WO2023159638A1 - 电池故障检测***、方法和设备 - Google Patents

电池故障检测***、方法和设备 Download PDF

Info

Publication number
WO2023159638A1
WO2023159638A1 PCT/CN2022/078479 CN2022078479W WO2023159638A1 WO 2023159638 A1 WO2023159638 A1 WO 2023159638A1 CN 2022078479 W CN2022078479 W CN 2022078479W WO 2023159638 A1 WO2023159638 A1 WO 2023159638A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
battery
fault detection
time
real
Prior art date
Application number
PCT/CN2022/078479
Other languages
English (en)
French (fr)
Inventor
刘宏阳
赵微
Original Assignee
宁德时代新能源科技股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 宁德时代新能源科技股份有限公司 filed Critical 宁德时代新能源科技股份有限公司
Priority to PCT/CN2022/078479 priority Critical patent/WO2023159638A1/zh
Priority to CN202280020546.6A priority patent/CN117015774A/zh
Publication of WO2023159638A1 publication Critical patent/WO2023159638A1/zh

Links

Images

Definitions

  • the present application relates to the field of battery technology, in particular to a battery fault detection system, method and device.
  • the battery may often fail after a long period of operation, which will affect the safe operation of new energy vehicles. Therefore, the fault detection of the battery is particularly important.
  • the present application provides a battery fault detection system, method and device, which can take into account the accuracy and real-time performance of battery fault detection.
  • the embodiment of the present application proposes a battery fault detection system, including: a first data warehouse for storing battery historical operation data; a fault detection module for obtaining corresponding battery historical operation data from the first data warehouse Perform feature extraction on the data to obtain corresponding fault detection feature data; batch calculation module: used to perform batch calculation on the fault detection feature data through the batch calculation engine to obtain corresponding intermediate parameters; stream computing platform, used to pass the stream calculation engine The intermediate parameters and the real-time operation data of the battery are acquired in real time for fusion calculation to obtain the fault detection result of the corresponding battery.
  • the first data warehouse can store a large amount of historical operation data, from which the fault detection module can obtain historical operation data with a sufficient time span for feature extraction, and obtain more accurate and reliable fault detection feature data.
  • the batch calculation module performs batch calculation on a large amount of the fault detection characteristic data through the batch calculation engine to quickly obtain the corresponding intermediate parameters, so as to ensure that the subsequent stream computing platform can obtain the intermediate parameters in time when performing real-time fusion calculation.
  • the stream computing platform can obtain intermediate parameters and battery real-time operation data in real time through the stream computing engine for fusion calculation, and obtain the real-time fault detection results of the corresponding battery.
  • the flow computing platform can perform real-time acquisition and calculation processing of large-scale flow data (that is, the real-time operation data of the battery) based on the flow computing engine, and the batch calculation module provides a fast calculation of the intermediate parameters before the calculation processing.
  • Computing capabilities further ensure the real-time execution of computing processing. Therefore, the embodiment of the present application can ensure the real-time performance of fault detection while taking into account the reliability of the detection results, which facilitates timely fault response to the fault risk of the battery, and provides guarantee for the safe operation of the power battery.
  • the system further includes a second data warehouse, the second data warehouse is used to store intermediate parameters and battery real-time operating data in real time for acquisition by the stream computing platform.
  • the system of this embodiment can use the data warehouse to real-time
  • the intermediate parameters calculated by the batch calculation module and the real-time battery operation data obtained from the terminal are stored, so as to provide guarantee for the data reading of the stream computing platform during the stream fusion calculation process.
  • the stream computing platform includes: a data interface, used to collect real-time running data of the corresponding battery from each terminal through the Internet of Things protocol; a first transmission unit, used to transmit the real-time running data to the second data warehouse , to update and obtain the existing battery real-time operation data stored in the second data warehouse; the second transmission unit is used to transmit the real-time operation data to the first data warehouse for storage, so as to update the battery historical operation data.
  • the data platform is based on the Internet of Things protocol, collects real-time battery operation data from each terminal through the data interface, and completes big data collection for batch battery risk identification, thereby facilitating the realization of large-scale battery risk identification management.
  • the obtained real-time operation data is transmitted to the second data warehouse through the first transmission unit to continuously update the existing real-time operation data of the battery, which can continuously provide the data basis for fusion computing for the flow computing platform in real time and improve the risk of large-scale batteries Real-time and reliability of identification management.
  • the real-time operation data is also transmitted to the first data warehouse for storage through the second transmission unit, and the historical operation data of the battery is updated to provide the sequential data basis for the fault detection module to perform feature extraction and accumulation, thereby facilitating the improvement of the fault detection module based on The robustness of massive data calculations, while improving the accuracy of battery fault detection.
  • the fault detection module includes: a data acquisition unit, configured to acquire historical battery operating data corresponding to the sampling period from the first data warehouse at a preset sampling time, and input them into the feature extraction unit according to time sequence;
  • the extraction unit is configured to perform multi-dimensional feature extraction on historical battery operation data through the preset first fault detection model, and output corresponding fault detection feature data.
  • the fault detection module includes a data acquisition unit and a feature extraction unit, wherein the data acquisition unit is used to acquire battery historical operating data from the first data warehouse at a preset sampling time, and input the feature extraction unit according to time sequence, so that The feature extraction unit can perform multi-dimensional feature extraction on the battery historical operating data according to the time series through the preset first fault detection model.
  • the influence of multidimensionality and timing of fault feature sources can be fully integrated, and more accurate fault detection feature data can be obtained, which in turn can help to obtain more reliable fault detection results.
  • the batch calculation module includes: a first calculation unit, configured to perform batch calculations on the fault detection feature data through a batch calculation engine, to obtain corresponding intermediate parameters; a generation unit, configured to generate corresponding time series The intermediate table; the third transmission unit, configured to transmit the intermediate table to the second data warehouse, so as to update the existing intermediate table stored in the second data warehouse.
  • the batch calculation module includes a first calculation unit, a generation unit, and a third transmission unit, wherein the first calculation unit is used to perform a large amount of fault detection feature data obtained from a large amount of historical battery operation data through a batch calculation engine Batch calculations can quickly obtain corresponding intermediate parameters, which is conducive to improving data calculation efficiency in big data processing scenarios.
  • the generation unit is used to generate an intermediate table corresponding to the time series according to the intermediate parameters, and then the third transmission unit transmits the intermediate table to the second data warehouse for storage to be invoked by the stream computing platform.
  • the intermediate parameters are stored in the second data warehouse through the intermediate table before being used for fusion calculation, which can be called faster and more conveniently, reduce the waiting time of real-time calculation, reduce the difficulty of data processing of stream computing, and help improve the real-time fault detection process Computational efficiency in .
  • the flow computing platform further includes: a real-time acquisition module, configured to acquire intermediate parameters and battery real-time operating data from the second data warehouse in real time through the flow computing engine; a determining module, configured to determine the The corresponding battery real-time status data; the splicing module is used to splice the real-time battery operation data and the intermediate parameters in the intermediate table; the calculation module is used to calculate the spliced data through the preset second fault detection model, And output the fault detection data corresponding to the battery and the identity information of the battery.
  • a real-time acquisition module configured to acquire intermediate parameters and battery real-time operating data from the second data warehouse in real time through the flow computing engine
  • a determining module configured to determine the The corresponding battery real-time status data
  • the splicing module is used to splice the real-time battery operation data and the intermediate parameters in the intermediate table
  • the calculation module is used to calculate the spliced data through the preset second fault detection model, And output the fault detection data corresponding to the battery and the
  • the real-time acquisition module of the flow computing platform can continuously obtain the intermediate parameters and real-time battery operating data from the second data warehouse in a flow computing manner through the flow computing engine, and through the determination module, the battery corresponding to the real-time operating data of the battery Real-time status data. Then the splicing module splices the real-time operating data of the battery with the intermediate parameters in the intermediate table to form a new data table corresponding to the time series, which is used for the calculation module to read the data in the data table, and perform fault detection through the preset second fault detection model. Calculate and output the corresponding fault detection data and the identity information of the battery corresponding to the data, so as to obtain the fault detection data corresponding to each battery in the big data battery risk identification management scenario.
  • the system also includes an early warning module and a service subsystem.
  • the early warning module is used to generate early warning information and send it to the service subsystem according to the fault detection data of the corresponding battery and the identity information of the battery; the service subsystem is used to According to the early warning information, the target terminal corresponding to the identity information of the battery is queried to generate service information corresponding to the target terminal and fault detection data.
  • the battery when the battery is faulty or abnormal in the fault detection data, it can generate early warning information and send it to the service subsystem, so that the service subsystem can query the target terminal corresponding to the battery identity information, and send the relevant service information to the service subsystem.
  • the target terminal for users to know, and realize the early warning of the risk of big data battery failure.
  • the embodiment of the present application provides a battery fault detection method, including:
  • the first data warehouse can store a large amount of historical operation data, and based on these large amount of historical operation data with a sufficiently long time span, feature extraction can be performed through the first fault detection model to obtain more accurate fault detection features data. Then, batch calculation is performed on a large amount of the fault detection characteristic data through the batch calculation engine to quickly obtain the corresponding intermediate parameters, so as to ensure that the subsequent flow calculation engine can obtain the intermediate parameters in time when performing real-time fusion calculation.
  • the intermediate parameters and real-time operating data of the battery can be obtained in real time through the stream computing engine, and the second fault detection model is used to perform fusion calculation on these data to obtain the fault detection result of the battery.
  • the embodiment of the present application can ensure the real-time performance of fault detection while taking into account the reliability of the detection results, which facilitates timely fault response to the fault risk of the battery, and provides guarantee for the safe operation of the power battery.
  • the method before using the preset first fault detection model to perform feature extraction on the battery historical operation data, the method further includes: acquiring the battery historical operation sample data within the target duration from the first data warehouse; Multi-dimensional feature extraction is performed on the historical operation sample data, and the first fault detection model corresponding to the fault features is constructed.
  • a large amount of historical operation data within the target duration stored in the first data warehouse can be used as samples (that is, battery historical operation sample data) to perform multi-dimensional feature extraction, and fault features of each dimension can be obtained to construct The corresponding first fault detection model is obtained. Because there are a large amount of historical operating data as training samples, when the constructed data model is applied to feature extraction, more accurate extraction results can be obtained.
  • feature extraction is performed on historical battery operating data through a preset first fault detection model to obtain corresponding fault detection feature data, including: at a preset sampling time, historical battery operating data corresponding to a sampling period according to The time series is input into the first fault detection model; through the first fault detection model, multi-dimensional feature extraction is performed on the battery historical operation data to obtain the corresponding fault detection feature data.
  • the battery historical operation data corresponding to the sampling period is input into the first fault detection model according to the time sequence, so that the first fault detection model can perform multi-dimensional feature extraction on the battery historical operation data according to the time sequence .
  • the influence of multidimensionality and timing of fault feature sources can be fully integrated, and more accurate fault detection feature data can be obtained, which in turn can be beneficial to obtain fault detection results with higher reliability.
  • the method further includes: generating an intermediate table corresponding to the timing according to the intermediate parameters; inputting the intermediate table into the preset first In the second data warehouse, the existing intermediate table stored in the second data warehouse is updated, and the second data warehouse is a real-time data warehouse.
  • the intermediate parameters are stored in the second data warehouse in the form of an intermediate table before being used for fusion calculation, which can be called faster and more conveniently, reducing the waiting time for real-time calculation, and reducing the data processing of streaming calculation difficulty, which is conducive to improving the computational efficiency in the process of real-time fault detection.
  • the intermediate parameters and battery real-time operation data are acquired in real time through the stream computing engine, and the fusion calculation is performed using the second fault detection model to obtain the fault detection result of the corresponding battery, including: through the stream computing engine, from the second The data warehouse acquires intermediate parameters and battery real-time operating data in real time; determines the corresponding battery real-time status data according to the real-time battery operating data; splices the real-time battery status data with the intermediate parameters in the intermediate table; through the preset second fault detection model , calculate the spliced data, and output the fault detection data of the corresponding battery and the identity information of the battery.
  • the intermediate parameters and real-time battery operating data are continuously acquired in real time from the second data warehouse through the stream computing engine, and the corresponding real-time battery status data is continuously updated according to the real-time battery operating data, and then the real-time battery status data is combined with the intermediate
  • the intermediate parameters in the table are spliced to form a new data table corresponding to the time series, based on the support of the stream computing engine, the second fault detection model can continue to read the data in the data table for calculation, and output the fault detection data corresponding to the battery And the identity information of the battery, so that in the big data battery failure risk identification management scenario, the fault detection data corresponding to each battery can be obtained in real time, and the response ability to the battery failure risk can be improved.
  • the method further includes: generating early warning information according to the fault detection data of the corresponding battery and the identity information of the battery; sending the early warning information to the service subsystem, so that the service subsystem According to the early warning information, the system queries the target terminal corresponding to the identity information of the battery, and generates service information corresponding to the target terminal and fault detection data.
  • the battery when the battery is faulty or abnormal in the fault detection data, it can generate early warning information and send it to the service subsystem, so that the service subsystem can query the target terminal corresponding to the battery identity information, and send the relevant service information to the service subsystem.
  • the target terminal for users to know, and realize the early warning of the risk of big data battery failure.
  • the embodiment of the present application provides an electronic device, including a processor, a memory, and a program or instruction stored in the memory and operable on the processor.
  • the program or instruction is executed by the processor, the first aspect is implemented.
  • an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the battery failure detection method as described in any embodiment of the first aspect is implemented A step of.
  • the embodiment of the present application provides a chip, the chip includes a processor and a communication interface, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run programs or instructions to achieve the first aspect The steps of the battery fault detection method described in any embodiment.
  • the embodiment of the present application provides a computer program product.
  • the instructions in the computer program product are executed by the processor of the electronic device, the electronic device executes the battery fault detection method described in any embodiment of the first aspect. step.
  • FIG. 1 is a schematic structural diagram of a battery fault detection system disclosed in an embodiment of the present application
  • Fig. 2 is a schematic diagram of a battery fault detection system performing fault detection in a specific embodiment of the present application
  • Fig. 3 is a schematic flowchart of a battery fault detection method disclosed in an embodiment of the present application.
  • Fig. 4 is a schematic diagram of a hardware structure of an electronic device disclosed by an embodiment of the present application.
  • a conventional solution is generally to shorten the cycle of data collection, so as to reduce the amount of data used for fault detection, thereby improving the real-time performance of fault detection. But this is at the expense of the accuracy of fault detection, because the fault of the power battery has a strong correlation with the timing, and the data in a short period of time often cannot reflect the true state of the battery.
  • the embodiment of the present application provides a battery fault detection system, method and equipment, which can quickly obtain
  • the historical operating data corresponds to the intermediate parameters of the fault detection features
  • the stream computing engine obtains the intermediate parameters and the real-time operating data of the battery in real time for fusion calculation, which can quickly obtain more accurate real-time fault detection results, and can be used in large-scale battery risk identification and management scenarios Among them, it has both the accuracy and real-time performance of the test results.
