Battery diagnostic device and its operating method

Abstract

A battery diagnostic device according to one embodiment disclosed herein may include: an encoder that converts time-series data related to the state of a battery into latent variables; a decoder that converts the latent variables into first reconstructed data in the time domain; a first discriminator that discriminates between the time-series data and the first reconstructed data and outputs a discrimination result; a first calculator that compares the time-series data and the first reconstructed data and calculates a reconstruction error; a second calculator that calculates an anomaly score based on the discrimination result and the reconstruction error; and a diagnostic device that diagnoses an anomaly in the battery based on the anomaly score and critical values.

Description

Technical Field

[0001] This application claims the benefit of priority based on Korean Patent Application No. 10-2023-0080445, filed on June 22, 2023, and all the contents disclosed in the literature of the Korean Patent Application are incorporated herein by reference in their entirety.

[0002] The embodiments disclosed in this document relate to a battery diagnostic device and an operating method thereof.

Background Art

[0003] Recently, research and development related to secondary batteries have been actively conducted. Here, a secondary battery is a battery capable of charging and discharging, and includes all conventional Ni / Cd batteries, Ni / MH batteries, etc., and recent lithium ion batteries. Among secondary batteries, lithium ion batteries have the advantage of much higher energy density compared to conventional Ni / Cd batteries, Ni / MH batteries, etc. Also, lithium ion batteries can be made small and lightweight, so they are used as power sources for mobile devices, and recently, their usage range has expanded as power sources for electric vehicles, and they have attracted attention as next-generation energy storage media.

[0004] Also, a secondary battery can generally be used as a battery pack including a battery module in which a plurality of battery cells are connected in series and / or in parallel. Also, a secondary battery can be used as a battery rack including a plurality of battery modules and a rack frame for housing such battery modules.

[0005] Such battery cells, battery modules, battery packs, or battery racks can be utilized in various devices. As an example, batteries can be utilized not only in mobile devices such as mobile phones, laptop computers, smartphones, smart pads, etc., but also in fields such as electric vehicles (EV, HEV, PHEV) driven by electricity and large-capacity power storage devices (ESS).

[0006] Such batteries can have their status and operation managed and controlled by a battery management system (BMS). The battery management system can be included with the battery in a single device.

[0007] Furthermore, the battery management system can manage and control the battery while being separated from the device containing the battery. For example, the battery management system can be implemented on a separate server device. In this case, the battery management system can collect battery data and vehicle data from the vehicle, and use the collected data to manage and control the battery. [Overview of the project] [Problems that the invention aims to solve]

[0008] If a short circuit or other type of failure occurs inside the battery, the likelihood of damage to the battery-containing device (e.g., EV, ESS) increases. Therefore, there is a need for a method that can detect abnormal battery conditions and reduce the likelihood of damage to the battery-containing device.

[0009] Traditional long-term battery cycle tests diagnosed battery degradation or abnormalities by comparing the capacity after the test to the capacity before the test. Such tests, requiring at least 100 cycles, were difficult for humans to continuously monitor, and the time-series data collected included sensor noise, leading to numerous misdiagnoses.

[0010] The embodiments disclosed in this document provide a battery diagnostic device and a method for operating it that can diagnose battery abnormalities using an artificial intelligence model that has learned previously unlearned patterns as abnormal patterns, taking into account instantaneous fluctuations in battery state values.

[0011] The embodiments disclosed herein provide a battery diagnostic device and its operating method that can diagnose battery abnormalities based on correlations between battery state variables by learning various variables, not just battery voltage, as input factors.

[0012] The technical problems of the embodiments disclosed herein are not limited to those mentioned above, and other technical problems not mentioned can be clearly understood by those skilled in the art from the following description. [Means for solving the problem]

[0013] A battery diagnostic device according to one embodiment disclosed herein may include: an encoder that converts time-series data related to the state of a battery into latent variables; a decoder that converts the latent variables into first reconstructed data in the time domain; a first discriminator that discriminates between the time-series data and the first reconstructed data and outputs a discrimination result; a first calculator that compares the time-series data and the first reconstructed data and calculates a reconstruction error; a second calculator that calculates an anomaly score based on the discrimination result and the reconstruction error; and a diagnostic device that diagnoses an anomaly in the battery based on the anomaly score and critical values.

[0014] A battery diagnostic device according to one embodiment disclosed herein may further include a second discriminant that determines the performance of the encoder based on the latent variables and random noise in the latent space.

[0015] In a battery diagnostic device according to one embodiment disclosed in this document, the decoder converts random noise in the latent space into second reconstructed data in the time domain, and the first discriminant can output the discriminant result based on a discriminant model that has been trained to discriminate between the time series data and the second reconstructed data.

