Battery thermal runaway risk identification method, early warning method, system and medium

Abstract

This invention relates to a method, early warning method, system, and medium for identifying battery thermal runaway risks in the field of battery diagnostics. The risk identification method includes: acquiring multi-source monitoring data with different update cycles, and generating a set of synchronized judgment frames based on source timestamps, data age, and freshness thresholds at a unified decision-making time. Feature judgment is performed only when all necessary inputs for the feature to be judged are valid, and the feature event that first meets the conditions and its trigger time are latched. Each candidate risk state is judged in parallel based on the latched events and their order, with the risk state being the higher of the historical state and the highest candidate risk state in the current cycle. This invention reduces the risk of temporal distortion and misidentification caused by directly combining data from different time sources by forming a set of synchronized judgment frames. By latching risk feature events, the continuity of risk evidence, the timeliness of risk identification results, and the traceability of the risk evolution process are improved.

Description

Technical Field

[0001] This invention relates to the field of battery diagnostic technology, and in particular to a method, early warning method, system and medium for identifying battery thermal runaway risk. Background Technology

[0002] Batteries may experience thermal runaway under abnormal conditions such as overcharging, internal short circuits, external heating, and mechanical damage. During the development of thermal runaway, monitored parameters such as voltage, temperature, structural expansion force, and the concentration of released gases typically change accordingly. Identifying the battery's risk evolution state through multi-source monitoring data is a crucial technical means for battery safety monitoring. Different types of monitoring data are usually acquired through independent acquisition links and have different update cycles. Signals such as voltage, temperature, and expansion force can be continuously acquired, while data such as gas concentration may be received asynchronously through digital communication interfaces. If data from different sources are directly combined according to their arrival order, or if historical data is used when it is not updated in a timely manner, expired data may be used in conjunction with current data for feature determination, leading to distortion of the data's temporal relationship and affecting the identification of change rates, inflection points, local extrema, and the chronological order of events. Furthermore, online monitoring signals may also be affected by noise and instantaneous fluctuations. If the risk state is updated only based on the determination results of the current cycle, confirmed features may disappear as the signal declines, and the established risk state may be unexpectedly reduced. When multiple candidate risk conditions are met within the same decision-making cycle, updating step-by-step according to fixed intermediate states may delay the output of higher-risk states. Furthermore, anomalies such as sensor disconnection, data timeout, discontinuous acquisition, and communication frame errors can affect feature calculation. If data source failure and battery risk share state variables, acquisition failure may directly cause changes in risk state or overwrite risk information already formed before the failure occurs; if there is a lack of input validity management for different features, local data source anomalies may also cause feature determinations supported by other valid data to be suspended simultaneously. Summary of the Invention

[0003] To address the technical problems of data temporal distortion and unstable risk characteristic evidence in the online processing of multi-source battery monitoring data with different update cycles, this invention provides a battery thermal runaway risk identification method, early warning method, system, and medium.

[0004] Firstly, this invention proposes a battery thermal runaway risk identification method based on multi-source synchronization and state latching. This method calculates the risk state to be identified based on at least two types of battery monitoring data with different update cycles. The battery thermal runaway risk identification method includes the following steps: At least two types of battery monitoring data with different update cycles are appended with source timestamps and cached independently. At the unified decision-making time, the most recent valid data with a source timestamp no later than the unified decision-making time is selected from each independent cache. If the data age of the most recent valid data is not greater than the corresponding freshness threshold, the corresponding most recent valid data is merged into and a synchronization judgment frame set is generated.

[0005] Unqualified data within the synchronized judgment frame set is filtered, and the most recent valid data after filtering is subjected to corresponding feature judgment based on data type. When the feature judgment is met for the first time, a risk feature event is generated and latched. Multiple latched risk feature events constitute a latched event set. Based on the latched event set and risk association rules, the risk level, represented by a quantitative grade, is generated for the current decision-making cycle. The risk level of the current decision-making cycle is compared with the risk state of the previous decision-making cycle, and the higher level is taken as the risk state of the current decision-making cycle.

