Method and apparatus for diagnosing integrity of energy storage system

Abstract

A method and apparatus for diagnosing the health of an ESS are disclosed. An ESS health diagnosis method according to one embodiment of the present invention may include the steps of: receiving data to be analyzed from each of a plurality of racks included in an Energy Storage System (ESS); inputting the received data to be analyzed into a pre-trained learning model; detecting that there is an abnormality in the ESS when an anomaly score, which is the output value of the learning model and indicates the degree to which the data to be analyzed deviates from normal, exceeds an alarm threshold value; and outputting a health alarm to a user based on the continuity of the detection result.

Description

Technology Field

[0001] The present invention relates to a method and apparatus for diagnosing the health of an Energy Storage System (ESS). More specifically, the present invention relates to a method and apparatus for diagnosing the health of an Energy Storage System (ESS) in a field using minimal data, and capable of detecting abnormalities and outputting alarms regarding health in stages. Background Technology

[0002] A battery pack installed in an electric vehicle, etc., includes at least one battery module, and a battery module may include at least one battery cell. A Battery Management System (BMS) connected to the battery pack and managing the battery overall can detect whether an abnormality has occurred in the battery based on data collected from various components included in the battery pack.

[0003] With the recent continuous occurrence of fires related to Energy Storage Systems (ESS), battery manufacturers are enhancing the internal logic of Battery Management Systems (BMS) for battery safety management; however, there are still cases where BMSs with relatively low reliability are used at installation sites. If a BMS fails to perform its functions properly, the reliability of the data collected by the BMS decreases, which can ultimately lead to battery failure. Therefore, there is a need for a method that can accurately detect abnormalities in the battery and provide step-by-step alarms accordingly. Prior art literature

[0004] Republic of Korea Published Patent Application No. 10-2023-0092116 The problem to be solved

[0005] The technical problem of the present invention is to provide a method and device for diagnosing the health of an ESS.

[0006] Another technical objective of the present invention is to provide an ESS health diagnosis method and device that can perform a relatively accurate diagnosis even when abnormal data occurs due to BMS malfunction by detecting the continuity of abnormal data occurrence and outputting alarms step by step.

[0007] The technical problems to be solved by the present invention are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below. means of solving the problem

[0008] According to one aspect of the present invention, an Energy Storage System (ESS) health diagnosis method may include the steps of: receiving data to be analyzed from each of a plurality of racks included in the ESS; inputting the received data to be analyzed into a pre-trained learning model; detecting that there is an abnormality in the ESS when an anomaly score, which is the output value of the learning model and indicates the degree to which the data to be analyzed deviates from normal, exceeds an alarm threshold value; and outputting a health alarm to a user based on the continuity of the detection result.

[0009] In the above-described ESS health diagnosis method, the step of receiving the data to be analyzed is to classify and receive the data to be analyzed into a first data group corresponding to any one of the cell data included in the plurality of rack units and a second data group corresponding to all of the cell data for each of the racks, and the cell data may include cell voltage, temperature, current, rack voltage, and SOC (State Of Charge).

[0010] In the above ESS health diagnosis method, the step of inputting into the previously trained learning model may be a step of inputting any one of the minimum, maximum, or average values ​​of the received data to be analyzed over a predetermined period of time into the learning model.

[0011] In the above ESS health diagnosis method, the learning model is trained using normal data from which abnormal data has been removed from the battery data by performing preprocessing on the battery data of the ESS that has been stored. The preprocessing can be performed by removing the abnormal data that falls outside the normal operating range of the battery that has been set, and by sequentially applying the Interquartile Range (IQR) and Sigma techniques.

[0012] In the above method for diagnosing the health of an ESS, when the learning is performed on the learning model, the method may include the steps of reducing the dimensionality of the battery data that has undergone the preprocessing and generating K clusters (where K is a natural number) using the K-Means algorithm, measuring the Mahalanobis distance for each of the clusters, setting the average value of the top N values ​​(where N is a natural number) for which the Mahalanobis distance exceeds a predetermined boundary as a predetermined threshold value, and scoring and outputting the degree to which the battery data of the ESS input to the learning model deviates from normal based on the predetermined threshold value.

[0013] In the above ESS health diagnosis method, the alarm generation threshold value may be set to a value that satisfies a predetermined condition among the values ​​obtained by sorting the abnormal scores of the abnormal data and the normal data in ascending order.

[0014] In the above ESS health diagnosis method, the health alarm includes a first stage, a second stage, and a third stage alarm, and the step of outputting the health alarm includes: a step of outputting the first stage alarm when the abnormal score does not exceed the alarm occurrence threshold value; a step of outputting the first stage alarm when the abnormal score exceeds the alarm occurrence threshold value and the continuity index is less than the first standard; a step of outputting the second stage alarm when the abnormal score exceeds the alarm occurrence threshold value and the continuity index is greater than or equal to the first standard and less than the second standard; and a step of outputting the third stage alarm when the abnormal score exceeds the alarm occurrence threshold value and the continuity index is greater than or equal to the second standard, wherein the continuity index is set to 1 when the detection result is not the same as the previous ESS abnormal detection result, and can be set to a value obtained by adding 1 to the previous continuity index when the detection result is the same as the previous ESS abnormal detection result.