[0016] In a battery diagnostic device according to one embodiment disclosed in this document, the first calculator can calculate the reconstruction error according to a combined comparison method that compares the time series data and the first reconstruction data over the entire time interval included in the time series data.

[0017] In a battery diagnostic device according to one embodiment disclosed in this document, the first calculator can calculate the reconstruction error in accordance with a segmented comparison method that compares the time series data with the first reconstruction data in a specified portion of the total time intervals included in the time series data.

[0018] In a battery diagnostic device according to one embodiment disclosed in this document, the second calculator can calculate the number of abnormal points based on the following mathematical formula 1. [Mathematical formula 1]

number

[0019] In a battery diagnostic device according to one embodiment disclosed in this document, the second calculator can calculate the number of abnormal points based on the following mathematical formula 2. [Mathematical formula 2]

number

[0020] In a battery diagnostic apparatus according to an embodiment disclosed in this document, the diagnostic device increases the threshold value and compares the number of abnormal points with the threshold value to diagnose an abnormality of the battery. When the threshold value is greater than or equal to a specified value, an optimal threshold value can be determined based on all the threshold values used during the abnormality diagnosis of the battery.

[0021] In a battery diagnostic apparatus according to an embodiment disclosed in this document, the diagnostic device calculates an AUROC (Area Under ROC curve) value corresponding to each of all the threshold values, and can determine the threshold value corresponding to the maximum AUROC value among the AUROC values as the optimal threshold value.

[0022] A battery diagnostic method according to an embodiment disclosed in this document may include an operation of converting time-series data related to the state of a battery into a latent variable, an operation of converting the latent variable into first reconstructed data in the time domain, an operation of discriminating the time-series data and the first reconstructed data to calculate a discrimination result, an operation of comparing the time-series data and the first reconstructed data to calculate a reconstruction error, an operation of calculating an anomaly score based on the discrimination result and the reconstruction error, and an operation of diagnosing an abnormality of the battery based on the anomaly score and a threshold value.

[0023] A battery diagnostic method according to an embodiment disclosed in this document may further include an operation of converting random noise in the latent space into second reconstructed data in the time domain, and the operation of calculating the discrimination result may include an operation of calculating the discrimination result based on a discrimination model learned to discriminate the time-series data and the second reconstructed data.

[0024] In a battery diagnostic method according to one embodiment disclosed herein, the operation for calculating the reconstruction error may include an operation for calculating the reconstruction error according to a combined comparison scheme that compares the time series data and the first reconstruction data over the entire time interval included in the time series data.

[0025] In a battery diagnostic method according to one embodiment disclosed herein, the operation for calculating the reconstruction error may include an operation for calculating the reconstruction error according to a segmented comparison method that compares the time series data with the first reconstruction data in a specified portion of the total time intervals included in the time series data.

[0026] A battery diagnostic method according to one embodiment disclosed herein may further include an operation to diagnose a battery abnormality by increasing the critical value and comparing the number of abnormal points with the critical value, and an operation to determine an optimal critical value based on all critical values ​​used during the battery abnormality diagnosis if the critical value is greater than or equal to a specified value.

[0027] In a battery diagnostic method according to one embodiment disclosed herein, the operation for determining the optimal critical value may include the operation of calculating an AUROC (Area Under ROC curve) value corresponding to each of the critical values, and the operation of determining the critical value corresponding to the maximum AUROC value among the AUROC values ​​as the optimal critical value. [Effects of the Invention]

[0028] According to the embodiments disclosed herein, the manual effort and time required to diagnose battery abnormalities can be reduced.

[0029] According to the embodiments disclosed in this document, it is possible to reduce compensation costs incurred in the field by detecting battery abnormality patterns that were previously undetectable.

[0030] Furthermore, various other effects can be understood directly or indirectly from this document. [Brief explanation of the drawing]

[0031] [Figure 1] This is a block diagram of a battery diagnostic device according to one embodiment. [Figure 2] This figure shows a battery diagnostic system according to one embodiment. [Figure 3] This figure illustrates various examples of how a battery diagnostic system according to one embodiment calculates reconstruction errors. [Figure 4] This is a flowchart showing the operation of a battery diagnostic device according to one embodiment. [Figure 5] This is a flowchart showing the operation of a battery diagnostic device according to one embodiment. [Figure 6] This is a flowchart showing the operation of a battery diagnostic device according to one embodiment. [Modes for carrying out the invention]

[0032] Various embodiments of the present invention are described below with reference to the accompanying drawings. However, this should be understood not as limiting the present invention to any particular embodiment, but rather as including various modifications, equivalents, and / or alternatives to the embodiments of the present invention.