[0006] Secondly, this invention proposes a battery thermal runaway early warning method based on multi-source synchronization and state latching. This method uses the battery thermal runaway risk identification method based on multi-source synchronization and state latching from the first aspect to obtain the risk state of the current decision-making cycle. Then, it triggers corresponding early warning strategies based on different risk states.

[0007] Thirdly, this invention proposes a battery thermal runaway risk identification system with multi-source synchronization and state latching, which uses the battery thermal runaway risk identification method with multi-source synchronization and state latching described in the first aspect. This battery thermal runaway risk identification system includes: a multi-source signal acquisition module, a synchronization judgment frame generation module, a risk feature event latching module, and a risk state determination module.

[0008] The system comprises several modules: a multi-source signal acquisition module to acquire at least two types of battery monitoring data with different update cycles, and independently cache each data point after appending a source timestamp; a synchronization judgment frame generation module to select the most recent valid data with a source timestamp no later than the unified decision time from each independent cache at the unified decision time; and a risk feature event latching module to perform feature judgment on the most recent valid data in the synchronization judgment frame set according to its data type, generating and latching a risk feature event when the feature judgment is first satisfied. Multiple latched risk feature events constitute a latched event set. A risk status judgment module to generate the risk level, represented by a quantitative grade, for the current decision period based on the latched event set and risk association rules. It compares the risk level of the current decision period with the risk status of the previous decision period, taking the highest level as the risk status of the current decision period.

[0009] Fourthly, the present invention provides a computer-readable storage medium having a computer program / instructions stored thereon. When executed by a processor, the computer program / instructions implement the steps of the battery thermal runaway early warning method with multi-source synchronization and state latching in the first aspect.

[0010] The beneficial effects of this invention are as follows: 1. This invention uses source record validity flags, source timestamps, data age, and freshness thresholds to conduct time validity admission on asynchronous, multi-source monitoring data at different frequencies. It forms a set of synchronous judgment frames at a unified decision time, preventing invalid or expired data from participating in new risk feature judgments. This reduces the risk of time-series distortion and misidentification caused by direct combination of data from different time sources. Furthermore, it transforms the risk feature that first meets the criteria into a latched risk feature event containing the first trigger time and data validity information. Based on the latched events and their order, it determines candidate risk states in parallel, allowing the risk state to be directly updated to the highest-ranking candidate risk state. This risk state does not automatically decrease due to subsequent signal decline or criterion withdrawal, improving the continuity of risk evidence, the timeliness of risk identification results, and the traceability of the risk evolution process.

[0011] 2. This invention implements local isolation of abnormal data through fault vectors independent of risk states, ensuring that abnormalities in data sources and acquisition links do not directly generate or aggravate risk states, nor do they clear existing feature events and risk states. When some data sources are abnormal, other necessary inputs with complete and valid features can still be evaluated; after the abnormality is resolved, relevant feature evaluations are restored through continuous accumulation of valid samples and reconstruction of the calculation window, thereby improving the system's operational continuity and state stability under conditions of channel and communication anomalies. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a battery thermal runaway risk identification method; Figure 2 This is a schematic diagram illustrating the generation of the synchronization determination frame set; Figure 3 This is a schematic diagram of the hardware for independent acquisition of multi-source signals and the data flow. Figure 4 This is a flowchart for generating latch risk characteristic events; Figure 5 This is a flowchart of risk state latching and fault isolation; Figure 6 This is a schematic diagram of the overcharge verification experimental setup and signal connections; Figure 7 This is a graph showing the results of the overcharge verification experiment; Figure 8 This is a schematic diagram of the connection between an electronic device and a multi-battery pack deployment in another embodiment. Detailed Implementation

[0014] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "or / and" as used herein includes any and all combinations of one or more of the associated listed items.

[0016] This embodiment provides a battery thermal runaway risk identification method based on multi-source synchronization and state latching. It calculates the risk state to be identified based on at least two types of battery monitoring data with different update cycles. It can complete data time validity verification before risk feature determination, transforming confirmed risk features into sustainably stored event evidence. Furthermore, it manages risk states and fault states independently to reduce the impact of expired data, signal fluctuations, and acquisition anomalies on the battery thermal runaway risk identification results. Specifically, the battery thermal runaway risk identification method includes the following steps: generating a synchronization determination frame set, risk feature event latching, risk state output, and fault vector masking of abnormal data.

