WO2022032674A1

WO2022032674A1 - Systems and methods for detecting abnormal charging events

Info

Publication number: WO2022032674A1
Application number: PCT/CN2020/109337
Authority: WO
Inventors: Jing Yang; Wei Guan
Original assignee: Beijing Didi Infinity Technology And Development Co., Ltd.
Priority date: 2020-08-14
Filing date: 2020-08-14
Publication date: 2022-02-17
Also published as: CN115315698A

Abstract

Embodiments of the disclosure provide systems and methods for detecting abnormal charging events. The system (100) may include a communication interface (202) configured to receive multivariant charging data of a plurality of charging events (702, 703, 704, 705), the charging data of each charging event (702, 703, 704, 705) comprising a plurality of variants each corresponding to a charging characteristic. The system (100) may further include at least one processor (204). The at least one processor (204) may be configured to determine differences between every two charging events (702, 703, 704, 705) based on the multivariant charging data of the two charging events (702, 703, 704, 705). The at least one processor (204) may be further configured to cluster the plurality of charging events (702, 703, 704, 705) based on the determined differences. The at least one processor (204) may be also configured to detect an abnormal charging event based on the clustering results.

Description

SYSTEMS AND METHODS FOR DETECTING ABNORMAL CHARGING EVENTS

TECHNICAL FIELD

The present disclosure relates to systems and methods for detecting abnormal charging events, and more particularly to systems and methods for detecting abnormal charging events based on clustering a plurality of charging events each characterized by a multivariate time series.

BACKGROUND

Electric power can be supplied by sources such as electric batteries. An electric battery is a device consisting of one or more electrochemical cells with external connections provided to power electrical devices such as mobile phones, flashlights and electric vehicles. When using electric batteries to provide electric power, electrochemical cells generate electrical energy from chemical reactions.

To provide enough electric power for driving large electrical apparatuses such as an electric vehicle, many battery cells are connected in series and/or in parallel to form a battery pack. For safety purposes, a battery management system of the electric vehicle may monitor characteristics of the battery pack and/or cells (e.g., voltage, current, and temperature) in real-time. If value of a battery characteristic meets a predetermined safety threshold under an operating or charging status, the battery management system may send a safety alert to a user. However, real-time monitoring the battery performance may not be sufficient to avoid a serious battery hazard (e.g., fire and/or explosion) . Other methods used big data techniques to analyze battery consistency (e.g., voltage inconsistency among battery cells) . But these methods cannot predict the battery hazards that are not caused by the battery inconsistency issues.

Embodiments of the disclosure address the above problems by providing systems and methods for detecting abnormal charging events based on clustering a plurality of charging events, to detect an abnormal charging event.

SUMMARY

Embodiments of the disclosure provide a system for detecting abnormal charging events. The system may include a communication interface configured to receive multivariant charging data of a plurality of charging events, the charging data of each charging event comprising a plurality of variants each corresponding to a charging characteristic. The system may further include at least one processor. The at least one processor may be configured to determine differences between every two charging events based on the multivariant charging data of the two charging events. The at least one processor may be further configured to cluster the plurality of charging events based on the determined differences. The at least one processor may be also configured to detect an abnormal charging event based on the clustering results.

Embodiments of the disclosure also provide a method for detecting abnormal charging events. The method may include receiving, by a communication interface, multivariant charging data of a plurality of charging events, the charging data of each charging event comprising a plurality of variants each corresponding to a charging characteristic. The method may further include determining, by at least one processor, differences between every two charging events based on the multivariant charging data of the two charging events. The method may also include clustering, by the at least one processor, the plurality of charging events based on the determined differences. The method may additionally include detecting, by the at least one processor, an abnormal charging event based on the clustering results.

Embodiments of the disclosure further provide a non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform a method for detecting abnormal charging events. The method may include receiving multivariant charging data of a plurality of charging events, the charging data of each charging event comprising a plurality of variants each corresponding to a charging characteristic. The method may further include determining differences between every two charging events based on the multivariant charging data of the two charging events. The method may also include the plurality of charging events based on the determined differences. The method may additionally include detecting, by the at least one processor, an abnormal charging event based on the clustering results.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates a schematic diagram of an exemplary system for detecting abnormal charging events, according to embodiments of the disclosure.

FIG. 2 illustrates a block diagram of an exemplary server for detecting abnormal charging events, according to embodiments of the disclosure.

FIG. 3 illustrates a flowchart of an exemplary method for detecting abnormal charging events, according to embodiments of the disclosure.

FIG. 4 illustrates a flowchart of an exemplary method for determining distances between every two charging events, according to embodiments of the disclosure.

FIG. 5 illustrates a flowchart of an exemplary method for clustering charging events, according to embodiments of the disclosure.

FIGs. 6A-6B illustrate two exemplary time series and the corresponding alignment matrix therebetween, according to embodiments of the disclosure.