  • the battery fault detection system provided by the embodiment of the present application will first be described below.
  • FIG. 1 shows a schematic structural diagram of a fault detection system provided by an embodiment of the present application. As shown in Figure 1, the system 100 includes:
  • the first data warehouse 101 is used to store historical battery operation data
  • the fault detection module 102 is configured to obtain corresponding historical battery operation data from the first data warehouse for feature extraction, and obtain corresponding fault detection feature data;
  • Batch calculation module 103 used to perform batch calculation on the fault detection feature data through the batch calculation engine to obtain corresponding intermediate parameters
  • the stream computing platform 104 is used to acquire the intermediate parameters and the real-time operation data of the battery in real time through the stream computing engine for fusion calculation to obtain the fault detection result of the corresponding battery.
  • the first data warehouse 101 can store a large amount of historical operation data, from which the fault detection module 102 can obtain historical operation data with a sufficient time span for feature extraction, and obtain more accurate and reliable fault detection feature data.
  • the batch calculation module 103 performs batch calculation on a large amount of the fault detection feature data through the batch calculation engine to quickly obtain the corresponding intermediate parameters, so as to ensure that the subsequent stream computing platform 104 can obtain the intermediate parameters in time when performing real-time fusion calculation.
  • the flow computing platform 104 can obtain intermediate parameters and real-time battery operation data in real time through the flow computing engine for fusion calculation, and obtain the real-time fault detection result of the corresponding battery.
  • the flow computing platform 104 can perform real-time acquisition and calculation processing of large-scale flow data (that is, battery real-time operation data) based on the flow computing engine, and the batch calculation module 103 provides intermediate parameters before the calculation processing.
  • the fast computing capability further guarantees the real-time execution of computing processing. Therefore, the embodiment of the present application can ensure the real-time performance of fault detection while taking into account the reliability of the detection results, which facilitates timely fault response to the fault risk of the battery, and provides guarantee for the safe operation of the power battery.
  • the first data warehouse 101 may be an offline data warehouse, such as a Hive warehouse.
  • the risk of battery failure also increases.
  • a large amount of historical battery operation data is stored in the offline data warehouse, so that the system can By calculating the historical operation data with a long enough time span, accurate fault detection results can be obtained, and the reliability of fault early warning can be improved.
  • the first data warehouse can store 1 ⁇ 2 years of vehicle power battery operating data to provide sufficient data samples for subsequent fault feature extraction, detection and identification.
  • the historical operation data of the battery may include the voltage value, current value, temperature value, state of charge (State Of Charge, SOC) value, and state of health (State Of Health) value of the battery at historical time, but is not limited thereto.
  • the first data warehouse 101 can clean and normalize the acquired massive battery data and then store it, so as to be called by the fault detection module 102 .
  • the system 100 further includes a second data warehouse 105 , which is used to store intermediate parameters and battery real-time operating data in real time for the stream computing platform 104 to acquire.
  • the system of this embodiment can use the data warehouse
  • the intermediate parameters calculated by the batch calculation module and the real-time battery operation data obtained from the terminal are stored in real time, thereby providing guarantee for the data reading of the stream computing platform in the process of streaming fusion computing.
  • the second data warehouse 105 may be a real-time data warehouse, such as an Hbase warehouse.
  • the second data warehouse 105 can acquire and store each frame of real-time operation data of the battery in real time, and correspondingly update the real-time operation data of the battery for calling by the flow computing platform 104 .
  • the stream computing platform 104 can continuously fuse and calculate the feature data related to the historical operating data of the battery and the real-time operating data of each frame through the stream computing engine, so as to obtain more accurate and reliable real-time fault detection results of the battery, thereby facilitating the realization of Real-time failure risk prediction.
  • the real-time operating data of the battery may include the real-time voltage value, current value, temperature value, state of charge (State Of Charge, SOC) value, and state of health (State Of Health, SOH) value of the battery, but is not limited thereto .
  • the real-time battery state data stored in the second data warehouse 105 may include voltage value, current value, temperature value, SOC value, SOH value and the like.
  • the stream computing platform 104 may include:
  • the data interface is used to collect the real-time operation data of the corresponding battery from each terminal through the Internet of Things protocol;
  • the first transmission unit is used to transmit the real-time operation data to the second data warehouse, so as to update and obtain the existing battery real-time operation data stored in the second data warehouse;
  • the second transmission unit is used to transmit the real-time operation data to the first data warehouse for storage, so as to update the historical operation data of the battery.
  • the stream computing platform 104 may be a processing platform including a Flink stream computing engine 106 (hereinafter referred to as “Flink platform” for short).
  • the Flink platform provides an open-source stream processing framework and a distributed stream computing engine 106, which can execute stream data programs in a data parallel and pipeline manner, so that it can continuously and real-time obtain real-time battery operation data of the terminal, and store the data in the first In the data warehouse 101 and the second data warehouse 105.
  • the new real-time operation data overwrites the existing data, so as to achieve the continuous update of the data.
  • both the new real-time operation data and the existing historical battery operation data are stored in the first data warehouse 101, so that the data can be continuously accumulated to update the overall data.
  • the fault detection module 102 acquires historical operating data from the first data warehouse 101 for feature extraction, obtains corresponding fault detection feature data for batch calculation, and stores intermediate parameters for generating these historical operating data into In the second data warehouse 105, the real-time operation data acquired by the second data warehouse 105 can participate in fusion calculation. At the same time, the real-time operation data acquired this time is stored in the first data warehouse 101 to prepare for the next real-time fault detection.
  • the flow computing engine based on the Flink platform can continuously and orderly obtain the above-mentioned real-time running data and intermediate parameters updated with time from the second data warehouse in the form of flow computing in the process of fusion computing, and perform data fusion calculate.
  • the time series attributes of the historical operation data and real-time operation data of the battery can be preserved, so that the fault detection results are strongly correlated with the time series of the data, which conforms to the objective law of battery operation status changing with time, and improves the accuracy of fault detection results. sex.
  • the Flink platform has the advantages of supporting high throughput, low latency, and high performance, and can ensure real-time data processing in large-scale battery failure detection scenarios.
  • the Flink platform can also efficiently process data according to continuous events, and the better fault tolerance based on the lightweight distributed snapshot (Snapshot) can guarantee the reliability and durability of the Flink platform.
  • the stream computing platform may also be a real-time platform using other types of stream computing engines, so as to realize the above-mentioned functions of the stream computing platform in this application.
  • the Flink platform establishes a communication connection with the terminal based on the IoT protocol of the data interface.
  • the IoT protocol can be the Message Queue Telemetry Transport (MQTT) protocol, the Constrained Application Protocol (CoAP) protocol, the Lightweight Machine-To-Machine (LwM2M) ) protocol, HyperText Transfer Protocol (HyperText Transfer Protocol, HTTP) protocol, narrowband Internet of Things (Narrow Band Internet of Things, NB-IoT) protocol, etc., which are not specifically limited in the embodiment of this application.
  • the terminal monitors the real-time operation data of the battery in real time, and uploads the monitored battery real-time operation data to the Flink platform in real time.
  • the Flink platform After receiving the data uploaded by the terminal, the Flink platform transmits the data to the first data warehouse 101 for storage through the second transmission unit, so as to update the historical battery operation data stored in the first data warehouse 101 .
  • the Flink platform transmits the data received from the terminal to the second data warehouse 105 for storage through the first transmission unit, so as to update the existing battery real-time status data stored in the second data warehouse 105 .
  • the data platform 105 collects real-time operating data of the battery from each terminal through the data interface based on the Internet of Things protocol, completes big data collection for batch battery risk identification, and facilitates the realization of large-scale battery risk identification management.
  • the acquired real-time operation data is transmitted to the second data warehouse 105 through the first transmission unit to continuously update the existing real-time operation data of the battery, thereby continuously providing the stream computing platform 104 with a data basis for fusion computing in real time, improving large-scale Real-time and reliability of battery risk identification management.
  • the real-time operation data is also transmitted to the first data warehouse 101 for storage through the second transmission unit, and the historical operation data of the battery is updated to provide the time-sensitive data basis for the fault detection module 102 to perform feature extraction and accumulation, which is beneficial to improve the fault detection module.
  • the robustness of calculations based on massive data is conducive to improving the accuracy of battery fault detection.
  • the fault detection module 102 may include:
  • the data acquisition unit is used to acquire the battery historical operating data corresponding to the sampling period from the first data warehouse 101 at the preset sampling time, and input them into the feature extraction unit according to the time sequence;
  • the feature extraction unit is configured to perform multi-dimensional feature extraction on the battery historical operating data through the preset first fault detection model, and output corresponding fault detection feature data.
  • the sampling period can be measured by time such as year, month, day, hour, minute, etc.
  • the preset sampling time can also be set by measuring intervals such as day, hour, minute, and second.
  • the fault detection module 102 obtains the historical battery operation data within the corresponding sampling period from the first data warehouse 101 at the preset sampling time through the data acquisition unit, and inputs the time sequence of the battery historical operation data into the feature extraction unit according to the time of the acquired battery historical operation data In , multi-dimensional feature extraction is performed through the preset first fault detection model to obtain fault detection characteristic data.
  • the feature extraction unit may first obtain enough samples of battery historical operating data from the first data warehouse 101, and extract voltage characteristics, temperature characteristics, current characteristics and corresponding fault codes (such as thermal runaway fault codes), etc., to build a multi-dimensional first fault detection model.
  • the first fault detection model can be used to obtain historical operating data of the battery for prediction and identification, and output fault detection characteristic data such as corresponding fault codes.
  • the constructed first fault detection model may be a matrix model, a neural network model, or a decision tree model, which is not limited in this embodiment.
  • the feature extraction unit acquires historical battery operation data within the sampling period for multi-dimensional feature extraction and identification, and outputs fault detection characteristic data.
  • the influence of multidimensionality and timing of fault feature sources can be fully integrated, and more accurate fault detection feature data can be obtained, which in turn can be beneficial to obtain fault detection results with higher reliability.
  • the batch calculation module 103 may specifically include:
  • the first calculation unit is used to perform batch calculation on the fault detection feature data through a batch calculation engine to obtain corresponding intermediate parameters
  • a generation unit configured to generate an intermediate table corresponding to the timing according to the intermediate parameters
  • the third transmission unit is used to transmit the intermediate table to the second data warehouse, so as to update the existing intermediate table stored in the second data warehouse.
  • the batch computing engine may be a Spark batch computing engine.
  • the Spark batch calculation engine has the advantages of being fast and versatile. It can perform batch calculations on massive fault detection feature data, obtain corresponding intermediate parameters, and generate intermediate tables corresponding to time series.
  • the intermediate parameters may include battery number id, detection time consumption, report time, current value, voltage value, SOC value, SOH value, fault code, battery status, etc., and store this data generation intermediate table in the first In the second data warehouse 105, it is convenient for the stream computing platform 104 to call when performing real-time fault detection, reducing the complexity of data calculation during stream computing.
  • the system 100 in the scenario where the big data platform performs battery failure risk warning, the system 100 usually needs to process the historical operation data of hundreds of thousands or even millions of vehicles for 1 to 2 years, and the amount of processed data reaches Pb level. If the big data platform in related technologies is used to process such a huge amount of data, the processing speed will be very slow, and it usually takes days or even weeks to process it. The response time is more than 1 day, so if the battery fails within 2 days after the early warning is triggered in practical applications, the related technology cannot meet the early warning requirements of failure risks.
  • the system 100 regularly obtains sufficient historical operating data of the battery from the offline first data warehouse 101 for fault detection by passing the fault detection module 102 in advance, and obtains the fault detection feature data batch calculation to generate an intermediate table and save it to
  • the second data warehouse 105 can be called directly when the flow computing platform 104 performs real-time fault detection, without additional calculation of the characteristics of historical operation data, which improves the rapid response capability in real-time fault risk identification and early warning scenarios.
  • the system 100 is based on the Spark batch calculation engine of the batch calculation module 106, and performs Spark batch calculation support during the feature extraction and identification process of the massive data by the fault detection module 102, so that the corresponding intermediate parameters can be quickly obtained, which is beneficial in large-scale Improve data computing efficiency in data processing scenarios.
  • the intermediate table generated by the intermediate parameters is transmitted to the second data warehouse 105 for storage for faster and more convenient call by the streaming computing platform 104, reducing the waiting time for real-time computing, reducing the difficulty of data processing for streaming computing, and further improving real-time fault detection Computational efficiency in the process.
  • the stream computing platform 104 may include:
  • the real-time acquisition module is used to obtain the intermediate parameters and battery real-time operation data from the second data warehouse in real time through the stream computing engine;
  • a determination module is configured to determine corresponding battery real-time state data according to the real-time battery operation data
  • Splicing module used for splicing the real-time battery status data and the intermediate parameters in the intermediate table
  • the calculation module is used to calculate the spliced data through the preset second fault detection model, and output the fault detection data corresponding to the battery and the identity information of the battery.
  • the real-time acquisition module of the stream computing platform can continuously and orderly acquire intermediate parameters and real-time battery operating data in real time from the second data warehouse in a stream computing manner through the stream computing engine 106 , and determine the corresponding real-time battery status data according to the real-time operating data of the battery through the determining module.
  • the real-time battery state data can be stored in the form of a table, so when splicing, the splicing module can splice the table of the real-time battery state data and the above-mentioned intermediate table in the second data warehouse 105 to form a new table, based on the new table , the calculation module reads the corresponding data and inputs the preset second fault detection model, which can identify and calculate faults based on the characteristics of historical operating data and real-time operating data, and obtain the fault detection data and identity information of each battery (such as Identity document, id ).
  • the flow computing engine can retain the timing attributes of the battery's historical operating data and real-time operating data, so that the fault detection results are strongly correlated with the timing of the data, which conforms to the objective law of battery operating status changing with time, and improves the accuracy of fault detection results sex.
  • the second fault detection model can be a pre-built matrix model, neural network model, or decision tree model, etc.
  • the calculation module inputs the spliced real-time battery state data and intermediate parameters into the corresponding second fault detection model It is processed in the battery, and the fault detection data and identity information of each battery are output.
  • the historical and real-time battery id, current value, voltage value, fault code, SOC value, etc. after splicing are input into the model, and data such as whether each battery id has a failure risk, the fault code of the fault risk, etc. are output, so that In the big data battery risk identification management scenario, the fault detection data corresponding to each battery can be obtained.
  • the system 100 may also include an early warning module 107 and a service subsystem 108, wherein:
  • the early warning module 107 is used to generate early warning information according to the fault detection data of the corresponding battery and the identity information of the battery and send it to the service subsystem 108; the service subsystem 108 is used to query the target terminal corresponding to the identity information of the battery according to the early warning information , to generate service information corresponding to the target terminal and fault detection data.
  • the service subsystem 108 may be an after-sales service system of the operator.
  • the data output by the flow computing platform 104 indicates that the power battery whose battery id is "123" will have a risk of thermal runaway, and the early warning model 107 can generate early warning information based on these data.
  • the early warning information can include the battery id and the identification of the failure risk, real-time Send to the service subsystem 108.
  • the service subsystem 108 can search for the target terminal corresponding to the battery id, and send the relevant service information to the target terminal for the user to know, so as to intercept the corresponding battery in real time through the terminal and instruct the battery to be Risk intervention, if indicated to replace the battery. In this way, the closed-loop process in the detection and early warning process of big data battery failure risks can be realized, with low false alarm rate and high reliability.
  • the fault detection system provided by the embodiment of the present application is especially suitable for the fault risk identification and early warning of the power battery of the new energy vehicle based on the big data platform, without increasing the hardware cost of the battery system (Battery Management System, BMS), and is easy to deploy.