[0017] 1. Generate a set of synchronization determination frames.

[0018] Please refer to Figure 1 This step can be performed locally on the battery pack. Battery monitoring data with different update cycles undergo time validity assessment before entering the risk characteristic determination stage. This is achieved through four stages: independent acquisition and caching, data selection and age determination, freshness gating, and generation of a synchronized judgment frame set. This time validity assessment process does not force all data to be resampled to the same frequency, nor does it wait for low-frequency data to arrive. Instead, it selects data with clear time relationships and still within the valid time limit at a unified decision-making moment.

[0019] like Figure 2As shown, in the first step, the continuous acquisition link and the asynchronous reception link operate separately to acquire at least two types of battery monitoring data with different update cycles. The update cycle includes continuous acquisition and asynchronous reception. The battery monitoring data includes at least two of the following: voltage, temperature, expansion force, and gas concentration. For continuously acquired data such as voltage, temperature, and expansion force signals, the data is written to a high-frequency independent buffer after input protection, signal conditioning, and analog-to-digital conversion. For asynchronously received data such as hydrogen concentration data, the data is received through a digital communication interface and written to a low-frequency independent buffer after checking the frame length, data fields, and verification information. The two links are independent of each other, and low-frequency data waiting and communication jitter do not block high-frequency signal acquisition. Each acquisition record includes at least a channel identifier, data value, source timestamp, and source record validity flag. The source record validity flag is used to characterize whether the original data record itself was received or acquired normally. Continuously acquired data generates a source timestamp after sampling is completed and the range check is passed. When the asynchronous data frame carries the measurement time, the measurement time is mapped to the local monotonic time reference. When it does not carry the measurement time, the complete reception time is used as the source timestamp, and the communication delay is included in the freshness threshold. Figure 3 As shown, in terms of hardware implementation, this embodiment uses an STM32H743 processor and three ADS1258 analog-to-digital converters to acquire 12 channels of individual unit voltage, 30 channels of temperature, and 3 channels of expansion force data, forming a 10 Hz effective update. The hydrogen sensor asynchronously transmits digital frames at 1 Hz. A hardware timer establishes a monotonic time base for acquisition scheduling, data age calculation, causal feature calculation, and event timing. An RTC is used to record absolute date and time. Different data sources are written to independent circular buffers. The processor locally maintains latched event registers, risk status registers, and fault vectors. The acquisition task, decision-making task, and communication output task are decoupled from each other, and communication interruptions do not change the local register state.

[0020] The second step is to select data and determine its age. Let R represent the j-th battery monitoring data acquisition record in channel i's cache. i，j The channel identifier is id. i The data value is x i，j The source timestamp is t i，j The valid flag for the source record is q. i，j The j-th battery monitoring data in channel i's cache can be represented as: R i，j =(id i , x i，j , t i，j , q i，j ).

[0021] At the k-th unified decision time T k Only when the source record has a valid flag of q i，j True (q) i，j=1), and the source timestamp is t. i，j No later than the time of unified decision-making T k Collection Records R i，j Select source timestamp t i，j Biggest record: j * =argmax{t i，j |t i，j ≤T k q i，j =1}.

[0022] In the formula, j * To unify decision-making time T k The selected record number is the sequence number of the most recent valid data no later than the time of the unified decision. This selection rule excludes future data and avoids directly concatenating data from different sources according to the order of communication arrival. The data age A is calculated based on the selected most recent valid data. i(k) : A i(k) =T k -t i，j* .

[0023] The third step is to perform freshness gating. In this step, a corresponding freshness threshold τ is first set for each battery monitoring data. i Data Age A i(k) Not exceeding the freshness threshold τ i At that time, the data validity status v of the selected data i(k) Marked as valid (v) i(k) =1). Data age A i(k) Exceeding the freshness threshold τ i At that time, the data validity status v of the selected data i(k) Marked as invalid (v) i(k) =0), and stops participating in subsequent feature determination. Data valid state v i(k) It can be represented as:

[0024] The freshness threshold τ i The freshness threshold can be determined based on the nominal update cycle of the data source, communication jitter, unified decision cycle, and the allowable timing error of the feature. In this embodiment, the freshness threshold of continuously collected data is set to 0.30 s, and the freshness threshold of hydrogen concentration data is set to 2.00 s. As shown in Table 1, an admission result for a unified decision time is given to illustrate that battery monitoring data with different update frequencies are judged independently according to their respective thresholds.