FIGs. 7A-7E illustrate an exemplary clustering method, according to embodiments of the disclosure.

DETAILED DESCRIPTION

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

Embodiments of the present disclosure provide systems and methods for detecting abnormal charging events based on clustering a plurality of charging events. A charging event can be a process of any electrical devices (e.g., electric vehicles) recharging their rechargeable battery packs from an external source of electricity. Compared to existing solutions, the disclosed systems and methods do not rely on analyzing battery consistency data. Instead, the disclosed systems and methods may cluster charging data (e.g., a battery state of charge (SoC) , a charging current, and/or a charging voltage) of multiple charging events to detect one or more abnormal charging events that have charging behavior different from the others. Detecting these abnormal charging events may be helpful for monitoring the state of the electrical devices (e.g., electric vehicles) and avoiding serious battery hazards.

FIG. 1 illustrates a schematic diagram of an exemplary system 100 for detecting abnormal charging events (referred to as “system 100” hereafter) , according to embodiments of the disclosure. As shown in FIG. 1, system 100 may monitor charging events of vehicles 110 and/or charging stations 120 and detect abnormal charging events from the same. In some embodiments, system 100 may include a database 130, a server 140, and a display device 150. In some embodiments, server 140 may request/download charging data from database 130 through a network (not shown) . The charging data of the battery may be acquired from one or more vehicles 110 and/or charging stations 120, which describe characteristics of the respective charging events. Server 140 may detect abnormal charging events from the charging data and transmit a detection result 103 to a display device 150 for displaying. In some embodiments, server 140 may determine whether further actions such as a service recommendation of replacing at least one cell of the battery pack and/or a further diagnosis of the battery pack are needed, and include the recommendation in detection result 103 to be displayed on display device 150.

Consistent with some embodiments, each vehicle 110 may be an electric vehicle that has an electric motor, or a hybrid vehicle that includes an internal combustion engine and at least one electric motor. Vehicle 110 may have a battery pack for providing electrical power to the electric motor. The battery pack may have more than one cells connected in series and/or in parallel to provide a larger electric power output.

In some embodiments, vehicles 110 may be recharged at charging stations 120. In some embodiments, vehicle 110 may be equipped with sensor (s) (not shown) for detecting/measuring charging data 102 reflecting characteristics of the charging process of the battery pack and/or cells. In some embodiments, the sensor (s) may include an electrical sensor unit, such as a voltage sensor, a current, and/or a temperature sensor for the battery pack and/or cells.

In some embodiments, charging data 102 may be data indicative of the charging characteristics of cells within the battery pack while vehicle 110 is under charging status. For example, charging data 102 may include partial and/or complete data acquired under the control of a battery management system (BMS) which manages the rechargeable battery (e.g., battery pack and/or cells) , for protecting the battery system from operating outside its safe operating zone, monitoring its state, calculating secondary data, etc. It is to be understood that charging data 102 may include certain meta data (e.g., vehicle VIN information) for mapping charging events to the vehicles uploaded charging data.

In some embodiments, charging data 102 may include a multivariate time series (MTS) which has more than one time-varying variable that each corresponds to a charging characteristic (e.g., charging current of the battery pack) . For example, the time-varying variables can include, but not limited to, a battery SoC of the battery pack, a charging current of the battery pack, a BMS required current of the battery pack, a total charging voltage of the battery pack, a maximum charging voltage of the battery cells, and a maximum temperature of the battery cells. A battery SoC indicates the level of charge of an electric battery relative to its capacity. The units of SoC are percentage points (0%= empty; 100%= full) .

Table 115 of FIG. 1 illustrates an exemplary charging data segment that vehicle 110 may produce during a charging event. As shown in table 115, column “voltage1” includes a series of charging voltage values. Current1 may be charging current values of the battery pack corresponding to the charging voltage values under voltage1. Both charging voltages and charging currents may be indexed in time order. Columns “voltage2” and “voltage3” may include charging voltage values of two individual battery cells which can be used to calculate the maximum charging voltage of battery cells at a plurality of time points. In some alternative embodiments, the sensor (s) on vehicle 110 may calculate the maximum voltage of the battery cells so that raw charging data of individual battery cells may not be included in charging data 102. Similarly, the sensor (s) on vehicles 110 may calculate the maximum temperature of the battery cells at a plurality of time points.

Consistent with some embodiments, charging data 102 may be acquired from charging stations 120 as shown in FIG. 1. In some embodiments, at least one of charging stations 120 can be an infrastructure that supplies electric energy for recharging plug-in electric vehicles, including electric vehicles, neighborhood electric vehicles and plug-in hybrids. The charging station may be accessible to multiple electric vehicles and has additional current or connection sensing mechanisms to disconnect the power when an electric vehicle is not charging.

