The artificial intelligence neural network unit a cell temperature prediction model that receives the second field data and outputs a predicted value of the cell temperature; The neural network learning unit 4. The system for detecting abnormal behavior of a battery according to claim 3, wherein the second field data for a predetermined period of time during a normal operation section of the battery is added as new learning data to re-learn and update the cell temperature prediction model.

5. The first field data for predicting the cell voltage is time-series values ​​of rack current, atmospheric temperature, fan on / off information, module SOC, cell SOH, and cell voltage for each battery cell and battery module constituting the battery; The second field data for predicting the cell temperature is The system for detecting abnormal behavior of a battery according to claim 2, characterized in that the information is time-series values ​​of cell temperature, ambient temperature, and fan on / off information for each battery cell and battery module constituting the battery.

6. 1. A method for detecting abnormal battery behavior, a field data calculation step of measuring and calculating real-time battery status information data from an operational battery; a field data pre-processing step including a field data pre-processing step for deriving data for predicting cell voltages from the calculated field data; a step of generating a cell voltage prediction model based on the output of the field data pre-processing step, using field data corresponding values ​​for predicting the cell voltage calculated from a standard battery, not the battery for calculating the field data, as learning data; a real-time prediction process including a cell voltage prediction process of inputting data for predicting the cell voltage into a cell voltage prediction model and calculating a cell voltage prediction value for the next period based on the cell voltage prediction model; a battery abnormal behavior detection process including a cell voltage abnormal behavior detection process for comparing the cell voltage prediction value with the cell voltage value of the field data and generating an abnormal cell voltage behavior detection signal when the difference is equal to or greater than a predetermined range; A method for detecting abnormal battery behavior, including:

7. The field data preprocessing step comprises: The method further includes a step of pre-processing field data for predicting a cell temperature, extracting data for predicting a cell temperature from the calculated field data, The real-time prediction process includes: The method further includes a cell temperature prediction step of inputting data for predicting the cell temperature into a cell temperature prediction model to calculate a predicted value of the cell temperature for the next period, The process of detecting abnormal battery behavior includes:

7. The method for detecting abnormal behavior of a battery according to claim 6, further comprising a step of detecting abnormal behavior of a cell temperature, which compares the predicted value of the cell temperature with the cell temperature value of the field data, and generates an abnormal behavior detection signal of the cell temperature when the difference is equal to or greater than a predetermined range.

8. The cell voltage prediction model is a data corresponding value for predicting the cell voltage calculated from a standard battery, rather than a battery in operation for calculating the field data, is learned as learning data, and data for predicting the cell voltage is input from a battery in operation, and a predicted value of the cell voltage for the next period is output; The cell temperature prediction model is The method for detecting abnormal behavior of a battery as described in claim 7, characterized in that a data corresponding value for predicting the cell temperature calculated from a standard battery, rather than a battery in operation for calculating the field data, is learned as learning data, and the data for predicting the cell temperature is received from a battery in operation and a predicted value of the cell temperature for the next period is output.

9. and further including a prediction model updating step of adding field data during a predetermined period of a normal operation section of the battery in operation in which the field data is calculated as new learning data, and re-learning and updating the cell voltage prediction model and the cell temperature prediction model, The real-time prediction process includes:

9. The method for detecting abnormal behavior of a battery according to claim 8, wherein the data for predicting the cell voltage and cell temperature are input into the cell voltage prediction model and the cell temperature prediction model updated through the prediction model update procedure, and predicted values ​​of the cell voltage and cell temperature for the next period are calculated.

10. The data for predicting the cell voltage includes: time-series values ​​of rack current, atmospheric temperature, fan on / off information, module SOC, cell SOH, and cell voltage for each battery cell and battery module constituting the battery; The data for predicting the cell temperature includes: The method for detecting abnormal behavior of a battery according to claim 9, characterized in that the information is time-series values ​​of cell temperature, ambient temperature, and fan on / off information for each battery cell and battery module constituting the battery.

Images