[0025] Table 1: Access results of battery monitoring data at the unified decision-making time.

[0026]

[0027] The fourth step involves forming a set of synchronization decision frames. As shown in Table 1, the most recent data age for all four data categories has not exceeded the corresponding freshness threshold, therefore they are all written into the synchronization decision frame set as valid data. Invalid data is not included in the synchronization decision frame set, ensuring that invalid or expired data does not participate in new risk feature determinations, reducing the risk of temporal distortion and misidentification caused by directly combining data from different time sources. Table 1 also shows that the synchronization decision frame set does not require all data sources to update at the same time, but rather requires each data item within the frame to have a clear time source and valid status. Therefore, the resulting synchronization decision frame set includes at least the unified decision time, data value, source timestamp, data age, and data valid status.

[0028] II. Risk Characteristic Event Locking.

[0029] Please refer to this again. Figure 1 After obtaining the set of synchronization judgment frames, it is necessary to convert valid continuous signals into traceable risk feature events and latch the risk feature events that first meet the feature judgment criteria. This includes the following steps: First, the data participating in the feature judgment is checked for validity and filtered. The data participating in the feature judgment includes: data values ​​and data validity status in the set of synchronization judgment frames, and the required input set M of the j-th risk feature to be judged. j The required input set M j This includes at least one current data point, historical data point, or feature parameter calculated from relevant data that is essential for determining the occurrence of a corresponding risk characteristic or event. Required input set M j The input data items are pre-configured according to the judgment rules of the corresponding risk characteristics. For example, the required input set M j This can include: current sampled values, historical data, derived rates of change, extreme values, data required for threshold comparison, and time series data required for duration statistics. More specifically, for the voltage sag risk feature, the required input set must include the current voltage and historical voltage. For the expansion force decrease risk feature, the required input set must include continuous expansion force data. For the turning point risk feature, the required input set must include time series data before and after the turning point. For the extreme value risk feature, the required input set must include the data sequence within the effective observation window. Therefore, each risk feature requires a different required input set.

[0030] like Figure 4As shown, after knowing the data values ​​and data validity status in the synchronization judgment frame set and the values ​​in the required input set, the system checks whether all values ​​are valid. The standard for validity is whether the input data meets the requirements for completeness, validity, and temporal consistency for participating in the feature judgment of the current cycle. For example, the following situations can be considered invalid: data acquisition failure, data communication loss, data exceeding the allowed physical range, sensor failure, abnormal data timestamp or exceeding the allowed synchronization window, data preprocessing failure, insufficient historical data window for features requiring historical data, etc. When all the above checks are valid, the feature judgment of the corresponding risk feature is performed. Otherwise, it is filtered. Furthermore, if any required input set is invalid, no new risk feature event is formed in this decision cycle. Already latched risk feature events remain unchanged. This step places the data validity check before the risk feature calculation, which can avoid abnormal inputs forming new risk evidence.

[0031] After confirming that all data participating in feature determination is valid, feature determination is performed according to the data type. Feature determination is calculated based on current and historical data in chronological order, without invoking future data. The forms of feature determination include, but are not limited to: threshold determination, duration determination, rate of change determination, turning point determination, and extreme value determination. For example, the rate of change and turning point can be obtained using backward differencing and causal filtering. Among the generated risk feature events, turning point events are confirmed by the rate of change crossing positive and negative hysteresis thresholds. Extreme value events are confirmed by changes in the sign of the rate of change, extreme value significance, and minimum time interval. Threshold events are confirmed by a threshold condition and duration. When the above risk feature criteria are met for the first time, risk feature event E is generated. j The latched risk characteristic events are then written to the latch event register and latched. Multiple latched risk characteristic events constitute a latch event set. Subsequent signal fallback, criterion exit, or corresponding data source failure does not cancel the latched risk characteristic events. The recorded content of the latched risk characteristic events includes: event flag b. j First trigger time t j，first Trigger value z j The source timestamp T of the data used for judgment j and input validity snapshot q j Risk Characteristic Event E j It can be represented as: E j =(b j , t j，first , z j T j q j ).