Like vehicles 110, each charging station 120 may be additionally equipped with sensor (s) (not shown) for detecting/measuring charging data 102 of the battery pack and/or cells. For example, current sensors, coupled with charging stations 120, may monitor the power consumed and maintain the connection only if the demand is within a predetermined range. These sensors react more quickly, have fewer parts to fail and are potentially less expensive to design and implement. The sensors can use standard connectors and can help suppliers to monitor or charge for the electricity actually consumed. The sensors may record the charging currents (e.g., actual and BMS required currents) and charging voltages of the battery pack at a plurality of time points.

In some embodiments, some charging stations may not be equipped with a voltage and/or temperature sensor. Those charging stations may download charging voltage and/or temperature data from the sensors on the plug-in vehicles with permission. In some alternative embodiments, the plug-in vehicle may not allow the charging station to download the charging data from the vehicle sensors. As a result, the charging data acquired from the charging station may not the same as those acquired from the vehicle. Because both vehicles 110 and charging stations 120 can generate charging data of the charging event, duplicate data may exist in charging data 102. Charging data 102 may be cleaned/filtered during data pre-processing (also known as data cleaning) . The data cleaning process will be disclosed in greater details below.

In some embodiments, charging data 102 may be stored in a memory and/or a storage coupled to the sensor (s) . For example, charging data 102 may be stored in a format of *. xls, *. xlsx, *. csv, etc. It is to be understood that the format of storing charging data 102 is not limited to the formats disclosed herein and may be modified for other charging purposes.

In some embodiments, charging data 102 may be uploaded to database 130 in real-time (e.g., by streaming from the sensor (s) to database 130) , or collectively after a period of time through a network (not shown) . In some embodiments, the network may be a Wireless Local Area Network (WLAN) , a Wide Area Network (WAN) , wireless networks such as radio waves, a cellular network, a satellite communication network, and/or a local or short-range wireless network (e.g., Bluetooth ^TM or near-field communication) for transmitting charging-related information of the battery of vehicles 110. In some other embodiments, charging data 102 may also be uploaded to database 130 via a direct link (e.g., through a communication cable) . For example, a user of vehicles 110 (i.e., the driver/operator) may periodically drive/direct vehicle 110 to a terminal where database 130 is located for uploading the data.

In some embodiments, server 140 may download charging data 102 from database 130 in real-time through the same and/or a different network through which charging data 102 is uploaded to database 130, or via communication cables for downloading charging data 102 collectively (e.g., every few seconds, every few minutes, etc. ) . In some embodiments, server 140 may process charging data 102 and generate a detection result 103 of abnormal charging events based on the processed charging data 102. Server 140 may monitor the battery pack and determine actions to take to keep the safety and high performance of the battery pack. In some embodiments, the recommended actions may be included in detection result 103. In some embodiments, system 100 may optionally include a display device 150 for displaying detection result 103, e.g., to a manager of charging station 120 and/or user of vehicle 110. It is contemplated that system 100 may include more or less components compared to those shown in FIG. 1.

FIG. 2 illustrates a block diagram of an exemplary server 140 for detecting abnormal charging events (referred to as “server 140” hereafter) , according to embodiments of the disclosure. Consistent with the present disclosure, server 140 may receive charging data 102 and generate detection result 103 indicative of the abnormal charging events based on charging data 102. Although as shown in FIG. 2, server 140 is a physical stand-along apparatus, it is contemplated that in some embodiments, server 140 may be implemented as a cloud software, an application on database 130 and/or display device 150, a virtual server, or a distributed server that is implemented multiple devices. For example, in some embodiments, the charging data cleaning and pre-processing may be implemented by a database management system equipped with database 130 and the remaining functions may be implemented by display device 150. Consistent with the present disclosure, server 140 may be a general-purpose server or a proprietary device specially designed for detecting abnormal charging events.

In some embodiments, as shown in FIG. 2, server 140 may include a communication interface 202 and a processor 204. In some embodiments, server 140 may also include a memory 206, and a storage 208. In some embodiments, server 140 may have different modules in a single device, such as an integrated circuit (IC) chip (implemented as an application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA) ) , or separate devices with dedicated functions. In some embodiments, one or more components of server 140 may be located in a cloud computing environment or may be alternatively in a single location or distributed locations. Components of server 140 may be in an integrated device or distributed at different locations but communicate with each other through a network (not shown) .

Communication interface 202 may receive data (e.g., charging data 102) from database 130 and transmit data (e.g., detection result 103) to display device 150 via communication cables, a Wireless Local Area Network (WLAN) , a Wide Area Network (WAN) , wireless networks such as radio waves, a cellular network, satellite communication links, and/or a local or short-range wireless network (e.g., Bluetooth ^TM) , or other communication methods. In some embodiments, communication interface 202 can be an integrated services digital network (ISDN) card, cable modem, satellite modem, or a modem to provide a data communication connection. As another example, communication interface 202 can be a local area network (LAN) card to provide a data communication connection to a compatible LAN. Wireless links can also be implemented by communication interface 202. In such an implementation, communication interface 202 can send and receive electrical, electromagnetic or optical signals that carry digital data streams representing various types of information via a network.