[0033] The various embodiments and terminology used in this document should be understood not to limit the technical features described herein to any particular embodiment, but to include various modifications, equivalents, or substitutions of such embodiments. In relation to the description of the drawings, similar reference numerals may be used for similar or related components. The singular form of a noun corresponding to an item may include one or more of such items unless the context clearly indicates otherwise.

[0034] In this document, each of the phrases such as “A or B,” “at least one of A and B,” “at least one of A or B,” “A, B or C,” “at least one of A, B and C,” and “at least one of A, B, or C” may include any one of the items listed together in the phrase, or any possible combination thereof. Terms such as “first,” “second,” “primary,” “second,” “A,” “B,” “(a),” or “(b)” may be used simply to distinguish one component from other components and, in particular, do not limit the component in other aspects (e.g., importance or order) unless otherwise stated.

[0035] In this document, when a component (e.g., component 1) is referred to as being "coupled," "joined," or "connected" to another component (e.g., component 2), with or without such terms, it means that the component can be connected to the other component directly (e.g., by wire), wirelessly, or via a third component.

[0036] According to various embodiments, each of the above-described components (e.g., modules or programs) may include one or more individuals, some of which may be separated and arranged in other components. According to various embodiments, one or more of the above-described components or operations may be omitted, or one or more other components or operations may be added. Generally or further, multiple components (e.g., modules or programs) may be integrated into a single component. In such cases, the integrated component may perform one or more functions of each of the multiple components in the same or similar manner as they were performed by the components of the multiple components before the integration. According to various embodiments, operations performed by modules, programs or other components may be performed sequentially, in parallel, repeatedly, or heuristically, or one or more of the operations may be performed in a different order, omitted, or one or more other operations may be added.

[0037] Figure 1 is a block diagram of a battery diagnostic device according to one embodiment.

[0038] Referring to Figure 1, the battery diagnostic device 110 can be connected to the electronic device 120 and the user terminal 130 by wire and / or wirelessly.

[0039] The operation of the battery diagnostic device 110, described below, is, of course, performed by the vehicle's BMS (Battery Management System), but can also be performed by various devices such as servers, clouds, chargers, or chargers / dischargers.

[0040] According to one embodiment, the connection 101 between the battery diagnostic device 110 and the electronic device 120 can be a communication connection via a wired and / or wireless network. In one embodiment, the wired network can be based on LAN (local area network) communication or power line communication. In one embodiment, the wireless network can be based on a short-range communication network (e.g., Bluetooth®, WiFi (wireless fidelity), or IrDA (infrared data association)) or a long-range communication network (cellular network, 4G network, 5G network).

[0041] According to other embodiments, the connection 101 between the battery diagnostic device 110 and the electronic device 120 can be a connection via a communication method between devices (e.g., bus, GPIO (general purpose input and output), SPI (serial peripheral interface), or MIPI (mobile industry processor interface)).

[0042] According to one embodiment, the connection 102 between the battery diagnostic device 110 and the user terminal 130 can be a communication connection via a wired and / or wireless network.

[0043] According to one embodiment, the electronic device 120 can be a mobile device (e.g., a mobile phone, a laptop computer, a smartphone, a smartpad), an electric vehicle (e.g., an EV (electric vehicle), an HEV (hybrid EV), a PHEV (plug-in HEV), an FCEV (fuel cell EV)), an energy storage system (ESS), or a battery swapping system (BSS).

[0044] According to one embodiment, the electronic device 120 may include one or more battery units 121, 122, 123. Each of the one or more battery units 121, 122, 123 may be a battery cell, a battery module, a battery pack, or a battery rack.

[0045] According to one embodiment, the user terminal 130 can be a mobile device (e.g., a mobile phone, laptop computer, smartphone, smartpad) or a PC (personal computer). According to one embodiment, information related to the estimation results of the battery diagnostic device 110 can be provided to the user terminal.

[0046] According to one embodiment, the battery diagnostic device 110 may include a communication circuit 111, a sensor 112, a memory 113, and a processor 114. Depending on the embodiment, the battery diagnostic device 110 shown in Figure 1 may further include at least one component other than those shown in Figure 1 (e.g., a display, an input device, or an output device), or at least one component among those shown in Figure 1 (e.g., the sensor 112) may be omitted. For example, if the battery diagnostic device 110 is implemented on an external electronic device separate from the electronic device 120, such as a server or cloud, the battery diagnostic device 110 may use the communication circuit 111 to acquire status information of the battery units 121, 122, or 123, and the battery diagnostic device 110 may not include the sensor 112.