[0032] Among them, event flag b j Distinguish between confirmed events and current continuous signals. First trigger time t j，firstEstablish the event time sequence. Trigger value z j Used for reviewing criteria. Source timestamp T j and input validity snapshot q j The evidence proves that the event was formed from valid data. The meaning and function of each record are shown in Table 2.

[0033] Table 2: Record items for risk characteristic events.

[0034]

[0035] III. Risk Status Output.

[0036] After obtaining the set of latched events, the risk level C, characterized by quantitative levels, is generated based on the set of latched events and risk association rules in the current decision-making cycle. (k) The risk association rule is a correlation rule between risk characteristic events, trigger thresholds, and quantification levels. It can be pre-set based on expert experience or actual operating condition testing. For example, the risk characteristic events in the risk association rule might be a voltage plateau event and an expansion force decrease inflection point event. The required input sets for these two risk characteristic events are already known. Therefore, the quantification level can be obtained by comparing the values ​​in the required input sets with the corresponding trigger thresholds in the risk association rule. For example, if the trigger threshold is a voltage plateau period greater than or equal to 20 seconds and the first appearance of an expansion force inflection point, the quantification level for determining the risk degree is C1. Similarly, if the risk characteristic event is a rapid change in expansion force, and the trigger threshold is an initial expansion force change rate > 0.5 kg / s and a state duration ≥ 5 seconds, the quantification level is C2. Another risk characteristic event might be a maximum expansion force event and a safety valve area temperature anomaly event. The trigger threshold is a safety valve temperature > 75℃ and a hydrogen concentration > 100 ppm, and a state duration ≥ 5 seconds, resulting in a quantification level of C3. The above examples only illustrate the expression form of the risk association rule and do not limit the number of states, event types, or quantification methods.

[0037] like Figure 5 As shown in (a), at least two candidate risk states are set in the current decision-making period k. Each candidate risk state corresponds to a risk association rule. All risk association rules are determined in parallel based on risk characteristic events. When at least one risk association rule is determined to be true, the candidate risk state with the highest quantification level is selected as the highest risk level C in the current decision-making period. (k),max : .

[0038] In the formula, r represents the state number arranged according to the degree of risk. E (k) This represents the set of latched events for the current decision period k. Let represent the candidate risk association rule for the r-th candidate risk state. Then, compare the highest risk level C in the current decision-making cycle. (k),max Risk status L compared to the previous decision-making cycle (k-1) The highest value of L is taken as the risk state L for the current decision-making cycle. (k) : L (k) =max{ L (k-1) C (k),max}

[0039] In parallel decision-making, when one or more candidate risk states higher than the current risk state are established, the system directly updates to the one with the highest risk level, without requiring the process to go through intermediate risk states sequentially. Established higher-risk states are not automatically reduced due to subsequent signal decline, criterion exit, or data source failure, thereby improving the continuity of risk evidence, the timeliness of risk identification results, and the traceability of the risk evolution process. Finally, the risk state L output in this decision cycle is... (k) Some structured output information (such as trigger time, trigger event, etc.) can be output to external systems, and perform corresponding actions or other purposes through host computers, execution systems, etc.

[0040] IV. Fault vector masking of abnormal data.

[0041] Please refer to Figure 5 In (b), while generating the risk state, battery monitoring data that exceeds the range, exceeds the data age limit, is discontinuously acquired, or has an incorrect data frame are respectively set to the corresponding fault bits in the fault vector for masking and alarming. Specifically, the fault vector updates the input effective mask, only masking new event judgments that depend on abnormal battery monitoring data, while other features with complete inputs continue to be judged. The fault vector is not used as the trigger for the establishment of candidate state rules, the latched event register is maintained, and the current risk state continues to increase with the support of other effective events. After the anomaly is cleared, if the number of effective samples continuously obtained from the corresponding battery monitoring data reaches the maximum number of samples required for the calculation window of each relevant feature dependent on the data source, and filtering, differencing, and reconstruction of the relevant feature calculation window are completed, then participation in feature judgment is restored. For battery monitoring data i, the expression for the number of effective samples required for restoration is: N i =max{W j |i∈M j}