[0015] An Energy Storage System (ESS) health diagnosis device according to another aspect of the present invention includes a receiving unit, a control unit, and an output unit that each receive data to be analyzed from a plurality of racks included in the ESS. The control unit inputs the received data to be analyzed into a pre-trained learning model, and if the anomaly score, which is the output value of the learning model and indicates the degree to which the data to be analyzed deviates from normal, exceeds an alarm threshold value, the device detects that there is an abnormality in the ESS and can output a health alarm to a user through the output unit based on the continuity of the detection result.

[0016] In the above-described ESS health diagnosis device, when the receiving unit receives the data to be analyzed, it receives the data to be analyzed by classifying it into a first data group corresponding to any one of the cell data included in the plurality of rack units and a second data group corresponding to all of the cell data for each of the racks, and the cell data may include cell voltage, temperature, current, rack voltage, and SOC (State Of Charge).

[0017] In the above-described ESS health diagnosis device, the learning model is trained using normal data from which abnormal data has been removed from the battery data by performing preprocessing on the battery data of the ESS that has been stored. The preprocessing can be performed by removing the abnormal data that falls outside the normal operating range of the battery that has been set, and by sequentially applying the Interquartile Range (IQR) and Sigma techniques.

[0018] According to another aspect of the present invention, software or a computer-readable medium having executable instructions for performing an ESS health diagnosis method and a computer program stored on said medium may be provided.

[0019] The features briefly summarized above regarding the present invention are merely exemplary aspects of the detailed description of the present invention that follows and do not limit the scope of the present invention. Effects of the invention

[0020] According to the present invention, a method and device for diagnosing the health of an ESS can be provided.

[0021] In addition, according to the present invention, a method and apparatus for diagnosing the health of an ESS can be provided, which can perform a relatively accurate diagnosis even when abnormal data occurs due to a malfunction of the BMS by detecting the continuity of abnormal data occurrence and outputting an alarm step by step.

[0022] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0023] FIG. 1 is a flowchart illustrating a method for diagnosing the health of an ESS according to an embodiment of the present invention. FIG. 2 is a diagram illustrating a method for selecting a data group in diagnosing the health of an ESS according to an embodiment of the present invention. FIG. 3 is a flowchart illustrating the process of performing pretreatment according to one embodiment of the present invention. FIGS. 4 and 5 are drawings for explaining the use of clustering and Mahalanobis distance during model learning according to an embodiment of the present invention. FIG. 6 is a diagram illustrating the ESS health diagnosis process and the resulting effects according to an embodiment of the present invention. FIG. 7 is a drawing for explaining an ESS health diagnosis device according to an embodiment of the present invention. FIG. 8 is a drawing for illustrating a computing device capable of wholly or partially implementing an ESS health diagnosis device according to an embodiment of the present invention. FIG. 9 is a flowchart illustrating the alarm output process based on the ESS health diagnosis result according to one embodiment of the present invention. Specific details for implementing the invention

[0024] Hereinafter, embodiments of the present invention are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein.

[0025] In describing the embodiments of the present invention, if it is determined that a detailed description of known configurations or functions may obscure the essence of the invention, such detailed description is omitted. Additionally, parts of the drawings unrelated to the description of the invention have been omitted, and similar parts are denoted by similar reference numerals.

[0026] In the present invention, when it is stated that a component is "connected," "combined," or "joined" with another component, this may include not only a direct connection but also an indirect connection in which another component exists in between. Furthermore, when it is stated that a component "includes" or "has" another component, this means that, unless specifically stated otherwise, it does not exclude the other component but may include an additional component.

[0027] In the present invention, terms such as first, second, etc. are used solely for the purpose of distinguishing one component from another component and do not limit the order or importance of the components unless specifically stated otherwise. Accordingly, within the scope of the present invention, a first component in one embodiment may be referred to as a second component in another embodiment, and likewise, a second component in one embodiment may be referred to as a first component in another embodiment.

[0028] In the present invention, distinct components are intended to clearly explain their respective features and do not necessarily imply that the components are separated. That is, multiple components may be integrated to form a single hardware or software unit, or a single component may be distributed to form multiple hardware or software units. Therefore, such integrated or distributed embodiments are included within the scope of the present invention, even without separate mention.

[0029] In the present invention, the components described in various embodiments are not necessarily essential, and some may be optional. Accordingly, embodiments composed of a subset of the components described in one embodiment are also included within the scope of the present invention. Furthermore, embodiments including additional components in addition to the components described in various embodiments are also included within the scope of the present invention.