  • BMS Battery Management System
  • the offline-based batch processing engine can train historical operating data, optimize fault detection feature data parameters, improve the accuracy of fault feature recognition, and sample real-time operating data of each frame based on stream computing for fusion calculation to obtain more reliable fault detection As a result, an early warning guarantee is provided for the safety of the user and the vehicle battery.
  • FIG. 3 is a schematic flowchart of a fault detection method provided by an embodiment of the present application. As shown in Figure 3, the method may include S101-S104:
  • S101 acquires corresponding battery historical operation data from the first data warehouse
  • S103 performs batch calculation on the fault detection feature data through the batch calculation engine to obtain corresponding intermediate parameters
  • S104 obtains the intermediate parameters and the real-time operation data of the battery in real time through the flow calculation engine, and uses the second fault detection model to perform fusion calculation to obtain the fault detection result of the corresponding battery.
  • the first data warehouse can store a large amount of historical operation data, and based on these large amount of historical operation data with a sufficiently long time span, feature extraction can be performed through the first fault detection model to obtain more accurate fault detection features data. Then, batch calculation is performed on a large amount of the fault detection characteristic data through the batch calculation engine to quickly obtain the corresponding intermediate parameters, so as to ensure that the subsequent flow calculation engine can obtain the intermediate parameters in time when performing real-time fusion calculation.
  • the intermediate parameters and real-time operation data of the battery can be obtained in real time through the stream computing engine, and the second fault detection model is used to perform fusion calculation on these data to obtain the fault detection result of the battery.
  • the embodiment of the present application can ensure the real-time performance of fault detection while taking into account the reliability of the detection results, which facilitates timely fault response to the fault risk of the battery, and provides guarantee for the safe operation of the power battery.
  • the first data warehouse can be an offline data warehouse, such as the Hive warehouse, which can store a large amount of historical battery operation data, and provide historical operation data with a long enough time span for the relevant calculation of fault detection, so as to obtain accurate The results of fault detection can improve the reliability of fault early warning.
  • the historical operation data of the battery may include the voltage value, current value, temperature value, state of charge (State Of Charge, SOC) value, and state of health (State Of Health) value of the battery at historical time, but is not limited thereto.
  • the method may further include: collecting a large amount of historical battery operating data, performing cleaning and normalization processing, and storing them in the first data warehouse, so that Called by the first fault detection model.
  • the method may further include:
  • Multi-dimensional feature extraction is performed on the battery historical operation sample data, and the first fault detection model corresponding to the fault characteristics is constructed.
  • the battery historical operation sample data is to take enough historical battery operation data as samples, extract voltage characteristics, temperature characteristics, current characteristics and corresponding fault codes (such as thermal runaway fault codes) from the samples, and construct a multi-dimensional The first fault detection model of .
  • the first fault detection model can be used to obtain historical battery operation data for prediction and identification, and output fault detection characteristic data corresponding to different fault risks, such as corresponding fault codes.
  • the constructed first fault detection model may be a matrix model, a neural network model, or a decision tree model, which is not limited in this embodiment. And it should be understood that constructing matrix models, neural network models, or decision tree models based on sample data are all mature technologies in the field, and will not be repeated here.
  • a large amount of historical operation data within the target duration stored in the first data warehouse can be used as samples (that is, battery historical operation sample data) to perform multi-dimensional feature extraction, and fault features of each dimension can be obtained to construct The corresponding first fault detection model is obtained. Because there are a large amount of historical operating data as training samples, when the constructed data model is applied to feature extraction, it can ensure the effective execution of feature extraction and obtain more accurate extraction results.
  • step 102 performs feature extraction on historical battery operating data through the preset first fault detection model to obtain corresponding fault detection feature data, which may specifically include:
  • the sampling period can be measured by time such as year, month, day, hour, minute, etc.
  • the preset sampling time can also be set by measuring intervals such as day, hour, minute, and second.
  • the battery historical operation data corresponding to the sampling period is input into the first fault detection model according to the time sequence, so that the first fault detection model can perform multi-dimensional characteristics on the battery historical operation data according to the time sequence extract. In this way, the influence of multidimensionality and timing of fault feature sources can be fully integrated, and more accurate fault detection feature data can be obtained, which in turn can be beneficial to obtain fault detection results with higher reliability.
  • step S103 is executed, through batch
  • the calculation engine performs batch calculation on the fault detection feature data, which can improve the system’s ability to process massive data.
  • it can efficiently process the fault detection feature data obtained by extracting the historical operation data of massive batteries.
  • the batch calculation engine performs batch calculation on the fault feature data to obtain the corresponding intermediate parameters, and the method may further include:
  • the intermediate table is input into the preset second data warehouse to update the existing intermediate table stored in the second data warehouse, and the second data warehouse is a real-time data warehouse.
  • the batch computing engine may be a Spark batch computing engine.
  • the Spark batch calculation engine has the advantages of being fast and versatile. It can perform batch calculations on massive fault detection feature data, obtain corresponding intermediate parameters, and generate intermediate tables corresponding to time series.
  • the intermediate parameters may include battery number id, detection time consumption, report time, current value, voltage value, SOC value, SOH value, fault code, battery status, etc., and store this data generation intermediate table in the first In the second data warehouse 105, it is convenient for the flow calculation engine to call when performing real-time fault detection calculations, reducing the complexity of data calculation during flow calculations.
  • the method uses the first fault detection model to regularly obtain sufficient historical operating data of the battery from the offline first data warehouse for fault detection, obtain fault detection characteristic data and perform batch calculation to generate an intermediate table, and save it to the second Two data warehouses, which can be directly called when the stream computing engine performs real-time fault detection, without additional calculation of the characteristics of historical operating data, reducing the waiting time for real-time computing, which is conducive to improving the computing efficiency in the real-time fault detection process and improving real-time faults Rapid response capabilities in risk identification and early warning scenarios.
  • the second data warehouse is a real-time data warehouse, such as an Hbase warehouse. Since the real-time operation data of the battery is continuously updated, in order to enable the flow computing platform to obtain real-time and timely and effective processing of the real-time operation data generated by the battery, the system of this embodiment can store the batch computing engine in real time through the second data warehouse The calculated intermediate parameters and the real-time operating data of the battery obtained from the terminal provide guarantee for the data reading of the stream computing engine during the fusion computing process.
  • the second data warehouse adopts the Hbase warehouse, which can obtain each frame of real-time operation data of the battery, and correspondingly update the real-time status data of the battery for invocation by the flow computing engine, so that the second fault detection model can be based on the historical operation data of the battery
  • the relevant characteristic data and each frame of real-time operation data are fused and calculated to obtain more accurate and reliable real-time fault detection results of the battery, which in turn facilitates the realization of real-time fault risk prediction.
  • the real-time operation data of the battery may include the real-time voltage value, current value, temperature value, state of charge (State Of Charge, SOC) value, and state of health (State Of Health, SOH) value of the battery, but not limited to this.
  • the real-time battery state data stored in the second data warehouse may include a voltage value, a current value, a temperature value, an SOC value, and a state of health SOH value, and the like.
  • the data acquired by the first data warehouse and the second data warehouse may be collected by the stream computing platform.
  • the stream computing platform may be a processing platform including a Flink stream computing engine (hereinafter referred to as "Flink platform"). Therefore, the real-time operating data of the battery of the terminal can be obtained continuously and in real time, and the data are stored in the first data warehouse and the second data warehouse respectively.
  • the new real-time operation data covers the existing data, so as to achieve the continuous update of the data.
  • the new real-time operation data and the existing battery historical operation data are stored in the first data warehouse, so that the data can be continuously accumulated to update the overall data.
  • the first fault detection model obtains historical operation data from the first data warehouse for feature extraction, obtains corresponding fault detection feature data for batch calculation, and generates intermediate parameters of these historical operation data and stores them in In the second data warehouse, it can participate in the fusion calculation together with the real-time operation data acquired by the second data warehouse this time.
  • the real-time operation data acquired this time is stored in the first data warehouse to prepare for the next real-time fault detection.
  • the flow computing engine based on the Flink platform can continuously and orderly obtain the above-mentioned real-time running data and intermediate parameters updated with time from the second data warehouse in the form of flow computing in the process of fusion computing, and perform data fusion calculate.
  • the time series attributes of the historical operation data and real-time operation data of the battery can be preserved, so that the fault detection results are strongly correlated with the time series of the data, which conforms to the objective law of battery operation status changing with time, and improves the accuracy of fault detection results. sex.
  • step S104 acquires the intermediate parameters and real-time battery operating data in real time through the flow computing engine, and uses the second fault detection model to perform fusion calculations on the real-time battery state data and intermediate parameters to obtain the fault of the corresponding battery Test results may include:
  • the spliced data is calculated, and the fault detection data corresponding to the battery and the identity information of the battery are output.
  • the stream computing engine can continuously obtain intermediate parameters and real-time battery operating data from the second data warehouse in real-time through stream computing, and determine the corresponding real-time battery status data according to the real-time battery operating data.
  • the real-time battery status data can be stored in the second data warehouse in the form of a table. Therefore, when splicing, the table of real-time battery status data and the above-mentioned intermediate table in the second data warehouse can be spliced together to form a new data table corresponding to time series.
  • the second fault detection model can continue to read the data in the data table for calculation, and obtain the fault detection data and identity information (such as Identity document, id) of each battery, thereby facilitating the identification and management of battery fault risks in big data
  • the fault detection data corresponding to each battery is obtained in real time to improve the response capability to the risk of battery failure.
  • it can retain the timing attributes of the battery's historical operating data and real-time operating data, so that the fault detection results are strongly correlated with the timing of the data, which conforms to the objective law of battery operating status changing with time, and improves the accuracy of fault detection results sex.
  • the second fault detection model may be a pre-built matrix model, neural network model, or decision tree model, etc., and the construction method is similar to that of the first fault detection model, which will not be repeated here.
  • the spliced real-time battery state data and intermediate parameters are input into the corresponding second fault detection model for training and decision-making, and the fault detection data and identity information of each battery can be output.
  • the historical and real-time battery id, current value, voltage value, fault code, SOC value, etc. after splicing are input into the model, and data such as whether each battery id has a failure risk, the fault code of the fault risk, etc. are output, so that In the big data battery risk identification management scenario, the fault detection data corresponding to each battery can be obtained.
  • the method may further include:
  • the early warning information is sent to the service subsystem, so that the service subsystem queries the target terminal corresponding to the identity information of the battery according to the early warning information, and generates service information corresponding to the target terminal and fault detection data.
  • the service subsystem may be the after-sales service system of the operator.
  • the data output by the second fault detection model indicates that the power battery whose battery id is "123" will have a risk of thermal runaway
  • the generated early warning model information may include the battery id and the identification of the fault risk.
  • the early warning information is sent to the service subsystem in real time, so that after receiving the early warning information, the service subsystem can find the target terminal corresponding to the battery id, and send the relevant service information to the target terminal for the user to know, so that the corresponding battery can be monitored through the terminal.
  • Real-time risk interception indicating risk intervention on the battery, such as indicating that the battery should be replaced. In this way, the closed-loop process in the detection and early warning process of big data battery failure risks can be realized, with low false alarm rate and high reliability.
  • the fault detection method provided by the embodiment of the present application is especially suitable for the fault risk identification and early warning of the power battery of the new energy vehicle based on the big data platform.
  • the cost of failure risk identification and early warning is low.
  • the offline-based batch processing engine can train historical operating data, optimize fault detection feature data parameters, improve the accuracy of fault feature recognition, and sample real-time operating data of each frame based on stream computing for fusion calculation to obtain more reliable fault detection As a result, an early warning guarantee is provided for the safety of the user and the vehicle battery.
  • FIG. 4 shows a schematic structural diagram of the electronic device provided by the embodiment of the present application.
  • the electronic device includes: a processor 401 and a memory 402 storing computer program instructions;
  • processor 401 When the processor 401 executes computer program instructions, it implements the fault detection method in any one of the foregoing embodiments.
  • processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
  • CPU central processing unit
  • ASIC Application Specific Integrated Circuit
  • Memory 402 may include mass storage for data or instructions.
  • memory 402 may include a hard disk drive (Hard Disk Drive, HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (Universal Serial Bus, USB) drive or two or more Combinations of multiple of the above.
  • Storage 402 may include removable or non-removable (or fixed) media, where appropriate. Under appropriate circumstances, the storage 402 can be inside or outside the comprehensive gateway disaster recovery device.
  • memory 402 is a non-volatile solid-state memory.
  • Memory 402 may include read only memory (ROM), random access memory (RAM), magnetic disk storage media devices, optical storage media devices, flash memory devices, electrical, optical, or other physical/tangible memory storage devices.
  • ROM read only memory
  • RAM random access memory
  • magnetic disk storage media devices magnetic disk storage media devices
  • optical storage media devices flash memory devices
  • electrical, optical, or other physical/tangible memory storage devices include one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software comprising computer-executable instructions, and when the software is executed (e.g., by one or multiple processors) operable to perform the operations described with reference to the method according to an aspect of the present application.
  • the electronic device may further include a communication interface 403 and a bus 410 .
  • the processor 401 , the memory 402 , and the communication interface 403 are connected through a bus 410 to complete mutual communication.
  • the communication interface 403 is mainly used to implement communication between modules, devices, units and/or devices in the embodiments of the present application.
  • Bus 410 includes hardware, software, or both, and couples the components of the electronic device to each other.
  • the bus may include Accelerated Graphics Port (AGP) or other graphics bus, Enhanced Industry Standard Architecture (EISA) bus, Front Side Bus (FSB), HyperTransport (HT) interconnect, Industry Standard Architecture (ISA) Bus, Infiniband Interconnect, Low Pin Count (LPC) Bus, Memory Bus, Micro Channel Architecture (MCA) Bus, Peripheral Component Interconnect (PCI) Bus, PCI-Express (PCI-X) Bus, Serial Advanced Technology Attachment (SATA) bus, Video Electronics Standards Association Local (VLB) bus or other suitable bus or a combination of two or more of these.
  • Bus 410 may comprise one or more buses, where appropriate. Although the embodiments of this application describe and illustrate a particular bus, this application contemplates any suitable bus or interconnect.
  • the embodiments of the present application also provide a readable storage medium, on which programs or instructions are stored, and when the programs or instructions are executed by the processor, the implementation as described in the above-mentioned embodiments can be realized.
  • the embodiment of the present application also provides a chip, the chip includes a processor and a communication interface, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, the processor is used to run programs or instructions, and realize the The steps of the battery fault detection method described above.
  • the embodiment of the present application also provides a computer program product.
  • the instructions in the computer program product are executed by the processor of the electronic device, the electronic device executes the steps of the method for detecting a battery fault as described in the above embodiments.
  • the functional modules shown in the above structural block diagrams may be implemented as hardware, software, firmware or a combination thereof.
  • it When implemented in hardware, it may be, for example, an electronic circuit, an application specific integrated circuit (ASIC), suitable firmware, a plug-in, a function card, or the like.
  • ASIC application specific integrated circuit
  • the elements of the present application are the programs or code segments employed to perform the required tasks.
  • Programs or code segments can be stored in machine-readable media, or transmitted over transmission media or communication links by data signals carried in carrier waves.
  • "Machine-readable medium” may include any medium that can store or transmit information.
  • machine-readable media examples include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, and the like.
  • Code segments may be downloaded via a computer network such as the Internet, an Intranet, or the like.