Consistent with some embodiments, communication interface 202 may further provide the received data to storage 208 for storage or to processor 204 for processing. Communication interface 202 may also receive detection result 103 generated by processor 204 and provide detection result 103 to display device 150.

Processor 204 may include any appropriate type of general-purpose or special-purpose microprocessor, digital signal processor, or microcontroller. Processor 204 may be configured as a separate processor module dedicated to process charging data 102. Alternatively, processor 204 may be configured as a shared processor module for performing other functions unrelated to detecting abnormal charging events (e.g., processor 204 may be a shared processor module on database 130 and/or a share processor module on display device 150) .

As shown in FIG. 2, processor 204 may include multiple modules, such as a data cleaning unit 210, a difference determination unit 212, a data clustering unit 214, an abnormality determination module 216, and the like. These modules (and any corresponding sub-modules or sub-units) can be hardware units (e.g., portions of an integrated circuit) of processor 204 designed for use with other components or software units implemented by processor 204 through executing at least part of a program. The program may be stored on a computer-readable medium, and when executed by processor 204, it may perform one or more functions. Although FIG. 2 shows units 210-216 all within one processor 204, it is contemplated that these units may be distributed among multiple processors located near or remotely with each other.

After receiving charging data 102 from database 130, data cleaning unit 210 may clean and pre-process (e.g., filter) the data. In some embodiments, corrupt, incorrect and/or inaccurate data (i.e., data with errors) may be discarded. For example, data that has predetermined types of error such as having typographical errors and/or formality errors (e.g., repetitions of one or more packets, misplacement of one or more packets, wrong payload format, payloads with empty values, etc. ) may be cleaned by data cleaning unit 210.

In some embodiments, data cleaning unit 210 may identify data segments of charging data 102 acquired from vehicles 110 and/or charging stations 120 and further filter charging data 102 based on the data segments. Consistent with some embodiments, charging data 102 may include duplicate charging data segments. It may happen when both the vehicle and the charging station uploaded charging data of a same charging event. Once duplicate data segments are detected, data cleaning unit 210 may remove the duplicate data segments.

In some embodiments, data cleaning unit 210 may process charging data 102 to obtain multi-variant time series that can be used for detecting abnormal charging events performed by other units of processor 204. For example, charging data 102 may include measured charging information such as temperatures of the battery cells at a plurality of time points, and data cleaning unit 210 may calculate the maximum temperature of the battery cells at each time point based on the measured temperature information.

In some embodiments, difference determination unit 212 may be configured to measure differences (or conversely, similarities) between every two charging events. Consistent with some embodiments, charging data of each charging event may include a multivariate time series (e.g., table 115 of FIG. 1) that includes time-varying data of multiple variants. For example, a charging event may be described using six temporal sequences of charging characteristics. Those charging characteristics can be the battery SoC, the charging current of the battery pack, the BMS required current of the battery pack, the total voltage of the battery pack, the maximum voltage of battery cells, and the maximum temperature of the battery cells. It is contemplated that a charging event can be described using more or less temporal sequences of charging characteristics. Also, the charging event can be described by charging characteristics other than above mentioned and/or together with some of above-mentioned charging characteristics.

In some embodiments, difference determination unit 212 may determine distances between every two charging events. The “distance” measures how similar or different the two charging events are in their respective charging characteristics. In some embodiments, the distance is smaller when the charging characteristics are similar to each other and larger when the charging characteristics are different from each other. In some embodiments, the distance is a collective measure of overall difference among the multiple variants indictive of the multiple charging characteristics.

In some embodiments, vehicles 110 and/or charging stations 120 may be equipped with sensors made by different manufacturers. For example, the sensors may operate differently, e.g., at different sampling frequencies. As a result, charging data 102 acquired from different vehicles and/or charging stations may vary and cannot directly map to each other. In addition, some data values may be missing or erroneous during data transmission. To compensate for these inconsistencies within charging data 102, difference determination unit 212 may use a dynamic time wrapping (DTW) to calculate distances between every two charging events in a charging data space. Generally, DTW is an algorithm for measuring difference/similarity between two temporal sequences (e.g., time series) . The calculation of distances between every two charging events will be described in greater details in connected with FIG. 4 and FIGs. 6A-6B.