[0030] Hereinafter, an ESS health diagnosis method and apparatus according to an embodiment of the present invention will be described with reference to the drawings.

[0031] A Battery Management System (BMS) is capable of performing functions such as measuring and controlling data like battery voltage, temperature, and current. Since a failure of the BMS to function properly poses a risk of battery failure, the BMS is equipped with algorithms to detect abnormal data. For example, the BMS has a structure that uses internal rule-based mechanisms to trigger an alarm if battery data exceeds specific upper and lower limits.

[0032] However, since it is difficult to accurately identify abnormal data using rule-based diagnosis alone, an advanced detection algorithm utilizing artificial intelligence is required. In other words, as abnormal data can occur even when operating within the rule scope, an advanced detection algorithm capable of determining the presence of data anomalies is required even within that range.

[0033] Furthermore, abnormal data is frequently detected due to BMS malfunctions even when the battery equipment is in good condition. Since classifying such malfunctions based on specific values, as is done in existing alarm generation methods, increases the probability of inaccurate alarms, it is necessary to systematize the alarm generation method.

[0034] According to one embodiment of the present invention, various data provided by the battery are analyzed to detect the presence or absence of abnormalities, and alarms (e.g., Good / Warning / Critical) based on hardware health levels may be provided so that an administrator can efficiently manage the detection results.

[0035] FIG. 1 is a flowchart illustrating a method for diagnosing the health of an ESS according to an embodiment of the present invention.

[0036] The ESS health diagnosis method illustrated in FIG. 1 can be performed by an ESS health diagnosis device described later.

[0037] Referring to FIG. 1, according to an embodiment of the present invention, an ESS health diagnosis method can receive data to be analyzed from each of the multiple racks included in the ESS (Energy Storage System) (S1).

[0038] According to one example, when the above S1 step is performed, the data to be analyzed may be received by classifying it into a first data group corresponding to any one of the cell data included in the plurality of rack units and a second data group corresponding to all of the cell data for each of the plurality of racks. At this time, the cell data may include cell voltage, temperature, current, rack voltage, and SOC (State Of Charge). The details regarding the classification and reception of data groups are described in more detail in FIG. 2.

[0039] FIG. 2 is a diagram illustrating a method for selecting a data group in diagnosing the health of an ESS according to an embodiment of the present invention.

[0040] Referring to FIG. 2, each of the multiple racks (Rack 1, Rack 2, …, Rack N) may include multiple modules (Module 1, Module 2, …, Module N). Additionally, there may be M data corresponding to cell data, such as cell voltage, temperature, current, rack voltage, and SOC, which represent each rack. At this time, the M data, such as cell voltage, temperature, current, and rack voltage, may each be referred to as a tag.

[0041] According to one embodiment, the Rack imbalance data group may be a group that selects data from a plurality of racks (in the case of FIG. 2, Rack 1 to Rack N) as the subject of analysis based on a specific tag (e.g., cell voltage) corresponding to any one of the cell data, and is illustrated by a dotted line in FIG. 2. For example, the Rack imbalance data group may be a cell voltage group from Rack 1 to Rack N or a temperature group from Rack 1 to Rack N. That is, the Rack imbalance data group may be a grouping of any one of the cell data corresponding to each of the plurality of racks.

[0042] In addition, the data corresponding to each tag may be a value that can represent the rack. For example, in the case of cell voltage, it may be any one of the average, minimum, or maximum values ​​of multiple cell voltages within the rack measured over a predetermined period, but is not necessarily limited thereto.

[0043] In addition, the group analysis data group may be a group that selects data corresponding to multiple tags (M tags such as cell voltage, temperature, and current in the case of FIG. 2) as analysis targets based on a specific rack (e.g., Rack 1), and is illustrated by a dashed line in FIG. 2.

[0044] The Rack imbalance data group and the group analysis data group may each refer to the first data group and the second data group described in step S1 of FIG. 1.

[0045] That is, by receiving data to be analyzed by dividing it into two data groups as shown in Fig. 2, or by analyzing the received data to be analyzed by dividing it into two data groups, the ESS health diagnosis method according to one embodiment of the present invention has the effect of enabling more accurate anomaly detection.

[0046] In addition, referring again to FIG. 1, according to an embodiment of the present invention, the ESS health diagnosis method can input data to be analyzed into a pre-trained learning model (S2).

[0047] At this time, according to one embodiment, the input in step S2 may be any one of the minimum value, maximum value, or average value of the data to be analyzed received in step S1 during a predetermined time period.

[0048] In the case of data analysis performed in the ESS health diagnosis method according to one embodiment of the present invention, a structure may be established to set a normal range based on normal data and to determine data outside the normal range as abnormal data. Accordingly, for the generation of a learning model according to one embodiment of the present invention, preprocessing may be performed after removing abnormal data (outliers) from past data. That is, battery data transmitted from the ESS is considered as data including normal data and abnormal data, and a learning model may be generated using only normal data from which abnormal data has been removed.