Landscapes

  • Tests Of Electric Status Of Batteries (AREA)

Abstract

本申请实施例提供一种电池故障检测***、方法和设备。其中***包括：第一数据仓库，用于存储电池历史运行数据；故障检测模块，用于从第一数据仓库中获取对应的电池历史运行数据进行特征提取，得到对应的故障检测特征数据；批量计算模块：用于通过批计算引擎对所述故障检测特征数据进行批量计算，得到对应的中间参数；流计算平台，用于通过流计算引擎，实时获取所述中间参数和电池实时运行数据进行融合计算，得到对应电池的故障检测结果。

Description

电池故障检测***、方法和设备 技术领域
本申请涉及电池技术领域，特别是涉及一种电池故障检测***、方法和设备。
背景技术
电池作为新能源汽车的主要动力来源，在长时间的运作后可能会常发生故障问题，影响新能源汽车的安全运行，因此对该电池的故障检测尤为重要。
由于动力电池的故障与时序具有强烈的相关性，目前进行动力电池故障识别时，通常需要计算很长时间跨度的数据才能保证预警结果的可靠性。然而，对于动力电池故障检测来说，检测的实时性同样至关重要，因此如何兼顾可靠性以及实时性是该领域亟待解决的问题。
发明内容
本申请提供一种电池故障检测***、方法和设备，能够兼顾电池故障检测的准确性和实时性。
第一方面，本申请实施例提出了一种电池故障检测***，包括：第一数据仓库，用于存储电池历史运行数据；故障检测模块，用于从第一数据仓库中获取对应的电池历史运行数据进行特征提取，得到对应的故障检测特征数据；批量计算模块：用于通过批计算引擎对所述故障检测特征数据进行批量计算，得到对应的中间参数；流计算平台，用于通过流计算引 擎，实时获取所述中间参数和电池实时运行数据进行融合计算，得到对应电池的故障检测结果。
根据本申请实施例，第一数据仓库可以存储海量历史运行数据，供故障检测模块从中获取足够长时间跨度的历史运行数据进行特征提取，得到更准确可靠的故障检测特征数据。然后由批量计算模块通过批计算引擎对大量该故障检测特征数据进行批量计算，以快速得到对应的中间参数，保障后续流计算平台进行实时融合计算时能及时获取该中间参数。该流计算平台通过流计算引擎可以实时获取中间参数和电池实时运行数据进行融合计算，得到对应电池的实时故障检测结果。由于该故障检测结果是由电池的历史运行数据和实时运行数据通过对应的特征提取和融合计算得到，因此准确率更高，降低了电池风险识别的误报率，提高电池故障检测的可靠性。并且本申请实施例中，通过流计算平台能够基于流计算引擎对大规模流动数据(即电池实时运行数据)进行实时获取和计算处理，且在该计算处理前批量计算模块提供了中间参数的快速计算能力，进一步保证计算处理的实时执行。因此本申请实施例在兼顾检测结果可靠性的同时，还能够保证故障检测的实时性，利于对电池的故障风险及时作出故障响应，为动力电池的安全运行提供保障。
在一些实施例中，***还包括第二数据仓库，第二数据仓库，用于实时存储中间参数和电池实时运行数据，以供流计算平台获取。
根据本申请实施例，由于电池的实时运行数据是不断持续更新的，为了能够使流计算平台能够对电池产生的实时运行数据进行实时获取和及时有效处理，本实施例的***可以通过数据仓库实时存储批量计算模块计算得到的中间参数，以及从终端获取的电池实时运行数据，进而为流计算平台在进行流式融合计算过程中的数据读取提供保障。
在一些实施例中，流计算平台包括：数据接口，用于通过物联网协 议，从各终端采集对应电池的实时运行数据；第一传输单元，用于将实时运行数据传输至第二数据仓库中，以更新得到第二数据仓库中存储的已有电池实时运行数据；第二传输单元，用于将实时运行数据传输至第一数据仓库中存储，以更新电池历史运行数据。
根据本申请实施例，数据平台基于物联网协议，通过数据接口从各终端采集电池的实时运行数据，完成大数据采集以进行批量的电池风险识别，进而利于实现大规模电池风险识别管理。其中，获取的实时运行数据通过第一传输单元传输到第二数据仓库中以不断更新已有的电池实时运行数据，进而可以持续实时为流计算平台提供融合计算的数据基础，提高大规模电池风险识别管理的实时性和可靠性。并且，实时运行数据还通过第二传输单元传输到第一数据仓库存储，更新电池历史运行数据，为故障检测模块进行特征提取累积所需的具有时序性的数据基础，从而利于提高故障检测模块基于海量数据进行计算的鲁棒性，同时提高电池故障检测的准确性。
在一些实施例中，故障检测模块，包括：数据获取单元，用于在预设采样时间，从第一数据仓库中获取对应采样周期的电池历史运行数据，并按照时序输入特征提取单元中；特征提取单元，用于通过预设的第一故障检测模型，对电池历史运行数据进行多维度特征提取，并输出对应的故障检测特征数据。
根据本申请实施例，故障检测模块包括数据获取单元和特征提取单元，其中数据获取单元用于在预设采样时间从第一数据仓库中获取电池历史运行数据，并按照时序输入特征提取单元，使得特征提取单元能够通过预设的第一故障检测模型，按照时序对电池历史运行数据进行多维度特征提取。这样可以充分综合故障特征来源的多维性和时序性的影响，得到更准确的故障检测特征数据，进而能够利于得到可靠性更高的故障检测结 果。
在一些实施例中，批量计算模块包括：第一计算单元，用于通过批计算引擎对故障检测特征数据进行批量计算，得到对应的中间参数；生成单元，用于根据中间参数，生成对应时序的中间表；第三传输单元，用于将中间表传输至第二数据仓库中，以更新第二数据仓库中存储的已有中间表。
根据本申请实施例，批量计算模块包括第一计算单元、生成单元和第三传输单元，其中第一计算单元用于通过批量计算引擎，能够对由海量电池历史运行数据得到大量故障检测特征数据进行批量计算，快速得到对应的中间参数，利于在大数据处理场景中提高数据计算效率。生成单元用于根据中间参数生成对应时序的中间表，继而由第三传输单元将中间表传输到第二数据仓库中存储以供流计算平台调用。中间参数被用于融合计算前通过中间表存储在第二数据仓库中，可以更快更方便的被调用，降低实时计算的等待时长，降低流式计算的数据处理难度，利于提高实时故障检测过程中的计算效率。
在一些实施例中，流计算平还包括：实时获取模块，用于通过流计算引擎，从第二数据仓库实时获取中间参数和电池实时运行数据；确定模块，用于根据电池实时运行数据，确定对应的电池实时状态数据；拼接模块，用于将电池实时运行数据与中间表中的中间参数进行拼接；计算模块，用于通过预设的第二故障检测模型，对拼接后的数据进行计算，并输出对应电池的故障检测数据和电池的身份信息。
根据本申请实施例，流计算平台的实时获取模块通过流计算引擎，能够以流计算方式从第二数据仓库持续实时获取中间参数和电池实时运行数据，通过确定模块根据电池实时运行数据对应的电池实时状态数据。然后拼接模块将电池实时运行数据与中间表中的中间参数进行拼接，形成新 的对应时序的数据表，以供计算模块读取该数据表中的数据，通过预设的第二故障检测模型进行计算，输出对应的故障检测数据和该数据对应的电池的身份信息，以利于在大数据电池风险识别管理场景中，得到对应各电池的故障检测数据。
在一些实施例中，***还包括预警模块和服务子***，预警模块，用于根据对应电池的故障检测数据和电池的身份信息，生成预警信息并发送至服务子***；服务子***，用于根据预警信息，查询电池的身份信息对应的目标终端，以生成对应目标终端和故障检测数据的服务信息。
根据本申请实施例，可以在故障检测数据中表征电池有故障或异常状况时，生成预警信息发送给服务子***，以通过服务子***查询电池身份信息对应的目标终端，将相关的服务信息发给目标终端供用户知晓，实现大数据电池故障风险预警。
第二方面，本申请实施例提供了一种电池故障检测方法，包括：
从第一数据仓库中获取对应的电池历史运行数据；通过预设的第一故障检测模型对电池历史运行数据进行特征提取，得到对应的故障检测特征数据；通过批计算引擎对故障检测特征数据进行批量计算，得到对应的中间参数；通过批计算引擎，对故障检测特征数据进行批量计算，得到对应的中间参数；通过流计算平台实时获取中间参数和电池实时运行数据；利用第二故障检测模型对中间参数和电池实时运行数据进行融合计算，得到对应电池的故障检测结果。
根据本申请实施例，第一数据仓库可以存储海量历史运行数据，基于这些海量的、时间跨度足够长的历史运行数据，通过第一故障检测模型进行特征提取，可以得到更为准确的故障检测特征数据。然后通过批计算引擎对大量该故障检测特征数据进行批量计算，以快速得到对应的中间参数，保障后续流计算引擎进行实时融合计算时能及时获取该中间参数。通 过流计算引擎可以实时获取中间参数和电池实时运行数据，利用第二故障检测模型对这些数据进行融合计算，得到电池的故障检测结果。由于该故障检测结果由电池的历史运行数据和实时运行数据通过对应的特征提取和融合计算得到，因此准确率更高，降低了电池风险识别的误报率，提高电池故障检测的可靠性。并且本申请实施例中，能够基于流计算引擎对大规模流动数据(即电池实时运行数据)进行实时获取和计算处理，且在该计算处理前批计算引擎提供了中间参数的快速计算能力，进一步保证计算处理的实时执行。因此本申请实施例在兼顾检测结果可靠性的同时，还能够保证故障检测的实时性，利于对电池的故障风险及时作出故障响应，为动力电池的安全运行提供保障。
在一些实施例中，在通过预设的第一故障检测模型对电池历史运行数据进行特征提取之前，方法还包括：从第一数据仓库中，获取目标时长内的电池历史运行样本数据；对电池历史运行样本数据进行多维度特征提取，构建得到对应故障特征的第一故障检测模型。
根据本申请实施例，可以以第一数据仓库中存储的目标时长内的大量历史运行数据作为样本(即电池历史运行样本数据)，进行多维度的特征提取，可以得到各维度的故障特征，构建得到对应的第一故障检测模型。因为有大量历史运行数据作为训练样本，构建得到的数据模型应用于特征提取时，可以得到更为准确的提取结果。
在一些实施例中，通过预设的第一故障检测模型对电池历史运行数据进行特征提取，得到对应的故障检测特征数据，包括：在预设采样时间，将对应采样周期的电池历史运行数据按照时序输入第一故障检测模型中；通过第一故障检测模型，对电池历史运行数据进行多维度特征提取，得到对应的故障检测特征数据。
根据本申请实施例，在预设采样时间，将对应采样周期的电池历史 运行数据按照时序输入第一故障检测模型中，使得第一故障检测模型能够按照时序对电池历史运行数据进行多维度特征提取。这样可以充分综合故障特征来源的多维性和时序性的影响，得到更准确的故障检测特征数据，进而能够利于得到可靠性更高的故障检测结果。
在一些实施例中，在通过批计算引擎，对故障特征数据进行批量计算，得到对应的中间参数之后，方法还包括：根据中间参数，生成对应时序的中间表；将中间表输入预设的第二数据仓库中，以更新第二数据仓库中存储的已有中间表，所述第二数据仓库为实时数据仓库。
根据本申请实施例，中间参数被用于融合计算前以中间表的形式存储在第二数据仓库中，可以更快更方便的被调用，降低实时计算的等待时长，降低流式计算的数据处理难度，利于提高实时故障检测过程中的计算效率。
在一些实施例中，通过流计算引擎，实时获取中间参数和电池实时运行数据，并利用第二故障检测模型进行融合计算，得到对应电池的故障检测结果，包括：通过流计算引擎，从第二数据仓库实时获取中间参数和电池实时运行数据；根据电池实时运行数据，确定对应的电池实时状态数据；将电池实时状态数据与中间表中的中间参数进行拼接；通过预设的第二故障检测模型，对拼接后的数据进行计算，并输出对应电池的故障检测数据和电池的身份信息。
根据本申请实施例，通过流计算引擎从第二数据仓库持续地实时获取中间参数和电池实时运行数据，根据电池实时运行数据，不断更新对应的电池实时状态数据，然后将电池实时状态数据与中间表中的中间参数进行拼接，形成新的对应时序的数据表，以基于流计算引擎的支持，持续供第二故障检测模型读取该数据表中的数据进行计算，输出对应电池的故障检测数据和电池的身份信息，从而利于在大数据电池故障风险识别管理场 景中，实时得到对应各电池的故障检测数据，提高对电池故障风险的响应能力。