[0047] According to one embodiment, the communication circuit 111 creates a wired communication channel and / or wireless communication channel between the battery diagnostic device 110 and the electronic device 120 and / or user terminal 130, and can send and receive data with the electronic device 120 and / or user terminal 130 via the created communication channel.

[0048] According to one embodiment, the sensor 112 can measure information related to the status of the battery units 121, 122, and 123 of the electronic device 120.

[0049] According to one embodiment, the communication circuit 111 and / or sensor 112 can acquire time-series data related to the state of battery units 121, 122, and / or 123. In one embodiment, the time-series data related to the state of battery units 121, 122, and / or 123 may be data showing the voltage, current, resistance, state of charge (SOC), state of health (SOH), and / or temperature of battery units 121, 122, and / or 123 against time. For example, the time-series data may be univariate data for one of several variables related to the battery state (e.g., voltage, current, resistance, SOC, SOH, or temperature). As another example, the time-series data may be multivariate data for several variables related to the battery state.

[0050] According to one embodiment, the memory 113 may include volatile memory and / or non-volatile memory.

[0051] According to one embodiment, the memory 113 can store data used by at least one component of the battery diagnostic device 110 (e.g., the processor 114). For example, the data may include software (or related instructions), input data, or output data. In one embodiment, the instructions may cause the battery diagnostic device 110 to perform an operation defined by the instructions when executed by the processor 114.

[0052] According to one embodiment, the memory 113 can store a battery diagnostic system (e.g., the battery diagnostic system 200 in Figure 2) that receives time-series data related to the status of the battery units 121, 122, or 123 and outputs abnormality diagnosis results for the battery units 121, 122, or 123.

[0053] According to one embodiment, the processor 114 may include a central processing unit, an application processor, a graphics processing unit, a neural network processing unit (NPU), an image signal processor, a sensor hub processor, or a communication processor.

[0054] According to one embodiment, the processor 114 can execute software stored in memory 113 to control at least one other component (e.g., hardware or software component) of the battery diagnostic device 110 connected to the processor 114, and can perform various data processing or calculations. For example, the processor 114 can execute a battery diagnostic system (e.g., the battery diagnostic system 200 in Figure 2) stored in memory 113 to train the battery diagnostic system, or to diagnose abnormalities in the battery units 121, 122, and / or 123.

[0055] The battery diagnostic system stored in memory 113 can be described in detail below with reference to Figures 2 and 3. The operation of each component included in the battery diagnostic system described below can be performed by various processes or calculations of the processor 114.

[0056] Figure 2 shows a battery diagnostic system according to one embodiment. Figure 3 illustrates various examples of how the battery diagnostic system according to one embodiment calculates the reconstruction error.

[0057] Referring to Figure 2, the battery diagnostic system 200 may include an encoder 210, a latent space 220, a decoder 230, a first discriminant 240, a first calculator 250, a second calculator 260, a diagnostic instrument 270, and / or a second discriminant 280.

[0058] According to one embodiment, some components of the battery diagnostic system 200 (e.g., encoder 210, latent space 220, decoder 230, first discriminator 240, first calculator 250, second calculator 260, and second discriminator 280) can be implemented using a Time Series Anomaly Detection Using Generative Adversarial Networks (TadGAN) model.

[0059] According to one embodiment, the encoder 210 can convert time-series data related to the battery state into latent variables. Here, the latent variables may refer to data stored in a latent space 220 that is potentially shared between the time-series data and the latent variables (i.e., between different domains).

[0060] According to one embodiment, the encoder 210 can operate to separate various features contained in the time-series data onto the latent space 220. For example, if the time-series data is multivariate data for multiple variables related to the battery state, the encoder 210 can separate each of the multiple variables contained in the time-series data and generate multiple latent variables corresponding to each of the multiple variables. According to one embodiment, when the encoder 210 converts the input data into latent variables, it can learn to group similar features together on the latent space 220 and separate features that are not similar.

[0061] According to one embodiment, the decoder 230 can convert data existing in the latent space 220 into reconstructed data in the time domain.

[0062] According to one embodiment, the decoder 230 can convert the latent variable transformed by the encoder 210 into first reconstructed data in the time domain. Here, since the latent variable corresponds to one of several variables included in the time series data, the first reconstructed data can be time series reconstructed data for that one variable.

[0063] According to one embodiment, the decoder 230 can also convert random noise present in the latent space 220 into second reconstructed data in the time domain.