[0049] The preprocessing process for generating a learning model according to one embodiment may sequentially perform a Rule-based technique, an Interquartile Range (IQR) technique, and a Sigma technique, which are techniques for removing data outside the normal operating range of a battery having a preset upper and lower limit, but is not necessarily limited thereto. The specific preprocessing process will be explained in more detail in FIG. 3.

[0050] FIG. 3 is a flowchart illustrating the process of performing pretreatment according to one embodiment of the present invention.

[0051] Referring to FIG. 3, if there is battery normal operation data related to the data to be analyzed (including normal data and abnormal data) (S10 - 'Yes'), data that falls outside the upper or lower limit of the normal operation range may be identified as abnormal data (Outlier) and removed (S11). On the other hand, if there is no battery normal operation data related to the data to be analyzed (including normal data and abnormal data) or if the user is unaware of it (S10 - 'No'), an IQR technique may be applied so that data that falls outside the standard is considered abnormal data and removed (S20). Alternatively, after the above step S11 is performed, only normal data may be utilized and the above step S20 may be performed sequentially.

[0052] The IQR diagnostic technique is one of the techniques for identifying outliers and is a known technology; therefore, a detailed description of the IQR diagnostic technique is omitted in this specification.

[0053] And, after the above S20 step is performed, the Sigma technique is applied to the normal data after the above S20 step is performed, so that data deviating from the standard is considered abnormal data and can be removed (S30).

[0054] As a result of performing the aforementioned steps S10 to S30, the input data can be classified into normal data (data from which abnormal data has been removed) and abnormal data (outliers), respectively, to be used for creating a learning model.

[0055] Only normal data obtained through the preprocessing process illustrated in FIG. 3 can be utilized to perform the creation and training of a learning model according to one embodiment.

[0056] Specifically, according to one embodiment, a feature with dimensions of the number of racks (N) or the number of tags (M) can be reduced to a smaller number of dimensions by reducing the dimensions of the normal data of the data to be analyzed.

[0057] In addition, distance-based clustering algorithms such as K-means can be applied to the reduced dimension to generate K clusters. Then, specific samples belonging to each cluster can be utilized to measure the Mahalanobis distance. Clustering and Mahalanobis distance measurement are illustrated in FIGS. 4 and 5.

[0058] FIGS. 4 and 5 are drawings for explaining the use of clustering and Mahalanobis distance during model learning according to an embodiment of the present invention.

[0059] In Figure 4, normal data is shown in blue, outlier data is shown in yellow, and it can be seen that a specific cluster has been created in red.

[0060] Referring to Fig. 5, the Mahalanobis distance can be measured based on specific samples in each of the left and right clusters. The distance represented by max on the left side of Fig. 5 may be a threshold value, and thus, in the case of the red sample on the right side of Fig. 5, it exceeds the threshold distance on the left side, and the corresponding data may be determined as anomaly data.

[0061] Regarding the setting of the threshold value above, for each generated cluster, the average of the top P distances (P is a natural number) with the largest reference values ​​for Mahalanobis distance is calculated, and the corresponding average value can be set as the threshold value.

[0062] In addition, based on the Mahalanobis distance from the center point of each cluster, it can be determined that the probability of the corresponding data being anomaly data is higher the greater the value compared to the threshold value. Thus, in order to measure the degree to which the input data differs from normal data, according to one embodiment of the present invention, the degree of difference from normal based on the threshold value can be measured as an indicator called an anomaly score. The anomaly score may have a value from 0 to 100, and the closer it is to 100, the greater the value compared to the threshold value. The anomaly score can be calculated by the following Equation 1.

[0063] Formula 1)

[0064]

[0065] In the above Equation 1, the Top N average distance may refer to the average distance of the top P (P is a natural number) distances with the largest reference values ​​for Mahalanobis distance for each cluster. According to one embodiment of the present invention, the above abnormal score may be used as a reference point for alarm generation.

[0066] Referring again to FIG. 1, according to an embodiment of the present invention, in a method for diagnosing ESS health, if the anomaly score regarding the degree to which the data to be analyzed deviates from normal exceeds an alarm threshold value, it can be detected that there is an abnormality in the ESS (S3).

[0067] In addition, according to an embodiment of the present invention, the ESS health diagnosis method can output a health alarm to the user of the ESS health diagnosis device based on the continuity of the detection result of step S3 (S4).

[0068] That is, according to the present invention, not only is one type of alarm output simply because an abnormality is detected in the ESS, but different alarms can be output according to the health level based on the continuity of the detection results.