在一些实施例中，在得到对应电池的故障检测结果之后，方法还包括：根据对应电池的故障检测数据和电池的身份信息，生成预警信息；将预警信息发送至服务子***，以使服务子***根据预警信息，查询电池的身份信息对应的目标终端，并生成对应目标终端和故障检测数据的服务信息。
根据本申请实施例，可以在故障检测数据中表征电池有故障或异常状况时，生成预警信息发送给服务子***，以通过服务子***查询电池身份信息对应的目标终端，将相关的服务信息发给目标终端供用户知晓，实现大数据电池故障风险预警。
第三方面，本申请实施例提供了一种电子设备，包括处理器，存储器及存储在存储器上并可在处理器上运行的程序或指令，程序或指令被处理器执行时实现如第一方面任意实施例所述的电池故障检测方法的步骤。
第四方面，本申请实施例提供了一种可读存储介质，可读存储介质上存储程序或指令，程序或指令被处理器执行时实现如第一方面任意实施例所述的电池故障检测方法的步骤。
第五方面，本申请实施例提供了一种芯片，芯片包括处理器和通信接口，芯片包括处理器和通信接口，通信接口和处理器耦合，处理器用于运行程序或指令，实现如第一方面任意实施例所述的电池故障检测方法的步骤。
第六方面，本申请实施例提供了一种计算机程序产品，计算机程序产品中的指令由电子设备的处理器执行时，使得电子设备执行如第一方面任意实施例所述的电池故障检测方法的步骤。
上述说明仅是本申请技术方案的概述，为了能够更清楚了解本申请 的技术手段，而可依照说明书的内容予以实施，并且为了让本申请的上述和其它目的、特征和优点能够更明显易懂，以下特举本申请的具体实施方式。
附图说明
为了更清楚地说明本申请实施例的技术方案，下面将对本申请实施例中所需要使用的附图作简单地介绍：
图1是本申请一实施例公开的一种电池故障检测***的结构示意图；
图2是本申请一具体实施例中的电池故障检测***进行故障检测的示意图；
图3是本申请一实施例公开的一种电池故障检测方法的流程示意图；
图4是本申请一实施例公开的一种电子设备的硬件结构示意图。
具体实施方式
下面结合附图和实施例对本申请的实施方式作进一步详细描述。以下实施例的详细描述和附图用于示例性地说明本申请的原理，但不能用来限制本申请的范围，即本申请不限于所描述的实施例。
在本申请的描述中，需要说明的是，除非另有说明，“多个”的含义是两个以上；术语“上”、“下”、“左”、“右”、“内”、“外”等指示的方位或位置关系仅是为了便于描述本申请和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本申请的限制。此外，术语“第一”、“第二”、“第三”等仅用于描述目的，而不能理解为指示或暗示相对重要性。“垂 直”并不是严格意义上的垂直，而是在误差允许范围之内。“平行”并不是严格意义上的平行，而是在误差允许范围之内。
在本申请中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本申请所描述的实施例可以与其它实施例相结合。
在动力电池故障检测领域，由于动力电池的故障与时序具有强烈的相关性，目前进行动力电池故障识别时，通常需要计算很长时间跨度的数据才能保证预警结果的可靠性。因此，当前基于大数据故障检测平台做风险电池预警时，通常需要对几十万甚至上百万车辆1～2年的数据进行计算和处理。为了满足大数据存储及处理能力的要求，目前大都采用大数据平台来进行动力电池的故障检测，但即使大数据平台的计算资源非常强大，处理如此大批量的数据也会面临着处理速度过慢，需要几天甚至几周才能处理完如此量级的数据。对此，常规的解决方案一般是缩短数据采集的周期，以减少用于故障检测的数据量，从而提升故障检测的实时性。但这么做是以牺牲故障检测准确率代价，因为动力电池的故障与时序具有强烈的相关性，短时间内的数据往往无法反映电池的真实状态。
为了解决上述相关技术中存在的部分或全部问题，本申请实施例提供了一种电池故障检测***、方法和设备，基于离线仓库存储的海量电池历史运行数据进行特征提取和批量计算，能够快速得到历史运行数据对应故障检测特征的中间参数，由流计算引擎实时获取中间参数和电池的实时运行数据进行融合计算，能够快速得到更为准确的实时故障检测结果，以在大规模电池风险识别管理场景中，兼具检测结果的准确性和实时性。下面首先对本申请实施例提供的电池故障检测***进行阐述。
图1示出的是本申请实施例提供的一种故障检测***的结构示意图。如图1所示，该***100包括：
第一数据仓库101，用于存储电池历史运行数据；
故障检测模块102，用于从第一数据仓库中获取对应的电池历史运行数据进行特征提取，得到对应的故障检测特征数据；
批量计算模块103：用于通过批计算引擎对故障检测特征数据进行批量计算，得到对应的中间参数；
流计算平台104，用于通过流计算引擎，实时获取中间参数和电池实时运行数据进行融合计算，得到对应电池的故障检测结果。
根据本申请实施例，第一数据仓库101可以存储海量历史运行数据，供故障检测模块102从中获取足够长时间跨度的历史运行数据进行特征提取，得到更准确可靠的故障检测特征数据。然后由批量计算模块103通过批计算引擎对大量该故障检测特征数据进行批量计算，以快速得到对应的中间参数，保障后续流计算平台104进行实时融合计算时能及时获取该中间参数。该流计算平台104通过流计算引擎可以实时获取中间参数和电池实时运行数据进行融合计算，得到对应电池的实时故障检测结果。由于该故障检测结果是由电池的历史运行数据和实时运行数据通过对应的特征提取和融合计算得到，因此准确率更高，降低了电池风险识别的误报率，提高电池故障检测的可靠性。并且本申请实施例中，通过流计算平台104能够基于流计算引擎对大规模流动数据(即电池实时运行数据)进行实时获取和计算处理，且在该计算处理前批量计算模块103提供了中间参数的快速计算能力，进一步保证计算处理的实时执行。因此本申请实施例在兼顾检测结果可靠性的同时，还能够保证故障检测的实时性，利于对电池的故障风险及时作出故障响应，为动力电池的安全运行提供保障。
在本申请实施例中，第一数据仓库101可以为离线数据仓库，如 Hive仓库。随着电池使用时间的增长，电池的故障风险也随之升高，基于电池故障风险所具有的强时序相关性，本实施例中在离线数据仓库中存储海量的电池历史运行数据，使得***可以通过对足够长的时间跨度的历史运行数据进行计算，得到准确的故障检测结果，提高故障预警的可靠性。
在一些示例中，对于通过大数据平台对终端电池故障风险监测的场景，第一数据仓库可以存储大数据平台从几十万，甚至上百万辆车的终端(如车载终端)中获取的1～2年内的车辆动力电池运行数据，以提供足够的数据样本进行后续故障特征的提取和检测识别。
示例性的，电池历史运行数据可以包括在历史时间电池的电压值、电流值、温度值、荷电状态(State Of Charge，SOC)值以及健康状态(State OfHealth)值等，但不限于此。
在一些示例中，第一数据仓库101可以对获取的电池海量数据进行清洗和规范化处理后存储，以便于被故障检测模块102调用。
在一些实施例中，参考图2所示，***100还包括第二数据仓库105，第二数据仓库105，用于实时存储中间参数和电池实时运行数据，以供流计算平台104获取。
在本申请实施例中，由于电池的实时运行数据是不断持续更新的，为了能够使流计算平台能够对电池产生的实时运行数据进行实时获取和及时有效处理，本实施例的***可以通过数据仓库实时存储批量计算模块计算得到的中间参数，以及从终端获取的电池实时运行数据，进而为流计算平台在进行流式融合计算过程中的数据读取提供保障。
示例性的，第二数据仓库105可以为实时数据仓库，如Hbase仓库。第二数据仓库105可以实时获取并存储电池的每一帧实时运行数据，对应更新该电池的实时运行数据，以供流计算平台104调用。这样流计算平台104可以通过流计算引擎，持续地将基于电池历史运行数据相关的特 征数据和每一帧实时运行数据进行融合计算，得到更为准确可靠的电池的实时故障检测结果，进而利于实现实时的故障风险预测。
示例性的，电池实时运行数据可以包括电池实时的电压值、电流值、温度值、荷电状态(State Of Charge，SOC)值以及健康状态(State Of Health，SOH)值等，但不限于此。对应的，第二数据仓库105中存储的电池实时状态数据可以包括电压值、电流值、温度值、SOC值以及健康状态SOH值等。
为了保障数据处理的实时性，可选地，在本申请实施例中，流计算平台104可以包括：
数据接口，用于通过物联网协议，从各终端采集对应电池的实时运行数据；
第一传输单元，用于将实时运行数据传输至第二数据仓库中，以更新得到第二数据仓库中存储的已有电池实时运行数据；
第二传输单元，用于将实时运行数据传输至第一数据仓库中存储，以更新电池历史运行数据。
在一些示例中，参考图2所示，流计算平台104可以为包括Flink流计算引擎106的处理平台(下文简称为“Flink平台”)。Flink平台提供开源流处理框架和分布式流计算引擎106，能够以数据并行和流水线方式执行流数据程序，从而能够持续、实时地获取终端的电池实时运行数据，并将该数据分别存储到第一数据仓库101和第二数据仓库105中。其中，在第二数据仓库105中，新的实时运行数据覆盖已有的数据，达到数据的持续更新。在第一数据仓库101中，新的实时运行数据与已有的电池历史运行数据均存储在第一数据仓库101中，使数据持续累积，达到整体数据的更新。
具体地，在一次实时故障检测过程中，故障检测模块102从第一数 据仓库101获取历史运行数据进行特征提取，得到对应的故障检测特征数据进行批计算，生成这些历史运行数据的中间参数存储到第二数据仓库105中，以与本次第二数据仓库105获取的实时运行数据一起参与融合计算。同时，本次获取的实时运行数据存储到第一数据仓库101中准备参与下一次实时故障检测。
并且，基于Flink平台的流计算引擎，在融合计算过程中，能够以流计算方式，从第二数据仓库中持续、有序地获取上述随时间更新的实时运行数据和中间参数，进行数据的融合计算。这样融合计算得到的结果中，可以保留电池历史运行数据和实时运行数据的时序属性，使得故障检测结果与数据的时序强相关，符合电池运行状态随时间变化的客观规律，提高故障检测结果的准确性。
本示例中，Flink平台具有支持高吞吐、低延迟、高性能的优势，能够在大规模电池故障检测场景中保证数据处理的实时性。Flink平台还能够按照连续事件高效地处理数据，基于轻量级分布式快照(Snapshot)实现的较佳的容错能力，可以保障Flink平台的可靠性和持久性。
在其他示例中，流计算平台也可以为采用其他类型流计算引擎的实时平台，以实现本申请上述流计算平台的功能。
Flink平台基于数据接口的物联网协议与终端建立通信连接。可以理解的是，物联网协议可以为消息队列遥测传输(Message Queue Telemetry Transport，MQTT)协议、约束应用程序(Constrained Application Protocol，CoAP)协议、轻量级物联网(Lightweight Machine-To-Machine，LwM2M)协议、超文本传输(HyperText Transfer Protocol，HTTP)协议、窄带物联网(Narrow Band Internet of Things，NB-IoT)协议等中的一种或多种，本申请实施例不做具体限定。
终端实时监测电池的实时运行数据，并将监测得到的电池实时运行 数据实时上传到Flink平台。Flink平台接收终端上传的数据后，一方面通过第二传输单元将这些数据传输至第一数据仓库101存储，以更新第一数据仓库101中存储电池历史运行数据。另一方面Flink平台将从终端接收的数据通过第一传输单元传输至第二数据仓库105中存储，以对第二数据仓库105中存储的已有电池实时状态数据进行更新。
根据本申请实施例，数据平台105基于物联网协议，通过数据接口从各终端采集电池的实时运行数据，完成大数据采集以进行批量的电池风险识别，进而利于实现大规模电池风险识别管理。其中，获取的实时运行数据通过第一传输单元传输到第二数据仓库105中以不断更新已有的电池实时运行数据，进而可以持续实时为流计算平台104提供融合计算的数据基础，提高大规模电池风险识别管理的实时性和可靠性。并且，实时运行数据还通过第二传输单元传输到第一数据仓库101存储，更新电池历史运行数据，为故障检测模块102进行特征提取累积所需的具有时效性的数据基础，利于提高故障检测模块102基于海量数据进行计算的鲁棒性，同时利于促进提高电池故障检测的准确性。