In some embodiments, data clustering unit 214 may be configured to cluster the charging events based on the distances determined by difference determination unit 212. In some embodiments, data clustering unit 214 may assign each charging event to two clusters, one corresponding to normal charging and the other corresponding to abnormal charging. It is contemplated that the number of clusters may not be limited to two, but can be a number larger than two, in which case one or more clusters may be corresponding to normal charging and the remaining clusters corresponding to abnormal charging. Data clustering unit 214 may implement any suitable clustering methods to cluster the charging events.

Data clustering unit 214 may start the clustering process by initializing the cluster centers. In some embodiments, when two clusters are used, two charging events may be chosen as initial cluster centers. For example, the two charging events may have the largest distance among all pairs of charging events among the plurality of charging events. In another example, one charging event may be randomly chosen as a first initial cluster center, and a second cluster center may be selected to be another charging event, e.g., based on its distance to the first initial cluster center. For example, the second cluster center may be the charging event having a largest distance from the charging event selected as the first initial cluster center.

After initialing the cluster centers, data clustering unit 214 may be configured to associate the remaining charging events with one of the two initial cluster centers. For example, the assignments may be based on the determined distances calculated by difference determination unit 212. In some embodiments, data clustering unit 214 may dynamically recalculate the cluster centers based on the charging events associated with the respective cluster. Data clustering unit 214 may repeat steps of assigning remaining charging events and recalculating the cluster centers till all the charging events are clustered. Clustering the charging events will be disclosed in greater details in connected with FIG. 5 and FIGs. 7A-7E.

In some embodiments, abnormality determination unit 216 may be configured to detect abnormal charging events based on the clustering results obtained by data clustering unit 214. In some embodiments, abnormality determination unit 216 may label the clusters obtained by data clustering unit 214 as either normal charging or abnormal charging. In some embodiments, when two clusters are obtained, the cluster that has more charging events may be labeled as normal and the other one that has less charging events may be labeled as abnormal. Practically, the normal cluster may contain significantly more charging events than the abnormal cluster. For example, 16 out of 18 charging events may be associated with a first cluster and the remaining 2 charging events are associated with a second cluster, in which case the first cluster is labeled normal and the second is labeled abnormal. In some embodiments, when more than two clusters are obtained, abnormality determination unit 216 may label the clusters based on their cluster centers. For example, a cluster with a cluster center farthest from the other clusters may be determined as abnormal. The charging events associated with the abnormal cluster (s) , by data clustering unit 214, may be determined as abnormal charging events.

After determining the abnormal charging events, abnormality determination unit 216 may send detection results (e.g., detection result 103) to a display device (e.g., display device 150) through communication interface 202. In some embodiments, the display device may be an output device for presentation of information in visual form. For example, display device 150 may be installed on vehicle 110 and/or in charging station 120 for the user of the vehicle or the manager of the charging station to review detection result 103. Detection result 103 may include the cleaned charging data 102 and the corresponding detection labels. A detection label may be a binary value, which indicates if the corresponding charging data is from an abnormal charging event. In some embodiments, display device 150 may graphically display abnormal charging events and normal charging in a same chart. Display device 150 may also graphically display an individual dimension of the charging data. For example, display device 150 may illustrate the battery SoC of the abnormal and normal charging events in temporal sequences with different colors or marks. In some alternative embodiments, only the identification of the abnormal charging events and the corresponding vehicle 110 and/or charging station 120 may be provided to display 150 for display.

In some embodiments, server 140 may further include memory 206 and storage 208. Memory 206 and storage 208 may include any appropriate type of mass storage provided to store any type of information that processor 204 may need to process. Memory 206 and storage 208 may be a volatile or non-volatile, magnetic, semiconductor, tape, optical, removable, non-removable, or other type of storage device or tangible (i.e., non-transitory) computer-readable medium including, but not limited to, a ROM, a flash memory, a dynamic RAM, and a static RAM. Memory 206 and/or storage 208 may be configured to store one or more computer programs that may be executed by processor 204 to detect abnormal charging events disclosed herein. For example, memory 206 and/or storage 208 may be configured to store program (s) that may be executed by processor 204 to clean the charging data, to determine the distances between the cleaned charging data, and/or to cluster the charging data based on the distances.

Memory 206 and/or storage 208 may be further configured to store information and data used by processor 204. For instance, memory 206 and/or storage 208 may be configured to store the various types of data (e.g., the original charging data received from database 130, detection result 103, etc. ) . Memory 206 and/or storage 208 may also store intermediate data such as the cleaned/filtered charging data, the distances, the clustering labels, etc. The various types of data may be stored permanently, removed periodically, or disregarded immediately after certain data segments are processed.

FIG. 3 illustrates a flowchart of an exemplary method 300 for detecting abnormal charging events, according to embodiments of the disclosure. In some embodiments, method 300 may be implemented by system 100. Method 300 may include steps S302-S316 as described below. It is to be appreciated that some of the steps may be optional to perform the disclosure provided herein. Further, some of the steps may be performed simultaneously, or in a different order than shown in FIG. 3.