[0069] Specifically, an anomaly score can be utilized to set the alarm threshold value, which serves as the reference point for triggering an alarm. The anomaly scores of outlier data removed during the preprocessing stage and normal data used during model training are sorted in ascending order, and the Knee-point of the sorted values ​​can be set as the alarm threshold value. At this time, the setting of the Knee-point may vary depending on the user. The above-described setting of the alarm threshold value is merely an example and is not necessarily limited to what is described above.

[0070] The health alarm step according to one embodiment of the present invention may consist of a total of three steps.

[0071] If the abnormal score for the input data to be analyzed does not exceed the alarm trigger threshold value, it may be detected that there is no abnormality in the ESS, and at this time, a first-stage alarm may be output. The first-stage alarm may, for example, be output through a display in the form of 'Good'.

[0072] In addition, even if the input data subject to analysis exceeds the alarm trigger threshold value, the continuity between the detection result of the previous time unit and the current ESS anomaly detection result is determined, and if the continuity index is less than the first criterion, the first stage alarm ('Good') may be output. A detailed explanation of the continuity index will be provided later.

[0073] Additionally, if the above abnormal score exceeds the alarm trigger threshold value and the above continuity index is greater than or equal to the first criterion and less than the second criterion, the above second stage alarm (for example, 'Warning') may be output. In this case, the above second criterion may be a value greater than the above first criterion.

[0074] In addition, if the above abnormal score exceeds the alarm trigger threshold value and the above continuity index is greater than or equal to the above second criterion, the above third stage alarm (for example, 'Critical') may be output.

[0075] Meanwhile, the above continuity index may refer to an index that can determine whether the same anomaly is detected continuously by comparing the ESS anomaly detection result with the previous anomaly detection result. If the detection result is not the same as the previous ESS anomaly detection result, the continuity index is set to 1, and if the detection result is the same as the previous ESS anomaly detection result, the continuity index may be set to a value obtained by adding 1 to the previous continuity index.

[0076] In the above description, the previous ESS anomaly detection result may refer to the ESS anomaly detection result of a previous time unit based on the time unit in which the ESS anomaly detection result is performed. For example, when ESS anomaly detection according to one embodiment of the present invention is performed on collected data in 15-minute units, a continuity index may be determined by comparing it with the anomaly detection result derived from the immediately preceding 15-minute unit of collected data.

[0078] Each step of the aforementioned alarm output can be output through a display along with sound, and the criterion for continuity can be set differently depending on the user's settings. The process of outputting a health alarm to the user will be described in more detail later in the description of Fig. 9.

[0079] FIG. 6 is a diagram illustrating the ESS health diagnosis process and the resulting effects according to an embodiment of the present invention.

[0080] Referring to Fig. 6, the data to be analyzed received from the field ESS can be classified into RACK imbalance detection analysis and group analysis data, and analysis can be performed on each.

[0081] As a result of the rack imbalance detection analysis, abnormal bank numbers and tags can be identified, and as a result of the group analysis, abnormal bank numbers and rack numbers can be identified; therefore, by combining these, it is possible to identify not only the abnormal bank and rack numbers but also the problematic tags.

[0082] FIG. 7 is a drawing for explaining an ESS health diagnosis device according to an embodiment of the present invention.

[0083] Referring to FIG. 7, an ESS health diagnosis device (100) according to one embodiment of the present invention may include a receiving unit (110), a control unit (120), and an output unit (130), but is not necessarily limited thereto, and may further include unillustrated components for performing ESS health diagnosis.

[0084] The receiving unit (110) can receive data to be analyzed from each of the multiple racks included in the ESS. At this time, when the receiving unit (110) receives the data to be analyzed, it may classify the data to be analyzed into a first data group corresponding to any one of the cell data included in the multiple rack units and a second data group corresponding to all of the cell data for each of the racks. Additionally, the cell data may include cell voltage, temperature, current, rack voltage, and SOC (State Of Charge).

[0085] The first data group mentioned above may refer to the aforementioned rack imbalance data group, and the second data group mentioned above may refer to the aforementioned group analysis data group.

[0086] The control unit (120) can input the data to be analyzed received by the receiving unit (110) into a pre-trained learning model. At this time, when the control unit (120) inputs the data to be analyzed into the pre-trained learning model, it may input any one of the minimum, maximum, or average values ​​of the received data to be analyzed over a predetermined period of time into the learning model.

[0087] In addition, the above learning model may have been trained using normal data from which abnormal data has been removed by performing preprocessing on the battery data of the ESS that has been stored. At this time, when the learning of the above learning model is performed, the dimensionality of the battery data that has been preprocessed is reduced, and K clusters (where K is a natural number) are generated using the K-Means algorithm, and a Mahalanobis distance is measured for each of the clusters, and the average value of the top N values ​​(where N is a natural number) whose Mahalanobis distances exceed a predetermined boundary is set as a predetermined threshold value, and the degree to which the battery data of the ESS input to the above learning model deviates from normal based on the predetermined threshold value is scored and output from the above learning model.