[0042] In the formula, W jis the calculation window length for risk feature j. The corresponding data remains invalid until the window reconstruction is complete. If a fault occurs before event latching, new events dependent on this data source cannot be generated. If a fault occurs after event latching, existing events and risk states are not cleared, thereby reducing misjudgments caused by asynchronous frequency data, signal fluctuations, and acquisition failures, and improving the timing accuracy, state continuity, and fault tolerance of battery thermal runaway risk identification.

[0043] To verify the implementation of time-validity access, risk characteristic event latching, risk status determination and latching, and fault isolation in this invention, an overcharge experiment was conducted. Please refer to... Figure 6 A square hard-shell battery was placed inside a metal casing, and voltage, battery surface temperature, safety valve area temperature, expansion force, and hydrogen concentration were collected. A paperless recorder independently collected voltage, temperature, and expansion force for reference. An overcharge condition was applied by a charge-discharge cycler, and a camera recorded the opening, discharge, and combustion phenomena of the safety valve. The host computer recorded the synchronous judgment frame set, latched events, risk states, and fault vectors. This setup provides three types of evidence simultaneously: online system, independent reference, and external phenomena. The real experiment used a 1C overcharge method to trigger battery thermal runaway. The online system and the external independent reference used their own acquisition links and did not share event judgment results. The test results are as follows: Figure 7 As shown, Figure 7 The temperature, expansion force, and voltage comparison curves shown in (a)-(d) indicate that the acquisition module and the independent reference acquisition have the same trend and main characteristic locations before thermal runaway. Figure 7 The risk state curve in (e) gives the formation time of risk states C1 to C4 and the relative early identification time of thermal runaway. Figure 7 The fault vector curve in (f) shows the location of the fault after the temperature channel fails, while C4 remains unchanged.

[0044] Online event trigger time t i,online With offline reference time t i,offline The deviation is calculated using the following formula, with a positive value indicating that the online trigger is later than the offline reference.

[0045] Δt= t i,online -t i,offline .

[0046] The triggering results for online and offline times are shown in Table 3.

[0047] Table 3: Results and deviations of online event triggering in the 1C overcharge verification experiment.

[0048]

[0049] Table 3 shows the online trigger time and offline reference time for each risk characteristic event. For events with offline reference values, the absolute value of their online trigger deviation does not exceed 7 s. This result indicates that independent acquisition, synchronous judgment frame set, and risk characteristic calculation can reproduce the sequence of major events under online conditions. According to the risk association rules, risk states C1, C2, C3, and C4 are formed at 2207 s, 2821 s, 3689 s, and 4124 s, respectively. Taking 4145 s as the thermal runaway confirmation time, the advance identification times of risk states are 1938 s, 1324 s, 456 s, and 21 s, respectively. After thermal runaway, some temperature channels fail, marking the corresponding power monitoring data as invalid and setting the fault vector. New events dependent on this channel stop forming, while the already latched events and C4 remain unchanged.

[0050] In another embodiment, a battery thermal runaway early warning method based on multi-source synchronization and state latching is proposed. This method uses the battery thermal runaway risk identification method based on multi-source synchronization and state latching described in the above embodiment to obtain the risk state of the current decision-making cycle. Then, corresponding early warning strategies are triggered based on different risk states. For example, when the risk state is C1, the battery charging operation is immediately stopped. When the risk state is C2, charging is immediately stopped and nitrogen is simultaneously injected into the battery compartment to reduce the ambient oxygen concentration. When the risk state is C3, charging is immediately stopped and a forced ventilation exhaust system is activated to remove flammable and explosive gases such as hydrogen and carbon monoxide. The early warning strategies can be preset based on expert experience or actual conditions, and no further restrictions are imposed here.