In step S302, database 130 may receive charging data (e.g., charging data102) from vehicle 110 and/or charging stations 120. For example, database 130 may receive charging data 102 from vehicles 110 in real-time (e.g., by streaming from the sensor (s) to database 130) , or collectively after a period of time through a network (not shown) . In another example, database 130 may receive charging data 102 directly from vehicles 110 or charging stations 120 via a direct link (e.g., through a communication cable) . For example, a user of vehicles 110 (i.e., the driver/operator) may periodically drive/direct one of vehicles 110 to a terminal where database 130 is located for uploading charging data 102. In step S304, database 130 may store the received charging data 102.

In some embodiments, in step S306, server 140 may download the charging data from database 130. For example, server 140 may download/receive the charging data through communication cable or a network (e.g., the same and/or a different network through which charging data 102 is uploaded to database 130) in real-time or collectively (e.g., every few seconds, every few minutes, etc. ) .

In step S308, server 140 may pre-process (e.g., clean/filter) the data (e.g., charging data 102) implemented by data cleaning unit 210 of processor 204. In some embodiments, server 140 may discard/filter the corrupt, incorrect and/or inaccurate data (i.e., data with errors) . In some embodiments of step S308, duplicated charging data may also be filtered by data cleaning unit 210. Consistent with some embodiments, some intermediate data such as the maximum battery cell temperature may be calculated based on the temperature data of the battery cells at each time point.

In step S310, the distances between every two charging events may be calculated by difference determination unit 212 of processor 204. In some embodiments, step S310 may include three sub-steps as shown in FIG. 4. In sub-step S412, standardization may be applied to each charging characteristic data sequence, respectively, to remove the inconsistencies among the acquired charging data. For example, the battery SoC data may be rescaled to have a mean of 0 and a standard deviation of 1 (unit variance) . The charging current data may also be rescaled to have a mean of 0 and a standard deviation of 1 (unit variance) . In some alternative embodiments, normalization method may be used to rescale data instead of standardization method.

Consistent with some embodiments, in sub-step S414, DTW may be used by difference determination unit 212 to construct alignment matrices for every two charging events. For example, as shown in FIG. 6A, line 610 and line 620 are the data plots (variant values v. time) of two charging events. DTW may determine an optimal match (shown in dash lines of FIG. 6A) of data points between line 610 and line 620. The optimal match is associated with a minimal cost, where the cost is computed as a sum of absolute differences between the values of the variant in the respective matched pairs of data points. In some embodiments, the minimal cost becomes a distance between line 610 and line 620. FIG. 6B illustrates an alignment matrix 630 which indicates an optimal match path (gray squares) of line 610 and line 620.

In sub-step S416, the distances between every two charging events may be calculated based on the alignment matrices. For example, a distance between data at time points i and j of charging events t and r can be defined by Equation (1) :

where

1≤i≤L _t, and 1≤j≤L _r. k denotes number of charging characteristics that are included in the charging events t and r. L _t denotes the number of the time points of charging event t, and L _r is the number of the time points of charging event r. w is a predetermined weight for each charging characteristic, which indicates the importance of each charging characteristic in the clustering. For example, a user may pre-set the weight for the charging current to 1.2 and the weights for other charging characteristics to 1.0 to reflect that the charging current is a more important characteristic in determining abnormal charging events.

Using equation (1) , a distance between two charging events D (i, j) can be defined by Equation (2)

where D (1, 1) =d (1, 1) , D (1, 0) =D (2, 0) =…=D (i, 0) =0, D (0, 1) =D (0, 2) =…=D (0, j) =0. The two charging events have i time points and j time points, respectively. The distances between every two charging events can be calculated based on Equations (1) and (2) .

Returning to FIG. 3, in step S312, data clustering unit 214 of processor 204 may be configured to cluster the charging events based on distances obtained in step S310. Consistent with some embodiments, details of clustering are described in sub-steps S512-S522 in FIG. 5. FIGs. 7A-7E illustrate an exemplary clustering method, according to embodiments of the disclosure.

In sub-step S512, data clustering unit 214 may randomly select a charging event as a first initial cluster center as shown in FIG. 7A. For example, six charging events (e.g., events 701-706) in a charging space need to be clustered into two clusters. As shown in FIG. 7A, event 706 (solid square) may be selected randomly as the first cluster center. Data clustering unit 214 may be further configured to select a second charging event based on the distance to the first cluster center. In some embodiments, the charging event with a maximum distance to the first cluster center may be selected as the second cluster center. As shown in FIG. 7B, dash lines illustrate the distances between event 706 to the other events. Because event 701 (solid square) is the event that has a largest distance to event 706, it is selected as the second cluster center.