[0064] According to one embodiment, the decoder 230 can learn to generate reconstructed data (e.g., first reconstructed data) that is similar to the original data (e.g., time series data) based on the discrimination result output by the first discriminant 240 and / or the reconstruction error output by the first calculator 250.

[0065] According to one embodiment, the first discriminant 240 can determine whether the data input to the first discriminant 240 is time-series data input to the encoder 210, or whether the original data is first reconstructed data converted by the encoder 210 and decoder 230. For example, the first discriminant 240 can use data obtained by combining time-series data and first reconstructed data as reference data. The first discriminant 240 can compare the input data with the reference data to determine whether the input data is time-series data or first reconstructed data. For example, the time-series data input to the first discriminant 240 may be univariate time-series data corresponding to one variable included in the first reconstructed data among the time-series data input to the encoder 210. In this case, the first discriminant 240 can perform the determination for each of the multiple first reconstructed data corresponding to each of the multiple variables.

[0066] According to one embodiment, the first discriminant 240 can perform the discrimination based on a discriminant model that has been trained to discriminate between time-series data and second reconstructed data converted from random noise by the decoder 230.

[0067] According to one embodiment, the first discriminant 240 can distinguish between time-series data and first reconstructed data and output a discrimination result. For example, the first discriminant 240 can output the similarity between time-series data and first reconstructed data as a discrimination result. According to one embodiment, the first discriminant 240 can distinguish between time-series data and first reconstructed data based on the discrimination model and output a discrimination result.

[0068] In this way, the battery diagnostic system 200 can generate first reconstructed data similar to the time series data by repeating the first reconstructed data generation process by the encoder 210 and decoder 230 and the discrimination process by the first discriminator 240, and the first discriminator 240 can distinguish between the time series data and the first reconstructed data with greater precision. In other words, adversarial learning takes place within the model, and the quality of reconstructed data generation can be improved.

[0069] According to one embodiment, the first calculator 250 can calculate a reconstruction error by comparing time series data with first reconstruction data. Here, the reconstruction error can represent the difference between the time series data and the first reconstruction data. For example, the time series data input to the first calculator 250 can be univariate time series data corresponding to one variable included in the first reconstruction data from the time series data input to the encoder 210. In this case, the first calculator 250 can calculate a reconstruction error for each of the multiple first reconstruction data corresponding to each of the multiple variables.

[0070] According to one embodiment, the first calculator 250 can calculate the reconstruction error by comparing time series data and first reconstructed data over the same time interval. For example, the first calculator 250 can calculate the reconstruction error according to a combined comparison method that compares the time series data and first reconstructed data over the entire time interval included in the time series data. As another example, the first calculator 250 can calculate the reconstruction error according to a split comparison method that compares the time series data and first reconstructed data over a specified portion of the entire time interval included in the time series data. The combined comparison method and split comparison method used by the first calculator 250 can be specifically described below with reference to Figure 3.

[0071] Figure 3 illustrates the original data 310, the reconstructed data 320, some of the original data 330 and 350, and some of the reconstructed data 340 and 360.

[0072] According to one embodiment, the first calculator 250 can calculate the reconstruction error according to a combined comparison method that compares the original data 310 and the reconstructed data 320 over the entire time interval T of the original data 310 and the reconstructed data 320. According to the combined comparison method, the first calculator 250 can calculate the reconstruction error by calculating the difference between the original data 310 and the reconstructed data 320 over the entire time interval T.

[0073] According to one embodiment, the first calculator 250 can calculate the reconstruction error according to a segmented comparison method that compares the original data 330 or 350 and the reconstructed data 340 or 360 over a portion of the time interval T1 or T2 of the original data 310 and the reconstructed data 320.

[0074] For example, the first calculator 250 can extract the original data 330 of the first part corresponding to the first time interval T1 of the entire time interval T from the original data 310, and extract the reconstructed data 340 of the first part corresponding to the first time interval T1 from the reconstructed data 320. The first calculator 250 can compare the original data 330 of the first part and the reconstructed data 340 of the first part to calculate the first reconstruction error. As another example, the first calculator 250 can extract the original data 350 of the second part corresponding to the second time interval T2 of the entire time interval T from the original data 310, and extract the reconstructed data 360 of the second part corresponding to the second time interval T2 from the reconstructed data 320. The first calculator 250 can compare the original data 350 of the second part and the reconstructed data 360 of the second part to calculate the second reconstruction error. According to one embodiment, the first calculator 250 can calculate the first reconstruction error or the second reconstruction error as the reconstruction error. According to another embodiment, the first calculator 250 can calculate the reconstruction error based on the first reconstruction error and the second reconstruction error. For example, the first calculator 250 can calculate the maximum, minimum, or average value of the first reconstruction error and the second reconstruction error as the reconstruction error.