[0051] In another embodiment, a battery thermal runaway risk identification system with multi-source synchronization and state latching is proposed, which uses the battery thermal runaway risk identification method with multi-source synchronization and state latching described in the above embodiment. This battery thermal runaway risk identification system includes: a multi-source signal acquisition module, a synchronization judgment frame generation module, a risk feature event latching module, and a risk state determination module. The multi-source signal acquisition module acquires at least two types of battery monitoring data with different update cycles, and caches them independently after attaching source timestamps. The synchronization judgment frame generation module selects the most recent valid data with a source timestamp no later than the unified decision time from each independent cache at a unified decision time. If the data age of the most recent valid data is not greater than the corresponding freshness threshold, the corresponding most recent valid data is incorporated into the synchronization judgment frame set. The risk feature event latching module performs corresponding feature determination on the most recent valid data in the synchronization judgment frame set according to the data type, and generates and latches a risk feature event when the feature determination is first satisfied. Multiple latched risk feature events constitute a latched event set. The risk state determination module generates the risk level characterized by quantitative levels in the current decision period based on the latched event set and risk association rules. Among them, the risk association rules are the association rules between risk characteristic events, trigger thresholds, and quantification levels. The risk level of the current decision-making cycle is compared with the risk status of the previous decision-making cycle, and the higher level is taken as the risk status of the current decision-making cycle.

[0052] In another embodiment, an energy storage battery pack is also proposed. This energy storage battery pack includes a battery thermal runaway risk identification system with multi-source synchronization and state latching as described above. Upon obtaining the risk state of the current decision cycle, a corresponding early warning strategy is triggered.

[0053] In another embodiment, an electronic device, such as Figure 8 As shown, it includes a processor, a memory, a multi-source signal acquisition interface, and a communication output interface. The memory stores computer programs / instructions. When the stored computer programs / instructions are executed by the processor, they implement the steps of the battery thermal runaway early warning method with multi-source synchronization and state latching as described in the above embodiments. The multi-source signal acquisition interface is used to receive power monitoring data from each battery pack, which is independently managed according to module identifiers, when multiple battery packs are deployed. The communication output interface is used to output risk identification results to the host computer, BCU, BMS, or EMS via structured communication frames. The structured communication frame includes at least the module identifier, unified decision time, source timestamp or data age, data validity status, risk characteristic event, risk status, and fault vector. Finally, the corresponding action is executed through an external execution system.

[0054] In another embodiment, a computer-readable storage medium is also proposed, which stores a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the battery thermal runaway risk identification method with multi-source synchronization and state latching described in the above embodiments are implemented. The computer-readable storage medium may include, but is not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0055] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0056] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for identifying battery thermal runaway risk using multi-source synchronization and state latching, characterized in that, It calculates the risk status to be identified based on at least two types of battery monitoring data with different update cycles; Methods for identifying battery thermal runaway risks include: For at least two types of battery monitoring data with different update cycles, append source timestamps and cache them independently. At the unified decision-making moment, the most recent valid data with a source timestamp no later than the unified decision-making moment is selected from each independent cache; if the data age of the most recent valid data is not greater than the corresponding freshness threshold, the corresponding most recent valid data is incorporated into the synchronization decision frame set. Unqualified data within the set of synchronous judgment frames are filtered, and the most recent valid data after filtering is judged according to the data type. When the feature judgment is met for the first time, a risk feature event is generated and latched; multiple latched risk feature events constitute a latched event set. Based on the latched event set and risk association rules, the risk level in the current decision-making cycle is generated using quantitative levels; the risk level in the current decision-making cycle is compared with the risk status in the previous decision-making cycle, and the highest level is taken as the risk status in the current decision-making cycle.

2. The battery thermal runaway risk identification method based on multi-source synchronization and state latching according to claim 1, characterized in that, The source timestamps for collecting multiple battery monitoring data use the same monotonic time base; The data age is the difference between the unified decision time and the source timestamp of the most recently selected valid data; if the data age exceeds the corresponding freshness threshold, the data validity status of the corresponding data is marked as invalid and is not included in the synchronization judgment frame set. The set of synchronization determination frames includes the unified decision time, data value, source timestamp, data age, and data validity status.