As another example, data clustering unit 214 may be configured to select two charging events as the initial cluster centers. In some embodiments, the two charging events may have the largest distance among all distances calculated between any two charging events. For example, data clustering unit 214 may rank the differences calculated in step S310, identify the largest distance of all, and select the two charging events that are associated with that distance. For instance, as shown in FIG. 7B, the two charging events having the largest distance are

event

701 and 706.

In sub-step S514, data clustering unit 214 may be configured to associate remaining charging events (e.g., events 702-705 in FIG. 7C) to the nearest cluster center (e.g. event 701 or event 706 in FIG. 7C) . For example, as shown in FIG. 7C, event 702 is assigned to cluster center 701 because it has a shorter distance (dash line) to cluster center 701 than to cluster center 706. As shown in FIG. 7C,

events

702 and 703 are associated to cluster center 701, and

events

704 and 705 are associated to cluster center 706. For illustration purpose only, a solid line is used to separate the two clusters in FIG. 7C.

In sub-step S516, data clustering unit 214 may be configured to calculate the sum of distances of each charging event to the remaining events in the same cluster. For example, as shown in FIG. 7D, the distance between

events

704 and 705 is D45 (dash line) . Similarly, the distance between

events

704 and 706 is D46, and the distance between

events

705 and 706 is as D56. The sum of distances of event 704 to the remaining events (i.e., events 705 and 706) equals to a value of (D45+D46) . The sum of distances of event 705 to the remaining events equals to a value of (D45+D56) . The sum of distances of event 706 to the remaining events equals to a value of (D46+D56) .

In sub-step S518, data clustering unit 214 may be configured to set the charging event with the minimum sum of distances as a new cluster center for the cluster. For example, as shown in FIG. 7D, because the value of (D45+D56) is smaller than either the value of (D45+D46) or the value of (D46+D56) , event 705 is selected as the new cluster center. As shown in FIG. 7D, events 702 and 705 (solid squares) become the new cluster centers for the two clusters.

In sub-step S520, data clustering unit 214 may be configured to update clusters by associating remaining charging events to the nearest new cluster center. For example, as shown in FIG. 7E, event 701 is assigned to new cluster center 702; and

events

703, 704, and 706 are associated to new cluster center 705. Compared with the clustering assignment in FIG. 7D, event 703 changes its cluster label in FIG. 7E, while other events do not change their associated clusters.

In sub-step S522, data clustering unit 214 may be configured to determine if any charging event changes its cluster label. If no charging event changes its cluster label after step S520 compared to after step S514 (sub-step S522: No) , data clustering unit 214 may complete the clustering. If the cluster label of one or more charging events changes (e.g., event 703 changes its cluster label from being associated with the first cluster in FIG. 7D to be associated with the second cluster in FIG. 7E) (sub-step S522: Yes) , sub-steps S516-S522 may be repeated till no charging event changes its cluster label after associating with the nearest new cluster center.

Returning to FIG. 3, in step S314, abnormality determination unit 206 of processor 204 may be configured to detect abnormal charging events based on the clustering results obtained by data clustering unit 214. Consistent with some embodiments, abnormality determination unit 206 may label the clusters obtained by data clustering unit 214 as either normal charging or abnormal charging. For example, the cluster that has more charging events may be labeled as normal and the other one that has less charging events may be labeled as abnormal. In some alternative embodiments, abnormality determination unit 206 may label the clusters based on variances of the two clusters. The variance of the cluster describes how similar the charging events are to each other. Practically, the charging events of the normal cluster may have more similar charging behavior than those of the abnormal cluster.

In step S316, the detection result (e.g., detection result 103) may be displayed on a display device (e.g., display device 150) for a further review. Consistent with some embodiments, display device 150 may graphically display the normal cluster and abnormal cluster. Consistent with some embodiments, only the abnormal charging events and the corresponding vehicle 110 and/or charging station 120 may be provided to display 150 for display.

In some embodiments, detection result 103 obtained by method 300 can be used for training a machine learning model for detecting abnormal charging events. For example, charging data 102 of the detected abnormal charging events and the remaining normal charging events may be used a training data to train the machine learning model. The learning model may be used to detect abnormality from subsequently acquired charging data. For example, the trained machine learning model can be pre-programmed in a vehicle or at a charging station to detect abnormal charging events in real-time.

Another aspect of the disclosure is directed to a non-transitory computer-readable medium storing instructions which, when executed, cause one or more processors to perform the methods, as discussed above. The computer-readable medium may include volatile or non-volatile, magnetic, semiconductor, tape, optical, removable, non-removable, or other types of computer-readable medium or computer-readable storage devices. For example, the computer-readable medium may be the storage device or the memory module having the computer instructions stored thereon, as disclosed. In some embodiments, the computer-readable medium may be a disc or a flash drive having the computer instructions stored thereon.

It will be apparent to those skilled in the art that various modifications and variations can be made to the disclosed system and related methods. Other embodiments will be apparent to those skilled in the art from consideration of the specification and practice of the disclosed system and related methods.