[0075] Furthermore, referring to Figure 2, the second calculator 260 can calculate an anomaly score based on the discrimination result output by the first discriminant 240 and the reconstruction error output by the first calculator 250. According to one embodiment, the second calculator 260 can calculate an anomaly score corresponding to each of the multiple variables included in the time series data.

[0076] According to one embodiment, the second calculator 260 can calculate the abnormal score based on the following mathematical formula 1 or mathematical formula 2.

[0077] [Mathematical formula 1]

number

[0078] [Mathematical formula 2]

number

[0079] (In mathematical formulas 1 and 2, α is the weighted value, Z reconstruction The above reconstruction error, Z Cx (This is the result of the aforementioned determination.)

[0080] According to one embodiment, the diagnostic device 270 can diagnose abnormalities in the battery units 121, 122, and / or 123 based on the number of abnormal points and critical values ​​output by the second calculator 260. For example, the diagnostic device 270 can diagnose abnormalities in the battery units 121, 122, and / or 123 by comparing the number of abnormal points and critical values.

[0081] According to one embodiment, the diagnostic device 270 can diagnose abnormalities in the battery units 121, 122, and / or 123 by increasing the critical value and comparing the number of abnormalities with the critical value. For example, the diagnostic device 270 can repeat the process of increasing the critical value by a predetermined value (e.g., 0.5), comparing the increased critical value with the number of abnormalities, and performing an abnormality diagnosis.

[0082] According to one embodiment, the diagnostic device 270 can identify whether the critical value is greater than or equal to a specified value. According to one embodiment, if the critical value is greater than or equal to a specified value, the diagnostic device 270 can determine the optimal critical value based on all the critical values ​​used during the abnormality diagnosis of the battery units 121, 122, and / or 123. For example, the diagnostic device 270 can calculate an AUROC (Area Under ROC curve) value corresponding to each of the critical values ​​and determine the critical value corresponding to the largest AUROC value among the calculated AUROC values ​​as the optimal critical value.

[0083] According to one embodiment, the second discriminant 280 can determine the performance of the encoder 210. According to one embodiment, the second discriminant 280 can determine the mapping performance of the encoder 210 based on the latent variables transformed by the encoder 210 and random noise on the latent space 220.

[0084] Figure 4 is a flowchart of the operation of a battery diagnostic device according to one embodiment. Figure 4 can be used to explain the operation of the battery diagnostic device 110 of Figure 1, and can be explained using the configuration of Figure 1.

[0085] The embodiment shown in Figure 4 is just one embodiment, and the order of steps in various embodiments of the present invention may differ from that shown in Figure 4. Some of the steps shown in Figure 4 may be omitted, the order of the steps may be changed, or steps may be merged. For example, operation 420 in Figure 4 may be omitted.

[0086] Referring to Figure 4, in operation 405, the battery diagnostic device 110 can convert time-series data related to the battery state into latent variables. Here, latent variables may refer to data stored in a latent space 220 that is potentially shared between the time-series data and the latent variables (i.e., between different domains).

[0087] In operation 410, the battery diagnostic device 110 can convert the latent variables transformed in operation 405 into first reconstruction data in the time domain. Here, since the latent variables correspond to one of several variables included in the time series data, the first reconstruction data can be time series reconstruction data for that one variable.

[0088] In operation 415, the battery diagnostic device 110 can distinguish between time-series data and first reconstructed data and output the discrimination result. For example, the battery diagnostic device 110 can output the similarity between time-series data and first reconstructed data as the discrimination result. According to one embodiment, the battery diagnostic device 110 can distinguish between time-series data and first reconstructed data based on the discrimination model and output the discrimination result. Here, the discrimination model can be specifically explained with reference to Figure 5, which will be described later. For example, the time-series data used in operation 415 can be univariate time-series data corresponding to one variable included in the first reconstructed data among the time-series data used in operation 405. In this case, the battery diagnostic device 110 can perform the discrimination for each of the multiple first reconstructed data corresponding to each of the multiple variables.

[0089] In operation 420, the battery diagnostic device 110 can determine the performance of the encoder 210. According to one embodiment, the battery diagnostic device 110 can determine the mapping performance of the encoder 210 based on the latent variables transformed in operation 405 and random noise on the latent space 220.