[0088] In relation to the execution of the above preprocessing, it can be performed by removing the abnormal data that falls outside the preset normal operating range of the battery and sequentially applying the Interquartile Range (IQR) and Sigma techniques.

[0089] Additionally, the control unit (120) can detect that there is an abnormality in the ESS when the anomaly score, which is the degree to which the data to be analyzed deviates from normality and is the output value of the learning model, exceeds an alarm threshold value. At this time, the alarm threshold value may be set to a value that satisfies a predetermined condition among the values ​​obtained by sorting the anomaly scores of the abnormal data and the normal data in ascending order.

[0090] The output unit (130) can output a health alarm to the user based on the continuity of the detection results detected by the control unit (120).

[0091] Specifically, when the output unit (130) outputs the health alarm, if the abnormal score does not exceed the alarm generation threshold value, the first stage alarm may be output.

[0092] In addition, when the output unit (130) outputs the health alarm, if the abnormal score exceeds the alarm occurrence threshold value and the continuity index is less than the first standard, the first stage alarm may be output.

[0093] In addition, when the output unit (130) outputs the health alarm, if the abnormal score exceeds the alarm occurrence threshold value and the continuity index is greater than or equal to the first standard and less than the second standard, the second stage alarm may be output.

[0094] In addition, when the output unit (130) outputs the health alarm, if the abnormal score exceeds the alarm generation threshold value and the continuity index is greater than or equal to the second standard, the third stage alarm may be output.

[0095] The above continuity index may be set to 1 if the detection result is not the same as the previous ESS anomaly detection result, and may be set to a value obtained by adding 1 to the previous continuity index if the detection result is the same as the previous ESS anomaly detection result.

[0096] The process of outputting a health alarm to the user will be described in more detail later in the explanation of Fig. 9.

[0098] FIG. 8 is a block diagram of a computing device (800) capable of wholly or partially implementing an ESS health diagnosis device according to one embodiment of the present invention, and may be the ESS health diagnosis device (100) shown in FIG. 7.

[0099] As illustrated in FIG. 8, the computing device (800) includes at least one processor (801), a computer-readable storage medium (802), and a communication bus (803).

[0100] The processor (801) can cause the computing device (800) to operate according to the exemplary embodiment described above. For example, the processor (801) can execute one or more programs stored in a computer-readable storage medium (802). The one or more programs may include one or more computer-executable instructions, and the computer-executable instructions may be configured to cause the computing device (800) to perform operations according to the exemplary embodiment when executed by the processor (801).

[0101] A computer-readable storage medium (802) is configured to store computer-executable instructions or program code, program data and / or other suitable forms of information. A program (802a) stored in the computer-readable storage medium (802) includes a set of instructions executable by a processor (801). In one embodiment, the computer-readable storage medium (802) may be memory (volatile memory such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, other forms of storage media that are accessed by a computing device (800) and capable of storing desired information, or a suitable combination thereof.

[0102] The communication bus (803) interconnects various other components of the computing device (800), including the processor (801) and the computer-readable storage medium (802).

[0103] The computing device (800) may also include one or more input / output interfaces (805) and one or more network communication interfaces (806) that provide interfaces for one or more input / output devices (804). The input / output interfaces (805) and network communication interfaces (806) are connected to a communication bus (803).

[0104] The network communication interface (806) is an interface for in-vehicle communication or an interface for communication between the vehicle and other devices outside the vehicle, and may include, for example, a CAN (Controller Area Network), a MOST (Media Oriented Systems Transport) network, a LIN (Local Interconnect Network) and / or X-by-Wire (Flexray), Wi-Fi, Bluetooth, NFC, RFID, etc. The network may be a cellular network, such as GSM (Global System for Mobile Communications), EDGE (Enhanced Data Rates for GSM Evolution), GPRS (General Packet Radio Service), CDMA (Code Division Multiple Access), Time Division CDMA (TD-CDMA), UMTS (Universal Mobile Telecommunications System), LTE (Long Term Evolution), or any other cellular network.

[0105] An input / output device (804) may be connected to other components of a computing device (800) through an input / output interface (805). An exemplary input / output device (804) may include input devices such as a pointing device (such as a mouse or trackpad), a keyboard, a touch input device (such as a touchpad or touchscreen), a voice or sound input device, various types of sensor devices and / or imaging devices, and / or output devices such as a display device, a printer, a speaker and / or a network card. An exemplary input / output device (804) may be included inside the computing device (800) as a component constituting the computing device (800), or it may be connected to the computing device (800) as a separate device distinct from the computing device (800).

[0106] FIG. 9 is a flowchart illustrating the alarm output process based on the ESS health diagnosis result according to one embodiment of the present invention.

[0107] Referring to FIG. 9, for stepwise alarm output according to the ESS health diagnosis result according to one embodiment of the present invention, it can be checked whether the abnormal score exceeds an alarm occurrence threshold value (S10).