在一些实施例中，故障检测模块102可以包括：
数据获取单元，用于在预设采样时间，从第一数据仓库101中获取对应采样周期的电池历史运行数据，并按照时序输入特征提取单元中；
特征提取单元，用于通过预设的第一故障检测模型，对电池历史运行数据进行多维度特征提取，并输出对应的故障检测特征数据。
本实施例中，采样周期可以为年、月、日、小时、分钟等时间量度，预设采样时间也可以为以日、小时、分钟、秒等量度间隔设置。故障检测模块102通过数据获取单元，在预设采样时间从第一数据仓库101中获取对应采样周期时长内的电池历史运行数据，并根据获取的电池历史运行数据的时间，按照时序输入特征提取单元中，通过预设的第一故障检测 模型进行多维度特征提取，得到故障检测特性数据。
示例性的，特征提取单元对电池历史运行数据进行特征提取前，可以先从第一数据仓库101中获取足够多的电池历史运行数据样本，从样本中提取电压特性、温度特征、电流特性以及对应的故障码(如热失控故障码)等等，构建多维的第一故障检测模型。第一故障检测模型可以用于获取电池历史运行数据进行预测识别，输出对应的故障码等故障检测特征数据。在一些具体示例中，构建的第一故障检测模型可以为矩阵模型、神经网络模型或决策树模型等，本实施例不做唯一限定。
基于第一故障检测模型，特征提取单元获取采样周期时长内的电池历史运行数据进行多维度的特征提取和识别，输出故障检测特性数据。这样可以充分综合故障特征来源的多维性和时序性的影响，得到更准确的故障检测特征数据，进而能够利于得到可靠性更高的故障检测结果。
在一些实施例中，批量计算模块103具体可以包括：
第一计算单元，用于通过批计算引擎对故障检测特征数据进行批量计算，得到对应的中间参数；
生成单元，用于根据中间参数，生成对应时序的中间表；
第三传输单元，用于将中间表传输至第二数据仓库中，以更新第二数据仓库中存储的已有中间表。
示例性的，批计算引擎可以为Spark批计算引擎。Spark批计算引擎具有快速、通用的优势，能够对海量的故障检测特征数据进行批量计算，得到对应的中间参数，并生成对应时序的中间表。一个具体示例中，中间参数可以包括电池编号id、检测耗时、报告时间、电流值、电压值、SOC值、SOH值、故障码、电池状态等等，将这种数据生成中间表存储到第二数据仓库105中，以便于流计算平台104进行实时故障检测时调用，降低流计算时的数据计算复杂度。
本申请实施例中，大数据平台进行电池故障风险预警场景下，***100通常需要对几十万甚至上百万车辆1～2年的历史运行数据进行处理，处理的数据量级达到Pb级别。如果采用相关技术中的大数据平台来处理如此巨大的数据量，处理速度会非常慢，一般需要几天甚至几周的时间才能处理完，而如果处理完后还要进行触发预警措施，还需要1天以上的反应时间，这样若实际应用中触发预警后的2天内电池就发生了失效，相关技术是无法满足故障风险的提前预警需求的。但本申请实施例中，***100通过提前通过故障检测模块102，定期从离线的第一数据仓库101中获取电池足够的历史运行数据进行故障检测，得到故障检测特征数据批计算生成中间表保存到第二数据仓库105，这样在流计算平台104进行实时故障检测时直接调用即可，不必额外计算历史运行数据的特征，提高了实时故障风险识别和预警场景中的快速响应能力。
根据本申请实施例，***100基于批量计算模块106的Spark批计算引擎，在故障检测模块102进行海量数据的特征提取识别过程中进行Spark批计算支持，能够快速得到对应的中间参数，利于在大数据处理场景中提高数据计算效率。中间参数生成的中间表传输到第二数据仓库105中存储以供流计算平台104更快更方便的调用，降低实时计算的等待时长，降低流式计算的数据处理难度，利于进一步提高实时故障检测过程中的计算效率。
在一些实施例中，流计算平台104可以包括：
实时获取模块，用于通过流计算引擎，从第二数据仓库实时获取中间参数和电池实时运行数据；
确定模块，用于根据电池实时运行数据，确定对应的电池实时状态数据；
拼接模块，用于将电池实时状态数据与中间表中的中间参数进行拼 接；
计算模块，用于通过预设的第二故障检测模型，对拼接后的数据进行计算，并输出对应电池的故障检测数据和电池的身份信息。
本申请实施例中，参考图2所示，流计算平台的实时获取模块通过流计算引擎106，能够以流计算方式，从第二数据仓库持续、有序地实时获取中间参数和电池实时运行数据，并通过确定模块根据电池实时运行数据确定对应的电池实时状态数据。电池实时状态数据可以以表的形式存储，因此拼接模块在拼接时，可以将电池实时状态数据的表和第二数据仓库105中的上述中间表拼接起来，形成新的表，基于该新的表，计算模块读取对应数据并输入预设的第二故障检测模型，可以综合历史运行数据特征和实时运行数据的故障进行识别计算，得到各电池的故障检测数据和身份信息(如Identity document，id)。并且基于流计算引擎的支持，可以保留电池历史运行数据和实时运行数据的时序属性，使得故障检测结果与数据的时序强相关，符合电池运行状态随时间变化的客观规律，提高故障检测结果的准确性。
在一些具体示例中，第二故障检测模型可以为预构建的矩阵模型、神经网络模型或决策树模型等，计算模块将拼接后的电池实时状态数据与中间参数，输入到对应第二故障检测模型中进行处理，输出各电池的故障检测数据和身份信息。例如拼接后的历史和实时的电池id、电流值、电压值、故障码、SOC值等等输入到模型中，输出各电池id是否发生故障风险、发生的故障风险的故障码等等数据，从而可以在大数据电池风险识别管理场景中，得到对应各电池的故障检测数据。
为了实现提前预警，可选地，在一些实施例中，***100还可以包括预警模块107和服务子***108，其中：
预警模块107，用于根据对应电池的故障检测数据和电池的身份信 息，生成预警信息并发送至服务子***108；服务子***108，用于根据预警信息，查询电池的身份信息对应的目标终端，以生成对应目标终端和故障检测数据的服务信息。
示例性的，服务子***108可以为运行方的售后服务***。例如，流计算平台104输出的数据表征电池id为“123”的动力电池将发生热失控风险，则预警模型107根据这些数据可以生成预警信息，预警信息可以包括电池id和故障风险的标识，实时发送给服务子***108。
服务子***108接收到预警信息后，可以查找该电池id对应的目标终端，将相关的服务信息发给目标终端供用户知晓，以通过该终端对对应电池进行实时风险拦截，指示对该电池进行风险干预，如指示要对该电池进行更换。这样可以实现大数据电池故障风险的检测和预警过程中的闭环流程，误报率低，可靠性高。
本申请实施例提供的故障检测***，尤其适用于基于大数据平台对新能源车辆的动力电池的故障风险识别和预警，无需增加电池***(Battery Management System，BMS)的硬件成本，部署方便。并且基于离线的批处理引擎可以训练历史运行数据，优化故障检测特征数据参数，提高故障特征识别的准确率，并基于流计算的方式采样每帧实时运行数据进行融合计算，得到更可靠的故障检测结果，为用户和车辆电池的安全提供预警保障。
本申请实施例还提供了一种故障检测方法。图3示出的为本申请实施例提供的故障检测方法的流程示意图。如图3所示，该方法可以包括S101～S104：
S101从第一数据仓库中获取对应的电池历史运行数据；
S102通过预设的第一故障检测模型对电池历史运行数据进行特征提取，得到对应的故障检测特征数据；
S103通过批计算引擎，对故障检测特征数据进行批量计算，得到对应的中间参数；
S104通过流计算引擎，实时获取中间参数和电池实时运行数据，并利用第二故障检测模型进行融合计算，得到对应电池的故障检测结果。
根据本申请实施例，第一数据仓库可以存储海量历史运行数据，基于这些海量的、时间跨度足够长的历史运行数据，通过第一故障检测模型进行特征提取，可以得到更为准确的故障检测特征数据。然后通过批计算引擎对大量该故障检测特征数据进行批量计算，以快速得到对应的中间参数，保障后续流计算引擎进行实时融合计算时能及时获取该中间参数。通过流计算引擎可以实时获取中间参数和电池实时运行数据，利用第二故障检测模型对这些数据，进行融合计算，得到电池的故障检测结果。由于该故障检测结果由电池的历史运行数据和实时运行数据通过对应的特征提取和融合计算得到，因此准确率更高，降低了电池风险识别的误报率，提高电池故障检测的可靠性。并且本申请实施例中，能够基于流计算引擎对大规模流动数据(即电池实时运行数据)进行实时获取和计算处理，且在该计算处理前批计算引擎提供了中间参数的快速计算能力，进一步保证计算处理的实时执行。因此本申请实施例在兼顾检测结果可靠性的同时，还能够保证故障检测的实时性，利于对电池的故障风险及时作出故障响应，为动力电池的安全运行提供保障。
在一些实施例中，第一数据仓库可以为离线数据仓库，如Hive仓库，能够存储海量的电池历史运行数据，提供足够长的时间跨度的历史运行数据用于故障检测的相关计算，得到准确的故障检测结果，提高故障预警的可靠性。其中，电池历史运行数据可以包括在历史时间电池的电压值、电流值、温度值、荷电状态(State Of Charge，SOC)值以及健康状态(State Of Health)值等，但不限于此。
在一些示例中，在S101从第一数据仓库中获取对应的电池历史运行数据之前，方法还可以包括：采集电池海量历史运行数据并进行清洗和规范化处理，存储到第一数据仓库中，以便于被第一故障检测模型调用。
在一些实施例中，为保证特征提取的有效执行，在S102通过预设的第一故障检测模型对电池历史运行数据进行特征提取之前，方法还可以包括：
从第一数据仓库中，获取目标时长内的电池历史运行样本数据；
对电池历史运行样本数据进行多维度特征提取，构建得到对应故障特征的第一故障检测模型。
示例性的，电池历史运行样本数据即以足够多的电池历史运行数据为样本，从样本中提取电压特性、温度特征、电流特性以及对应的故障码(如热失控故障码)等等，构建多维的第一故障检测模型。第一故障检测模型可以用于获取电池历史运行数据进行预测识别，输出对应的故障码等对应不同故障风险的故障检测特征数据。在一些具体示例中，构建的第一故障检测模型可以为矩阵模型、神经网络模型或决策树模型等，本实施例不做唯一限定。并且应理解，基于样本数据构建矩阵模型、神经网络模型或决策树模型等均为本领域成熟技术，此处不再赘述。
根据本申请实施例，可以以第一数据仓库中存储的目标时长内的大量历史运行数据作为样本(即电池历史运行样本数据)，进行多维度的特征提取，可以得到各维度的故障特征，构建得到对应的第一故障检测模型。因为有大量历史运行数据作为训练样本，构建得到的数据模型应用于特征提取时，可以保证特征提取的有效执行，得到更为准确的提取结果。
示例性的，步骤102通过预设的第一故障检测模型对电池历史运行数据进行特征提取，得到对应的故障检测特征数据，具体可以包括：
在预设采样时间，将对应采样周期的电池历史运行数据按照时序 输入第一故障检测模型中；
通过第一故障检测模型，对电池历史运行数据进行多维度特征提取，得到对应的故障检测特征数据。
本实施例中，采样周期可以为年、月、日、小时、分钟等时间量度，预设采样时间也可以为以日、小时、分钟、秒等量度间隔设置。在本申请实施例中，在预设采样时间，将对应采样周期的电池历史运行数据按照时序输入第一故障检测模型中，使得第一故障检测模型能够按照时序对电池历史运行数据进行多维度特征提取。这样可以充分综合故障特征来源的多维性和时序性的影响，得到更准确的故障检测特征数据，进而能够利于得到可靠性更高的故障检测结果。