It is intended that the specification and examples be considered as exemplary only, with a true scope being indicated by the following claims and their equivalents.

Claims

A system for detecting abnormal charging events comprising:

a communication interface configured to receive multivariant charging data of a plurality of charging events, the charging data of each charging event comprising a plurality of variants each corresponding to a charging characteristic; and

at least one processor coupled to the communication interface and configured to:

determine differences between every two charging events based on the multivariant charging data of the two charging events;

cluster the plurality of charging events based on the determined differences; and

detect an abnormal charging event based on the clustering results.
The system of claim 1, wherein the multivariant charging data of each charging event comprises a multivariate time series including values of the plurality of variants at a plurality of time points.
The system of claim 1, wherein the plurality of variants correspond to at least two of a battery state of charge (SoC) , a charging current, a required current, a total voltage of a battery pack, a maximum voltage of battery cells, or a maximum temperature of the battery cells.
The system of claim 1, wherein to determine differences between every two charging events, the at least one processor is further configured to:

construct alignment matrices for the two charging events based on the multivariant charging data of the two charging events; and

compute a dynamic time warping (DTW) distance between the two charging events based on the alignment matrices.
The system of claim 4, wherein to compute the DTW distance between the two charging events, the at least one processor is further configured to:

compute multivariate differences each based on one variant of the multivariant charging data; and

compute the DTW distance as a weighted sum of the multivariate differences each weighted by a predetermined weight.
The system of claim 1, wherein to cluster the charging events, the at least one processor is configured to:

select two charging events as initial cluster centers;

associate remaining charging events with a cluster center nearest to it based on the determined differences; and

recalculate the cluster centers using the charging events associated with the respective cluster.
The system of claim 6, wherein to recalculate the cluster centers, the at least one processor is further configured to:

compute a sum of differences of each charging event associated with a cluster to the remaining charging events associated with the same cluster; and

set the charging event with a minimum sum of differences as a new cluster center for the cluster.
The system of claim 6, wherein the two charging events include a first charging event and a second charging event, wherein the difference between the first charging event and the second charging event is the largest among all differences among any two charging events among the plurality of charging events.
The system of claim 1, wherein the multivariant charging data are received from an electric vehicle being charged or a charging station for charging the electric vehicle.
A method of detecting abnormal charging events, comprising:

receiving multivariant charging data of a plurality of charging events, the charging data of each charging event comprising a plurality of variants each corresponding to a charging characteristic;

determining differences between every two charging events based on the multivariant charging data of the two charging events;

clustering the plurality of charging events based on the determined differences; and

detecting an abnormal charging event based on the clustering results.
The method of claim 10, wherein the multivariant charging data of each charging event comprises a multivariate time series including values of the plurality of variants at a plurality of time points.
The method of claim 10, wherein the plurality of variants correspond to at least two of a battery state of charge (SoC) , a charging current, a required current, a total voltage of a battery pack, a maximum voltage of battery cells, or a maximum temperature of the battery cells.
The method of claim 10, wherein determining differences between every two charging events further comprising:

constructing alignment matrices for the two charging events based on the multivariant charging data of the two charging events; and

computing a dynamic time warping (DTW) distance between the two charging events based on the alignment matrices.
The method of claim 13, wherein computing the DTW distance between the two charging events further comprising:

computing multivariate differences each based on one variant of the multivariant charging data; and

computing the DTW distance as a weighted sum of the multivariate differences each weighted by a predetermined weight.
The method of claim 10, wherein clustering the charging events further comprising: selecting two charging events as initial cluster centers;

associating remaining charging events with a cluster center nearest to it based on the determined differences; and

recalculating the cluster centers using the charging events associated with the respective cluster.
The method of claim 15, wherein recalculating the cluster centers further comprising:

computing a sum of differences of each charging event associated with a cluster to the remaining charging events associated with the same cluster; and

setting the charging event with a minimum sum of differences as a new cluster center for the cluster.
The method of claim 15, wherein the two charging events include a first charging event and a second charging event, wherein the difference between the first charging event and the second charging event is the largest among all differences among any two charging events among the plurality of charging events.
The method of claim 10, wherein the multivariant charging data are received from an electric vehicle being charged or a charging station for charging the electric vehicle.
A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform a method for detecting abnormal charging events, the method comprising:

receiving multivariant charging data of a plurality of charging events, the charging data of each charging event comprising a plurality of variants each corresponding to a charging characteristic;

determining differences between every two charging events based on the multivariant charging data of the two charging events;

clustering the plurality of charging events based on the determined differences; and

detecting an abnormal charging event based on the clustering results.
The method of claim 19, wherein the multivariant charging data of each charging event comprises a multivariate time series including values of the plurality of variants at a plurality of time points.