[0108] The alarm stages according to one embodiment of the present invention may be at least three stages, but are not necessarily limited thereto. For example, when the alarm stages are three stages, the three alarm levels may be Good, Warning, and Critical, and the level of warning to the user may increase as it moves from Good to Critical. That is, Warning and Critical are alarms output to the user for the purpose of warning, and Good may be an alarm output to the user simply to provide guidance on the situation. In the following description, it is assumed that the alarm stages according to one embodiment of the present invention are three stages, and that each alarm level is Good, Warning, and Critical. Furthermore, for convenience, the Good, Warning, and Critical alarm levels will be described as the first stage alarm, the second stage alarm, and the third stage alarm, respectively.

[0109] At this time, for the step-by-step alarm output illustrated in FIG. 9, at least two criteria may be used for which a step-by-step alarm can be output according to a continuity index calculated based on the continuity of the detection result. For example, if the alarm stages are Good, Warning, and Critical, there may be two criteria for distinguishing each stage. Hereinafter, the alarm distinguishing criteria according to one embodiment of the present invention are two, and will be described as a first criterion and a second criterion, respectively. At this time, the first criterion may have a smaller value than the second criterion.

[0110] The above alarm distinction criteria allow the user to adjust the alarm occurrence rate according to the present invention by changing the values. For example, the alarm occurrence rate according to the present invention decreases as the first criterion and the second criterion have larger values, and the alarm occurrence rate according to the present invention may increase as the first criterion and the second criterion have smaller values.

[0111] The above continuity index may refer to an index that can determine whether the same anomaly is detected continuously by comparing the ESS anomaly detection result with the previous anomaly detection result.

[0112] According to one embodiment of the present invention, if the abnormal score does not exceed the alarm generation threshold value (S10-'No'), a first stage alarm, a Good alarm level, may be output (S11).

[0113] If the above abnormal score exceeds the alarm trigger threshold value (S10-'Yes'), the identification of the continuity index described above is performed, thereby determining whether to trigger a step alarm (S20). That is, if the abnormal score exceeds the alarm trigger threshold value and a previous ESS abnormality detection result exists (S20-'Yes'), 1 is added to the previously stored continuity index (S21), and if the abnormal score exceeds the alarm trigger threshold value and a previous ESS abnormality detection result does not exist (S20-'No'), the continuity index can be set to 1 regardless of the previous continuity index (S22).

[0114] In the above description, the previous ESS anomaly detection result may refer to the ESS anomaly detection result of a previous time unit based on the time unit in which the ESS anomaly detection result is performed. For example, when ESS anomaly detection according to one embodiment of the present invention is performed on collected data in 15-minute units, a continuity index may be determined by comparing it with the anomaly detection result derived from the immediately preceding 15-minute unit of collected data.

[0115] And, according to one embodiment of the present invention, after a continuity index is determined (S21 or S22), it can be determined whether the continuity index is greater than or equal to a first criterion ('X') (S30).

[0116] If the determined continuity index is less than the first criterion (S30-'No'), the first stage alarm, the Good alarm level, may be output (S31).

[0117] On the other hand, if the determined continuity index is greater than or equal to the first criterion (S30-'Yes'), it can be determined whether the continuity index is greater than or equal to the second criterion ('Y') (S40).

[0118] If the determined continuity index is greater than or equal to the first criterion and less than the second criterion (S40-'No'), a Warning alarm level, which is a second stage alarm, may be output (S41).

[0119] On the other hand, if the determined continuity index is greater than or equal to the second criterion (S40-'Yes'), a third-stage alarm, the Critical alarm level, may be output (S42).

[0120] The description above in relation to Fig. 9 is merely an example, and the number and names of the alarm level steps may be set differently depending on the user.

[0121] The exemplary methods of the present invention are described as a series of operations for clarity of description, but this is not intended to limit the order in which the steps are performed, and if necessary, each step may be performed simultaneously or in a different order. To implement the method according to the present invention, additional steps may be included in addition to the steps exemplified, steps excluding some steps and including the remaining steps, or steps excluding some steps and including additional steps.

[0122] The various embodiments of the present invention are not intended to list all possible combinations but are intended to explain representative aspects of the invention, and the matters described in the various embodiments may be applied independently or in combination of two or more.

[0123] In addition, various embodiments of the present invention may be implemented by hardware, firmware, software, or a combination thereof. In the case of implementation by hardware, it may be implemented by one or more ASICs (Application Specific Integrated Circuits), DSPs (Digital Signal Processors), DSPDs (Digital Signal Processing Devices), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), general processors, controllers, microcontrollers, microprocessors, etc.