由于在大规模电池故障风险检测场景中，需要获取海量电池历史运行数据进行处理，为了提高数据处理效率，从而提高对故障风险的响应能力，本申请一些实施例中，执行步骤S 103，通过批计算引擎，对故障检测特征数据进行批量计算，可以提高***对海量数据的处理能力，这样在海量电池故障风险识别场景中，能够对海量电池历史运行数据提取得到的故障检测特征数据进行高效处理，快速得到关于故障检测特征的中间参数供流计算引擎调用。为了保证流计算引擎进行流计算的有效执行，示例性地，在步骤103通过批计算引擎，对故障特征数据进行批量计算，得到对应的中间参数之后，方法还可以包括：
根据中间参数，生成对应时序的中间表；
将中间表输入预设的第二数据仓库中，以更新第二数据仓库中存储的已有中间表，第二数据仓库为实时数据仓库。
示例性的，批计算引擎可以为Spark批计算引擎。Spark批计算引擎具有快速、通用的优势，能够对海量的故障检测特征数据进行批量计算，得到对应的中间参数，并生成对应时序的中间表。一个具体示例中， 中间参数可以包括电池编号id、检测耗时、报告时间、电流值、电压值、SOC值、SOH值、故障码、电池状态等等，将这种数据生成中间表存储到第二数据仓库105中，以便于流计算引擎进行实时故障检测计算时调用，降低流计算时的数据计算复杂度。
本申请实施例中，方法通过第一故障检测模型，定期从离线的第一数据仓库中获取电池足够的历史运行数据进行故障检测，得到故障检测特征数据并批计算生成中间表，并保存到第二数据仓库，这样在流计算引擎进行实时故障检测时直接调用即可，不必额外计算历史运行数据的特征，降低实时计算的等待时长，利于提高实时故障检测过程中的计算效率，提高了实时故障风险识别和预警场景中的快速响应能力。
示例性的，第二数据仓库为实时数据仓库，如Hbase仓库。由于电池的实时运行数据是不断持续更新的，为了能够使流计算平台能够对电池产生的实时运行数据进行实时获取和及时有效处理，本实施例的***可以通过第二数据仓库实时存储批计算引擎计算得到的中间参数，以及从终端获取的电池实时运行数据，进而为流计算引擎在进行融合计算过程中的数据读取提供保障。
因此，第二数据仓库采用Hbase仓库，可以获取电池的每一帧实时运行数据，对应更新该电池的实时状态数据，以供流计算引擎调用，这样第二故障检测模型可以将基于电池历史运行数据相关特征数据和每一帧实时运行数据进行融合计算，得到更为准确可靠的电池的实时故障检测结果，进而利于实现实时的故障风险预测。其中，示例性的，电池实时运行数据可以包括电池实时的电压值、电流值、温度值、荷电状态(State Of Charge，SOC)值以及健康状态(State Of Health，SOH)值等，但不限于此。对应的，第二数据仓库中存储的电池实时状态数据可以包括电压值、电流值、温度值、SOC值以及健康状态SOH值等。
示例性的，第一数据仓库和第二数据仓库获取的数据可以由流计算平台进行采集。流计算平台可以为包括Flink流计算引擎的处理平台(下文简称为“Flink平台”)。从而能够持续、实时地获取终端的电池实时运行数据，并将该数据分别存储到第一数据仓库和第二数据仓库中。其中，在第二数据仓库中，新的实时运行数据覆盖已有的数据，达到数据的持续更新。在第一数据仓库中，新的实时运行数据与已有的电池历史运行数据均存储在第一数据仓库中，使数据持续累积，达到整体数据的更新。
具体地，在一次实时故障检测过程中，第一故障检测模型从第一数据仓库获取历史运行数据进行特征提取，得到对应的故障检测特征数据进行批计算，生成这些历史运行数据的中间参数存储到第二数据仓库中，以与本次第二数据仓库获取的实时运行数据一起参与融合计算。同时，本次获取的实时运行数据存储到第一数据仓库中准备参与下一次实时故障检测。
并且，基于Flink平台的流计算引擎，在融合计算过程中，能够以流计算方式，从第二数据仓库中持续、有序地获取上述随时间更新的实时运行数据和中间参数，进行数据的融合计算。这样融合计算得到的结果中，可以保留电池历史运行数据和实时运行数据的时序属性，使得故障检测结果与数据的时序强相关，符合电池运行状态随时间变化的客观规律，提高故障检测结果的准确性。
在一些实施例中，步骤S104通过流计算引擎，实时获取所述中间参数和电池实时运行数据，并利用第二故障检测模型，对电池实时状态数据和中间参数进行融合计算，得到对应电池的故障检测结果，具体可以包括：
通过流计算引擎，从第二数据仓库实时获取中间参数和电池实时运行数据；
根据电池实时运行数据，确定对应的电池实时状态数据；
将电池实时状态数据与中间表中的中间参数进行拼接；
通过预设的第二故障检测模型，对拼接后的数据进行计算，并输出对应电池的故障检测数据和电池的身份信息。
本申请实施例中，通过流计算引擎，能够以流计算方式从第二数据仓库持续实时获取中间参数和电池实时运行数据，并根据电池实时运行数据确定对应的电池实时状态数据。电池实时状态数据可以以表的形式存储在第二数据仓库中。因此在拼接时，可以将电池实时状态数据的表和第二数据仓库中的上述中间表拼接起来，形成新的对应时序的数据表，基于该新的对应时序的数据表，在流计算引擎的支持下，可以持续供第二故障检测模型读取该数据表中的数据进行计算，得到各电池的故障检测数据和身份信息(如Identity document，id)，从而利于在大数据电池故障风险识别管理场景中，实时得到对应各电池的故障检测数据，提高对电池故障风险的响应能力。并且基于流计算引擎的支持，可以保留电池历史运行数据和实时运行数据的时序属性，使得故障检测结果与数据的时序强相关，符合电池运行状态随时间变化的客观规律，提高故障检测结果的准确性。
在一些具体示例中，第二故障检测模型可以为预构建的矩阵模型、神经网络模型或决策树模型等，构建方法与第一故障检测模型类似，此处不再赘述。
本实施例中，拼接后的电池实时状态数据与中间参数输入到对应第二故障检测模型中，进行训练决策，可以输出各电池的故障检测数据和身份信息。例如拼接后的历史和实时的电池id、电流值、电压值、故障码、SOC值等等输入到模型中，输出各电池id是否发生故障风险、发生的故障风险的故障码等等数据，从而可以在大数据电池风险识别管理场景中，得到对应各电池的故障检测数据。
为了实现提前预警，可选地，在一些实施例中，在得到对应电池的故障检测结果之后，方法还可以包括：
根据对应电池的故障检测数据和电池的身份信息，生成预警信息；
将预警信息发送至服务子***，以使服务子***根据预警信息，查询电池的身份信息对应的目标终端，并生成对应目标终端和故障检测数据的服务信息。
示例性的，服务子***可以为运行方的售后服务***。例如，第二故障检测模型输出的数据表征电池id为“123”的动力电池将发生热失控风险，则生成的预警模型信息可以包括电池id和故障风险的标识等。预警信息实时发送至服务子***，这样服务子***接收到预警信息后，可以查找该电池id对应的目标终端，将相关的服务信息发给目标终端供用户知晓，以通过该终端对对应电池进行实时风险拦截，指示对该电池进行风险干预，如指示要对该电池进行更换。这样可以实现大数据电池故障风险的检测和预警过程中的闭环流程，误报率低，可靠性高。
本申请实施例提供的故障检测方法，尤其适用于基于大数据平台对新能源车辆的动力电池的故障风险识别和预警，无需对电池***(Battery Management System，BMS)的硬件进行过多改造，实现故障风险识别和预警的成本较低。并且基于离线的批处理引擎可以训练历史运行数据，优化故障检测特征数据参数，提高故障特征识别的准确率，并基于流计算的方式采样每帧实时运行数据进行融合计算，得到更可靠的故障检测结果，为用户和车辆电池的安全提供预警保障。
本申请还提出了一种电子设备，如图4示出了本申请实施例提出了一种电子设备的结构示意图。该电子设备包括：处理器401以及存储有计算机程序指令的存储器402；
处理器401执行计算机程序指令时实现如上述任意一实施例中的故障检测方法。
具体地，上述处理器401可以包括中央处理器(CPU)，或者特定集成电路(Application Specific Integrated Circuit，ASIC)，或者可以被配置成实施本申请实施例的一个或多个集成电路。
存储器402可以包括用于数据或指令的大容量存储器。举例来说而非限制，存储器402可包括硬盘驱动器(Hard Disk Drive，HDD)、软盘驱动器、闪存、光盘、磁光盘、磁带或通用串行总线(Universal Serial Bus，USB)驱动器或者两个或更多个以上这些的组合。在合适的情况下，存储器402可包括可移除或不可移除(或固定)的介质。在合适的情况下，存储器402可在综合网关容灾设备的内部或外部。在特定实施例中，存储器402是非易失性固态存储器。
存储器402可包括只读存储器(ROM)，随机存取存储器(RAM)，磁盘存储介质设备，光存储介质设备，闪存设备，电气、光学或其他物理/有形的存储器存储设备。因此，通常，存储器包括一个或多个编码有包括计算机可执行指令的软件的有形(非暂态)计算机可读存储介质(例如，存储器设备)，并且当该软件被执行(例如，由一个或多个处理器)时，其可操作来执行参考根据本申请的一方面的方法所描述的操作。
在一个示例中，电子设备还可包括通信接口403和总线410。其中，如图4所示，处理器401、存储器402、通信接口403通过总线410连接并完成相互间的通信。
通信接口403，主要用于实现本申请实施例中各模块、装置、单元和/或设备之间的通信。
总线410包括硬件、软件或两者，将电子设备的部件彼此耦接在一 起。举例来说而非限制，总线可包括加速图形端口(AGP)或其他图形总线、增强工业标准架构(EISA)总线、前端总线(FSB)、超传输(HT)互连、工业标准架构(ISA)总线、无限带宽互连、低引脚数(LPC)总线、存储器总线、微信道架构(MCA)总线、***组件互连(PCI)总线、PCI-Express(PCI-X)总线、串行高级技术附件(SATA)总线、视频电子标准协会局部(VLB)总线或其他合适的总线或者两个或更多个以上这些的组合。在合适的情况下，总线410可包括一个或多个总线。尽管本申请实施例描述和示出了特定的总线，但本申请考虑任何合适的总线或互连。
另外，结合上述实施例中的故障检测方法，本申请实施例还提供了一种可读存储介质，可读存储介质上存储程序或指令，程序或指令被处理器执行时实现如上述实施例所述的电池故障检测方法的步骤。
并且，本申请实施例还提供了一种芯片，芯片包括处理器和通信接口，芯片包括处理器和通信接口，通信接口和处理器耦合，处理器用于运行程序或指令，实现如上述实施例所述的电池故障检测方法的步骤。
并且，本申请实施例还提供了一种计算机程序产品，计算机程序产品中的指令由电子设备的处理器执行时，使得电子设备执行如上述实施例所述的电池故障检测方法的步骤。
需要明确的是，本申请并不局限于上文所描述并在图中示出的特定配置和处理。为了简明起见，这里省略了对已知方法的详细描述。在上述实施例中，描述和示出了若干具体的步骤作为示例。但是，本申请的方法过程并不限于所描述和示出的具体步骤，本领域的技术人员可以在领会本申请的精神后，作出各种改变、修改和添加，或者改变步骤之间的顺序。
以上所述的结构框图中所示的功能模块可以实现为硬件、软件、 固件或者它们的组合。当以硬件方式实现时，其可以例如是电子电路、专用集成电路(ASIC)、适当的固件、插件、功能卡等等。当以软件方式实现时，本申请的元素是被用于执行所需任务的程序或者代码段。程序或者代码段可以存储在机器可读介质中，或者通过载波中携带的数据信号在传输介质或者通信链路上传送。“机器可读介质”可以包括能够存储或传输信息的任何介质。机器可读介质的例子包括电子电路、半导体存储器设备、ROM、闪存、可擦除ROM(EROM)、软盘、CD-ROM、光盘、硬盘、光纤介质、射频(RF)链路，等等。代码段可以经由诸如因特网、内联网等的计算机网络被下载。
还需要说明的是，本申请中提及的示例性实施例，基于一系列的步骤或者装置描述一些方法或***。但是，本申请不局限于上述步骤的顺序，也就是说，可以按照实施例中提及的顺序执行步骤，也可以不同于实施例中的顺序，或者若干步骤同时执行。
虽然已经参考优选实施例对本申请进行了描述，但在不脱离本申请的范围的情况下，可以对其进行各种改进并且可以用等效物替换其中的部件。尤其是，只要不存在结构冲突，各个实施例中所提到的各项技术特征均可以任意方式组合起来。本申请并不局限于文中公开的特定实施例，而是包括落入权利要求的范围内的所有技术方案。