[0090] In operation 425, the battery diagnostic device 110 can calculate the reconstruction error. According to one embodiment, the battery diagnostic device 110 can calculate the reconstruction error by comparing the time series data with the first reconstruction data converted in operation 410. Here, the reconstruction error can represent the difference between the time series data and the first reconstruction data. For example, the time series data used in operation 425 may be univariate time series data corresponding to one variable included in the first reconstruction data from the time series data used in operation 405. In this case, the battery diagnostic device 110 can calculate the reconstruction error for each of the multiple first reconstruction data corresponding to each of the multiple variables.

[0091] According to one embodiment, the battery diagnostic device 110 can calculate the reconstruction error by comparing time-series data and first reconstruction data over the same time interval. For example, the battery diagnostic device 110 can calculate the reconstruction error according to a combined comparison method that compares the time-series data and first reconstruction data over the entire time interval included in the time-series data. As another example, the battery diagnostic device 110 can calculate the reconstruction error according to a split comparison method that compares the time-series data and first reconstruction data over a specified portion of the entire time interval included in the time-series data.

[0092] In operation 430, the battery diagnostic device 110 can calculate an anomaly score based on the discrimination result calculated in operation 415 and the reconstruction error calculated in operation 425. According to one embodiment, the battery diagnostic device 110 can calculate an anomaly score corresponding to each of the multiple variables included in the time series data.

[0093] According to one embodiment, the battery diagnostic device 110 can calculate the number of abnormalities based on the mathematical formula 1 or mathematical formula 2.

[0094] In operation 435, the battery diagnostic device 110 can diagnose abnormalities in battery units 121, 122, and / or 123 based on the number of abnormalities and critical values ​​calculated in operation 430. For example, the battery diagnostic device 110 can diagnose abnormalities in battery units 121, 122, and / or 123 by comparing the number of abnormalities and critical values.

[0095] Figure 5 is a flowchart of the operation of a battery diagnostic device according to one embodiment. Figure 5 can be used to explain the operation of the battery diagnostic device 110 of Figure 1, and can be explained using the configuration of Figure 1.

[0096] The embodiment shown in Figure 5 is just one embodiment, and the order of steps in various embodiments of the present invention may differ from that shown in Figure 5. Some of the steps shown in Figure 5 may be omitted, the order of the steps may be changed, or steps may be merged.

[0097] Referring to Figure 5, in operation 505, the battery diagnostic device 110 can convert random noise present in the latent space 220 into second reconstruction data in the time domain.

[0098] In operation 510, the battery diagnostic device 110 can be trained to train a discrimination model to distinguish between time-series data and the second reconstructed data transformed in operation 510.

[0099] Figure 6 is a flowchart of the operation of a battery diagnostic device according to one embodiment. Figure 6 can be used to explain the operation of the battery diagnostic device 110 of Figure 1, and can be explained using the configuration of Figure 1.

[0100] The embodiment shown in Figure 6 is just one embodiment, and the order of steps in various embodiments of the present invention may differ from that shown in Figure 6. Some of the steps shown in Figure 6 may be omitted, the order of the steps may be changed, or steps may be merged.

[0101] Referring to Figure 6, in operation 605, the battery diagnostic device 110 can increase the critical value. For example, the battery diagnostic device 110 can increase the critical value by a predetermined value (e.g., 0.5).

[0102] In operation 610, the battery diagnostic device 110 can diagnose abnormalities in the battery units 121, 122, and / or 123 by comparing the number of abnormal points with critical values.

[0103] In operation 615, the battery diagnostic device 110 can identify whether the critical value increased in operation 605 is greater than or equal to a specified value.

[0104] If, in operation 615, the critical value is identified as being less than the specified value ("NO"), the battery diagnostic device 110 may perform operation 605 again.

[0105] If, in operation 615, the critical value is identified as being greater than or equal to a specified value ("YES"), then in operation 620, the battery diagnostic device 110 can determine the optimal critical value based on all the critical values ​​used during the abnormality diagnosis of battery units 121, 122, and / or 123. For example, the battery diagnostic device 110 can calculate an AUROC value corresponding to each of the aforementioned critical values ​​and determine the critical value corresponding to the largest AUROC value among the calculated AUROC values ​​as the optimal critical value.

[0106] The terms "contains," "constitutes," or "possesses," as used above, should be interpreted as meaning that the component may be inherent, and not as excluding other components, but as potentially including other components, unless otherwise specified. All terms, including technical or scientific terms, have the same meaning as that generally understood by a person of ordinary skill in the art to which the embodiments disclosed herein belong, unless otherwise defined. Commonly used terms, such as those defined in dictionaries, should be interpreted in accordance with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined herein.