3. The battery thermal runaway risk identification method based on multi-source synchronization and state latching according to claim 1, characterized in that, The feature determination is calculated based on current data and historical data in chronological order, and includes at least one of the following: threshold determination, duration determination, turning point determination, and extreme value determination; The recorded content of the risk characteristic event includes: event flag, first trigger time, trigger value, source timestamp of the data involved in the judgment, and input validity snapshot; the locked risk characteristic event will not be revoked due to subsequent decline in battery monitoring data values, criterion exit, or corresponding data source failure.

4. The battery thermal runaway risk identification method based on multi-source synchronization and state latching according to claim 1, characterized in that, Methods for filtering data involved in feature determination include: Determine whether the data values ​​and data validity status in the synchronization frame set are valid; Configure the required input set M for the j-th risk feature to be determined. j The required input set M j This includes at least one current data, historical data, or feature parameter calculated from the data that is necessary to determine the corresponding risk characteristics; When both the synchronous judgment frame set and the required input set are valid, the corresponding risk feature is judged; otherwise, it is filtered; and when any required input set is invalid, no new risk feature event is formed in this decision cycle.

5. The battery thermal runaway risk identification method based on multi-source synchronization and state latching according to claim 1, characterized in that, Risk association rules are the association rules between risk characteristic events, trigger thresholds, and quantification levels; Methods for generating the highest risk level in the current decision-making cycle include: Set at least two candidate risk states; each candidate risk state corresponds to a risk association rule; In the current decision-making cycle, all risk association rules are determined in parallel based on risk characteristic events; when at least one risk association rule is determined to be true, the candidate risk state with the highest quantification level is selected as the highest risk level in the current decision-making cycle.

6. The battery thermal runaway risk identification method based on multi-source synchronization and state latching according to claim 1, characterized in that, Battery monitoring data that exceeds the range, data age limit, data acquisition is discontinuous, or data frame is incorrect will be set to the corresponding fault bit in the fault vector for shielding and alarm. After the anomaly in the battery monitoring data ends, if the number of valid samples obtained continuously from the corresponding battery monitoring data reaches the maximum number of samples required for the calculation window of each relevant feature dependent on the battery monitoring data, then participation in feature determination will resume. During the shielding and recovery process, the battery monitoring data indicates that the risk status remains unchanged or continues to increase.

7. The battery thermal runaway risk identification method based on multi-source synchronization and state latching according to claim 1, characterized in that, The battery monitoring data includes at least two of the following: voltage, temperature, expansion force, and gas concentration; the update cycle includes two types: continuous acquisition and asynchronous reception. Continuously acquired data and asynchronously received data are written to their respective independent buffers; The asynchronously received data is appended with a source timestamp and a source record validity flag after the data frame integrity check.

8. A battery thermal runaway early warning method with multi-source synchronization and state latching, characterized in that, It uses the battery thermal runaway risk identification method with multi-source synchronization and state latching as described in any one of claims 1 to 7 to obtain the risk status of the current decision cycle; Different early warning strategies are triggered based on different risk states.

9. A battery thermal runaway risk identification system with multi-source synchronization and state latching, characterized in that, It uses the battery thermal runaway risk identification method with multi-source synchronization and state latching as described in any one of claims 1 to 7; The battery thermal runaway risk identification system includes: A multi-source signal acquisition module is used to acquire at least two types of battery monitoring data with different update cycles, and cache them independently after attaching source timestamps; The synchronization decision frame generation module is used to select the most recent valid data with a source timestamp no later than the unified decision time from each independent cache at the unified decision time; if the data age of the most recent valid data is not greater than the corresponding freshness threshold, the corresponding most recent valid data is incorporated into the synchronization decision frame set. The risk feature event latching module is used to perform corresponding feature judgment on the most recent valid data in the synchronization judgment frame set according to the data type. When the feature judgment is met for the first time, a risk feature event is generated and latched. Multiple latched risk feature events constitute a latched event set. The risk status determination module is used to generate the risk level in the current decision cycle by using quantitative levels based on the latched event set and risk association rules; compare the risk level in the current decision cycle with the risk status in the previous decision cycle, and take the highest level as the risk status in the current decision cycle.

10. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instruction is executed by the processor, it implements the steps of the battery thermal runaway risk identification method with multi-source synchronization and state latching as described in any one of claims 1 to 7.

Images