Claims (17)

  1. 一种电池故障检测***，包括：
    第一数据仓库，用于存储电池历史运行数据；
    故障检测模块，用于从所述第一数据仓库中获取对应的电池历史运行数据进行特征提取，得到对应的故障检测特征数据；
    批量计算模块：用于通过批计算引擎对所述故障检测特征数据进行批量计算，得到对应的中间参数；
    流计算平台，用于通过流计算引擎，实时获取所述中间参数和电池实时运行数据进行融合计算，得到对应电池的故障检测结果。
  2. 根据权利要求1所述的***，其中，所述***还包括第二数据仓库，
    所述第二数据仓库，用于实时存储所述中间参数和所述电池实时运行数据，以供所述流计算平台获取。
  3. 根据权利要求2所述的***，其中，所述流计算平台包括：
    数据接口，用于通过物联网协议，从各终端采集对应电池的实时运行数据；
    第一传输单元，用于将所述实时运行数据传输至所述第二数据仓库中，以更新得到所述第二数据仓库中存储的已有电池实时运行数据；
    第二传输单元，用于将所述实时运行数据传输至所述第一数据仓库中存储，以更新所述电池历史运行数据。
  4. 根据权利要求2所述的***，其中，所述故障检测模块，包括：
    数据获取单元，用于在预设采样时间，从所述第一数据仓库中获取对 应采用周期的电池历史运行数据，并按照时序输入特征提取单元中；
    所述特征提取单元，用于通过预设的第一故障检测模型，对所述电池历史运行数据进行多维度特征提取，并输出对应的故障检测特征数据。
  5. 根据权利要求4所述的***，其中，所述批量计算模块包括：
    第一计算单元，用于通过批计算引擎对所述故障检测特征数据进行批量计算，得到对应的中间参数；
    生成单元，用于根据所述中间参数，生成对应时序的中间表；
    第三传输单元，用于将所述中间表传输至所述第二数据仓库中，以更新所述第二数据仓库中存储的已有中间表。
  6. 根据权利要求5所述的***，其中，所述流计算平台，还包括：
    实时获取模块，用于通过流计算引擎，从所述第二数据仓库实时获取所述中间参数和电池实时运行数据；
    确定模块，用于根据所述电池实时运行数据，确定对应的电池实时状态数据；
    拼接模块，用于将所述电池实时运行数据与所述中间表中的中间参数进行拼接；
    计算模块，用于通过预设的第二故障检测模型，对拼接后的数据进行计算，并输出对应电池的故障检测数据和所述电池的身份信息。
  7. 根据权利要求6所述的***，其中，所述***还包括预警模块和服务子***，
    所述预警模块，用于根据所述对应电池的故障检测数据和所述电池的身份信息，生成预警信息并发送至所述服务子***；
    所述服务子***，用于根据所述预警信息，查询所述电池的身份信息对应的目标终端，以生成对应所述目标终端和所述故障检测数据的服务信息。
  8. 一种电池故障检测方法，包括：
    从第一数据仓库中获取对应的电池历史运行数据；
    通过预设的第一故障检测模型对所述电池历史运行数据进行特征提取，得到对应的故障检测特征数据；
    通过批计算引擎，对所述故障检测特征数据进行批量计算，得到对应的中间参数；
    通过流计算引擎，实时获取所述中间参数和电池实时运行数据，并利用第二故障检测模型进行融合计算，得到对应电池的故障检测结果。
  9. 根据权利要求8所述的方法，其中，
    在所述通过预设的第一故障检测模型对所述电池历史运行数据进行特征提取之前，所述方法还包括：
    从所述第一数据仓库中，获取目标时长内的电池历史运行样本数据；
    对所述电池历史运行样本数据进行多维度特征提取，构建得到对应故障特征的所述第一故障检测模型。
  10. 根据权利要求8所述的方法，其中，
    所述通过预设的第一故障检测模型对所述电池历史运行数据进行特征提取，得到对应的故障检测特征数据，包括：
    在预设采样时间，将对应采样周期的电池历史运行数据按照时序输入所述第一故障检测模型中；
    通过所述第一故障检测模型，对所述电池历史运行数据进行多维度特征提取，得到对应的故障检测特征数据。
  11. 根据权利要求10所述的方法，其中，
    在所述通过批计算引擎，对所述故障特征数据进行批量计算，得到对应的中间参数之后，所述方法还包括：
    根据所述中间参数，生成对应时序的中间表；
    将所述中间表输入预设的第二数据仓库中，以更新所述第二数据仓库中存储的已有中间表，所述第二数据仓库为实时数据仓库。
  12. 根据权利要求11所述的方法，其中，
    所述通过流计算引擎，实时获取所述中间参数和电池实时运行数据，并利用第二故障检测模型进行融合计算，得到对应电池的故障检测结果，包括：
    通过流计算引擎，从所述第二数据仓库实时获取所述中间参数和电池实时运行数据；
    根据所述电池实时运行数据，确定对应的电池实时状态数据；
    将所述电池实时状态数据与所述中间表中的中间参数进行拼接；
    通过预设的第二故障检测模型，对拼接后的数据进行计算，并输出对应电池的故障检测数据和所述电池的身份信息。
  13. 根据权利要求12所述的方法，其中，
    在所述得到对应电池的故障检测结果之后，所述方法还包括：
    根据所述对应电池的故障检测数据和所述电池的身份信息，生成预警信息；
    将所述预警信息发送至服务子***，以使所述服务子***根据所述预警信息，查询所述电池的身份信息对应的目标终端，并生成对应所述目标终端和所述故障检测数据的服务信息。
  14. 一种电子设备，包括处理器，存储器及存储在所述存储器上并可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求8-13任一所述的电池故障检测方法的步骤。
  15. 一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如权利要求8-13任一所述的电池故障检测方法的步骤。
  16. 一种芯片，所述芯片包括处理器和通信接口，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现如权利要求8-13任一所述的电池故障检测方法的步骤。
  17. 一种计算机程序产品，所述计算机程序产品中的指令由电子设备的处理器执行时，使得所述电子设备执行如权利要求8-13任一所述的电池故障检测方法的步骤。
PCT/CN2022/078479 2022-02-28 2022-02-28 电池故障检测***、方法和设备 WO2023159638A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2022/078479 WO2023159638A1 (zh) 2022-02-28 2022-02-28 电池故障检测***、方法和设备
CN202280020546.6A CN117015774A (zh) 2022-02-28 2022-02-28 电池故障检测***、方法和设备

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/078479 WO2023159638A1 (zh) 2022-02-28 2022-02-28 电池故障检测***、方法和设备

Publications (1)

Publication Number Publication Date
WO2023159638A1 true WO2023159638A1 (zh) 2023-08-31

Family

ID=87764526

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/078479 WO2023159638A1 (zh) 2022-02-28 2022-02-28 电池故障检测***、方法和设备

Country Status (2)

Country Link
CN (1) CN117015774A (zh)
WO (1) WO2023159638A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117725389A (zh) * 2024-02-08 2024-03-19 宁德时代新能源科技股份有限公司 电池出站方法及电池出站***

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107909508A (zh) * 2017-09-26 2018-04-13 国网辽宁省电力有限公司 一种配电变压器负荷异常告警方法
CN111368890A (zh) * 2020-02-26 2020-07-03 珠海格力电器股份有限公司 故障检测方法及装置、信息物理融合***
CN111985561A (zh) * 2020-08-19 2020-11-24 安徽蓝杰鑫信息科技有限公司 一种智能电表的故障诊断方法、***及电子装置
CN113344024A (zh) * 2021-04-22 2021-09-03 华中科技大学 锂离子电池热失控分级预警方法及预警***

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107909508A (zh) * 2017-09-26 2018-04-13 国网辽宁省电力有限公司 一种配电变压器负荷异常告警方法
CN111368890A (zh) * 2020-02-26 2020-07-03 珠海格力电器股份有限公司 故障检测方法及装置、信息物理融合***
CN111985561A (zh) * 2020-08-19 2020-11-24 安徽蓝杰鑫信息科技有限公司 一种智能电表的故障诊断方法、***及电子装置
CN113344024A (zh) * 2021-04-22 2021-09-03 华中科技大学 锂离子电池热失控分级预警方法及预警***

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117725389A (zh) * 2024-02-08 2024-03-19 宁德时代新能源科技股份有限公司 电池出站方法及电池出站***

Also Published As

Publication number Publication date
CN117015774A (zh) 2023-11-07

Similar Documents

Publication Publication Date Title
US20210389290A1 (en) Intelligent Monitoring and Analysis Method for Air Pollution and Device Thereof
CN106209405B (zh) 故障诊断方法及装置
CN103246265B (zh) 机电设备检测维护方法
CN113310647B (zh) 电池包泄露检测的方法、装置、电子设备及存储介质
CN110838075A (zh) 电网***暂态稳定的预测模型的训练及预测方法、装置
CN112492567B (zh) 一种应急指挥通信中的故障分析和解决方法及装置
WO2023159638A1 (zh) 电池故障检测***、方法和设备
CN110674009A (zh) 应用服务器性能监测方法、装置、存储介质及电子设备
CN110188888A (zh) 基于ahp和远程支持的电力设备缺陷管理方法、***和终端
CN112734977B (zh) 一种基于物联网的设备风险预警***及算法
CN108900319B (zh) 故障检测方法和装置
CN114356734A (zh) 服务异常检测方法和装置、设备、存储介质
CN114325405A (zh) 电池组一致性分析方法、建模方法、装置、设备及介质
CN114579659A (zh) 一种梯次利用动力电池利用潜力评估及分选***和方法
CN115509797A (zh) 一种故障类别的确定方法、装置、设备及介质
CN104485991A (zh) 一种光模块故障自动报警方法及***
CN110647070A (zh) 一种用于超大规模数据中心的动力环境监控***
CN115687406A (zh) 一种调用链数据的采样方法、装置、设备及存储介质
CN114398200A (zh) 一种***报错处理方法、装置、电子设备及存储介质
CN114003466A (zh) 一种用于微服务应用程序的故障根因定位方法
CN111737242A (zh) 一种监控海量数据处理过程的方法
CN117880055B (zh) 基于传输层指标的网络故障诊断方法、装置、设备及介质
EP4369204A1 (en) Graph comparison-based behavior analysis
KR102560907B1 (ko) 개선된 학습모델을 통해 재난발생여부를 판단하는 방법 및 그 장치
CN114422324B (zh) 一种告警信息的处理方法、装置、电子设备及存储介质

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 202280020546.6

Country of ref document: CN