Claims

1. An encoder that converts time-series data related to battery status into latent variables, A decoder that converts the aforementioned latent variables into first reconstructed data in the time domain, A first discriminant that distinguishes between the time-series data and the first reconstructed data and outputs the discrimination result, A first calculator that compares the time-series data with the first reconstructed data and calculates the reconstruction error, A second calculator that calculates the number of abnormal points based on the aforementioned discrimination result and the aforementioned reconstruction error, A battery diagnostic device, comprising a diagnostic instrument for diagnosing abnormalities in the battery based on the number of abnormal points and critical values.

2. The battery diagnostic device according to claim 1, further comprising a second discriminant for determining the performance of the encoder based on the latent variables and random noise in the latent space.

3. The decoder converts random noise in the latent space into second reconstructed data in the time domain, The battery diagnostic device according to claim 1, wherein the first discriminant outputs the discriminant result based on a discriminant model that has been trained to discriminate between the time-series data and the second reconstructed data.

4. The battery diagnostic device according to any one of claims 1 to 3, wherein the first calculator calculates the reconstruction error according to a combined comparison method that compares the time series data and the first reconstruction data over the entire time interval included in the time series data.

5. The battery diagnostic device according to claim 1, wherein the first calculator calculates the reconstruction error according to a segmented comparison method that compares the time series data with the first reconstruction data in a specified portion of the total time intervals included in the time series data.

6. The battery diagnostic device according to any one of claims 1 to 3, wherein the second calculator calculates the number of abnormal points based on the following mathematical formula 1. [Mathematical formula 1] [Math 5] (In mathematical formula 1, α is the weighted value, Z ｒｅｃｏｎｓｔｒｕｃｔｉｏｎ The above reconstruction error, Z Ｃｘ (This is the result of the aforementioned determination.)

7. The battery diagnostic device according to any one of claims 1 to 3, wherein the second calculator calculates the number of abnormal points based on the following mathematical formula 2. [Mathematical formula 2] [Math 6] (In mathematical formula 2, α is the weighted value, Z ｒｅｃｏｎｓｔｒｕｃｔｉｏｎ The above reconstruction error, Z Ｃｘ (This is the result of the aforementioned determination.)

8. The aforementioned diagnostic device is The critical value is increased and the number of abnormal points and the critical value are compared to diagnose the abnormality of the battery. A battery diagnostic device according to any one of claims 1 to 3, wherein if the critical value is greater than or equal to a specified value, the optimal critical value is determined based on all critical values ​​used during the abnormality diagnosis of the battery.

9. The aforementioned diagnostic device is The AUROC value corresponding to each of the aforementioned critical values ​​is calculated, The battery diagnostic device according to claim 8, wherein the critical value corresponding to the maximum AUROC value among the AUROC values ​​is determined as the optimal critical value.

10. The process of converting time-series data related to battery status into latent variables, The operation of converting the aforementioned latent variables into first reconstruction data in the time domain, The operation involves distinguishing between the aforementioned time-series data and the first reconstructed data, and calculating the discrimination result. The operation involves comparing the time-series data with the first reconstructed data to calculate the reconstruction error, An operation to calculate the number of abnormal points based on the aforementioned discrimination result and the aforementioned reconstruction error, A battery diagnostic method, comprising the operation of diagnosing an abnormality in the battery based on the number of abnormal points and critical values.

11. This further includes the operation of converting random noise in the latent space into second reconstructed data in the time domain, The battery diagnostic method according to claim 10, wherein the operation for calculating the discrimination result includes the operation for calculating the discrimination result based on a discrimination model that has been trained to distinguish between the time series data and the second reconstructed data.

12. The battery diagnostic method according to claim 10, wherein the operation for calculating the reconstruction error includes an operation for calculating the reconstruction error according to a combined comparison method that compares the time series data and the first reconstruction data over the entire time interval included in the time series data.

13. The battery diagnostic method according to any one of claims 10 to 12, wherein the operation for calculating the reconstruction error includes an operation for calculating the reconstruction error according to a segmented comparison method that compares the time series data with the first reconstruction data in a specified portion of the total time intervals included in the time series data.

14. The operation involves increasing the critical value and comparing the number of abnormal points with the critical value to diagnose an abnormality in the battery, A battery diagnostic method according to any one of claims 10 to 12, further comprising: if the critical value is greater than or equal to a specified value, determining an optimal critical value based on all critical values ​​used during the abnormal diagnosis of the battery.

15. The operation for determining the optimal critical value is as follows: The operation of calculating the AUROC value corresponding to each of the aforementioned critical values, The battery diagnostic method according to claim 14, further comprising the operation of determining a critical value corresponding to the maximum AUROC value among the AUROC values ​​as the optimal critical value.

Images