[0124] The scope of the present invention includes software or machine-executable instructions (e.g., an operating system, an application, firmware, a program, etc.) that enable an operation according to a method of various embodiments to be executed on a device or computer, and a non-transitory computer-readable medium on which such software or instructions, etc. are stored and which is executable on a device or computer. Explanation of the symbols

[0125] 100: ESS Health Diagnostic Device 110: Receiver 120: Control unit 130: Output section 800: Computing device 801: Processor 802: Computer-readable storage media 802a: program 803: Communications bus 804: Input / Output Devices 805: Input / Output Interface 806: Network communication interface

Claims

Claim 1 A method for diagnosing the health of an Energy Storage System (ESS), comprising: receiving data to be analyzed from each of a plurality of racks included in the ESS; inputting the received data to be analyzed into a pre-trained learning model; detecting that there is an abnormality in the ESS when an anomaly score, which is the output value of the learning model and indicates the degree to which the data to be analyzed deviates from normal, exceeds an alarm threshold value; and outputting a health alarm to a user based on the continuity between the detection result and the detection result of the previous time unit. Claim 2 In claim 1, the step of receiving the data to be analyzed is a step of receiving the data to be analyzed by classifying it into a first data group corresponding to any one of the cell data included in each of the plurality of rack units and a second data group corresponding to all of the cell data for each of the racks, and the cell data includes cell voltage, temperature, current, rack voltage and SOC (State Of Charge). Claim 3 A method for diagnosing ESS health, wherein, in claim 1, the step of inputting into the previously trained learning model is the step of inputting any one of the minimum, maximum, or average values ​​of the received data to be analyzed over a predetermined period of time into the learning model. Claim 4 A method for diagnosing the health of an ESS according to claim 1, wherein the learning model is trained using normal data from which abnormal data has been removed from the battery data by performing preprocessing on the battery data of the ESS that has been stored, and the preprocessing is performed by removing the abnormal data that falls outside the normal operating range of the battery that has been set, and sequentially applying the Interquartile Range (IQR) and Sigma techniques. Claim 5 A method for diagnosing the health of an ESS according to claim 4, wherein, when the learning is performed on the learning model, the method comprises the steps of: reducing the dimensionality of the battery data that has undergone the preprocessing and generating K clusters (where K is a natural number) using the K-Means algorithm; measuring the Mahalanobis distance for each of the clusters; setting the average value of the top N values ​​(where N is a natural number) whose Mahalanobis distance exceeds a predetermined boundary as a predetermined threshold value; and scoring and outputting the degree to which the battery data of the ESS input to the learning model deviates from normal based on the predetermined threshold value. Claim 6 In paragraph 4, the alarm generation threshold value is set to a value that satisfies a predetermined condition among the values ​​obtained by sorting the abnormal scores of the abnormal data and the normal data in ascending order, in an ESS health diagnosis method. Claim 7 A method for diagnosing ESS health, wherein, in claim 1, the health alarm includes a first stage, a second stage, and a third stage alarm, and the step of outputting the health alarm comprises: a step of outputting the first stage alarm when the abnormal score does not exceed the alarm occurrence threshold value; a step of outputting the first stage alarm when the abnormal score exceeds the alarm occurrence threshold value and the continuity index is less than the first standard; a step of outputting the second stage alarm when the abnormal score exceeds the alarm occurrence threshold value and the continuity index is greater than or equal to the first standard and less than the second standard; and a step of outputting the third stage alarm when the abnormal score exceeds the alarm occurrence threshold value and the continuity index is greater than or equal to the second standard, wherein the continuity index is set to 1 when the detection result is not the same as the previous ESS abnormal detection result, and is set to a value obtained by adding 1 to the previous continuity index when the detection result is the same as the previous ESS abnormal detection result. Claim 8 An Energy Storage System (ESS) health diagnosis device comprises: a receiving unit that receives data to be analyzed from each of a plurality of racks included in the ESS; a control unit; and an output unit; wherein the control unit inputs the received data to be analyzed into a pre-trained learning model, and if an anomaly score regarding the degree to which the data to be analyzed deviates from normal, which is the output value of the learning model, exceeds an alarm generation threshold value, the device detects that there is an abnormality in the ESS, and outputs a health alarm to a user through the output unit based on the detection result and the continuity of the detection result from the previous time unit. Claim 9 An ESS health diagnosis device according to claim 8, wherein the receiving unit receives the data to be analyzed by classifying the data to be analyzed into a first data group corresponding to any one of the cell data included in the plurality of rack units and a second data group corresponding to all of the cell data for each of the racks, and wherein the cell data includes cell voltage, temperature, current, rack voltage, and SOC (State Of Charge). Claim 10 In claim 8, the learning model is a device for ESS health diagnosis in which learning is performed using normal data from which abnormal data has been removed from battery data by performing preprocessing on the battery data of the ESS that has been stored, and the preprocessing is performed by removing the abnormal data that falls outside the normal operating range of the battery that has been set, and sequentially applying the Interquartile Range (IQR) and Sigma techniques.

Images