A lithium battery full-temperature-range accurate perception method based on an intelligent sensing system

Abstract

The present application relates to lithium battery electric parameter measurement, intelligent sensing and battery management data processing technical field, specifically to a kind of lithium battery full temperature domain accurate perception method based on intelligent sensing system.Voltage, current and temperature data output by different communication protocols are obtained, measurement semantic records containing source protocol identifier, sampling identifier, receiving time, update sequence and effective state are established, current excitation events are extracted, dynamic time delay and sampling phase deviation are determined according to current excitation, voltage response and temperature continuous change relationship, data is reconstructed to unified physical time axis, reconstruction residual is jointly decomposed, protocol time base deviation, temperature related measurement deviation and measurement uncertainty residual are obtained, and dynamic time delay and temperature deviation trajectory are iteratively updated.The present application can distinguish communication timing dislocation and temperature related measurement change, reduce cross-protocol data association error, and keep voltage, current and temperature measurement continuous and consistent under full temperature domain conditions.

Description

Technical Field

[0001] This invention relates to the fields of lithium battery electrical parameter measurement, intelligent sensing, and battery management data processing technology, specifically a method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system. Background Technology

[0002] Existing intelligent sensing systems for lithium batteries typically collect state parameters such as voltage, current, and temperature through the battery management system (BMS) and transmit the collected results to the control terminal for state analysis using communication protocols. In multi-protocol BMS integration applications, common technical solutions mainly employ protocol parsing and field mapping to convert measurement data from different sources into a unified data format. For example, different battery management nodes generate data messages containing voltage, current, and temperature information according to their respective communication protocols. After receiving the messages, the intelligent sensing system parses the corresponding data fields according to the protocol type and performs data normalization processing according to a preset data structure. After unifying the data format, the system typically sorts the data from different sources by time based on the message reception time. Through timestamp matching, linear interpolation, or fixed delay compensation, the data streams from different protocols are made into a continuous data sequence, and then the processed data is input into the battery state estimation, anomaly detection, or operation control process. This type of solution mainly focuses on whether the communication data can be parsed and transmitted normally, lacking further analysis of the internal sampling mechanisms of different protocols, data refresh logic, and the offset relationship between the actual sampling time of the source and the reception time. Therefore, in the process of multi-protocol data fusion, it is usually assumed that the data from different sources are consistent in the time dimension.

[0003] To address the full-temperature-range sensing requirements of lithium batteries, existing technologies also incorporate temperature compensation models to correct the collected voltage and current data. Since lithium batteries exhibit changes in electrochemical characteristics across different temperature ranges, some systems establish a correlation between temperature and measurement deviation, selecting appropriate compensation parameters based on the current temperature value to adjust the voltage and current measurement results. For example, in low-temperature operation, the system increases the corresponding compensation amount according to a preset temperature range to correct measurement offsets caused by polarization effects; in high-temperature operation, it corrects long-term drift based on historical calibration data. Simultaneously, some solutions collect data trends over multiple periods, using data smoothing, averaging, or historical trend prediction to reduce the impact of measurement fluctuations. These solutions can improve measurement stability within a single protocol, fixed sampling mechanism, and preset temperature range. However, measurement deviations caused by temperature changes are typically processed separately from factors such as communication delays and sampling phase differences, failing to consider how time misalignments during transmission, buffering, and refresh of data from different protocols alter the correlation between voltage, current, and temperature. Therefore, it is difficult to guarantee that the measurement data after multi-protocol fusion still corresponds to the same actual battery state.

[0004] During cross-protocol BMS data fusion, protocol transmission delays, sampling time offsets, and temperature-related measurement changes all affect voltage, current, and temperature data, making it difficult for intelligent sensing systems to distinguish between data time misalignment caused by communication processes and actual measurement changes caused by battery temperature variations. When data from different protocols are directly aligned according to reception time, voltage data received at a certain moment may correspond to an earlier battery state, while synchronously received current data may reflect a new operating state, resulting in a loss of true correspondence between voltage response and current excitation. When the system further uses a temperature compensation model to correct the data, because the compensation process is based on the already misaligned data relationship, it is easy to mistake changes caused by protocol delays for temperature-related deviations, causing the compensation result to deviate from the actual measurement state. Due to the lack of a joint identification mechanism for protocol time base deviations and temperature measurement deviations, existing solutions struggle to maintain a unified measurement time reference under different communication protocols, different temperature ranges, and dynamic current change conditions, leading to data association errors and decreased measurement consistency during the full-temperature-range state sensing of lithium batteries. Summary of the Invention

[0005] The purpose of this invention is to provide a method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system, which can solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system includes: acquiring voltage, current and temperature measurement data output by the BMS via different communication protocols, and establishing a measurement semantic record for each measurement data, including source protocol identifier, sampling identifier, receiving time, update sequence and valid state; Extract current excitation events from current measurement data, and determine the dynamic delay and sampling phase deviation of each source protocol data relative to the unified physical time axis based on the correspondence between the current excitation events and voltage response and continuous temperature change. The time position of each measurement data is reconstructed according to the determined dynamic time delay and sampling phase deviation. The residuals formed by the reconstructed voltage, current and temperature data are jointly decomposed to obtain the true trajectory of electrical parameters, protocol time base deviation and temperature-related measurement deviation. The reconstruction results are then iteratively corrected using the protocol time base deviation and the temperature-related measurement deviation.

[0007] Preferably, the establishment of the measurement semantic record includes: parsing the measurement value update relationship in the continuous messages according to the source protocol, marking the messages that continuously maintain the same measurement value and the update sequence advances as hold events, marking the messages whose measurement value changes and the update sequence changes synchronously as refresh events, and associating the sampling identifier continuity, the receiving time interval and the change of valid state with the hold events and refresh events to form the source end measurement event chain corresponding to each voltage, current and temperature measurement data; The source sampling time window that may correspond to the measurement value is defined by the hold interval between adjacent refresh events, and the source sampling time window is written into the corresponding measurement semantic record.

[0008] Preferably, the determination of the dynamic delay and the sampling phase deviation includes: generating current excitation event markers for amplitude change segments and direction change segments in the current measurement data; extracting the voltage change direction, change start position and local change sequence adjacent to the current excitation event from the voltage measurement data of each source protocol; and simultaneously extracting the continuous change interval of the corresponding temperature measurement data. Based on the current excitation event, candidate correspondences between current excitation and voltage response are established within the sampling time window of the candidate source end of each source protocol, and candidate correspondences that conflict with the continuous temperature change interval are eliminated to obtain the time offset sequence of each source protocol.

[0009] Preferably, the joint decomposition of the residuals includes: generating a cross-protocol reconstruction trajectory based on the reconstructed voltage, current, and temperature data; calculating the temporal residual of the voltage measurement trajectory relative to the cross-protocol reconstruction trajectory; establishing the residual that changes synchronously with the boundary of the current excitation event and changes with time offset correction as the protocol time base residual; establishing the residual that continuously migrates with the temperature range across multiple current excitation events as the temperature-related residual; and retaining the portion not explained by the two types of residuals as the measurement uncertainty residual. The dynamic delay is updated according to the protocol time base residual, the temperature deviation trajectory is updated according to the temperature-related residual, and then the cross-protocol reconstruction trajectory is regenerated with the updated dynamic delay and the temperature deviation trajectory.

[0010] Preferably, the limitation of the source sampling time window includes: for adjacent refresh events within the same source protocol, comparing the order of joint changes in voltage, current and temperature measurements before and after refresh, merging the segment where only message transmission occurs but the measurement value is not synchronously updated into the previous holding interval, and determining the segment where at least one electrical measurement value shows a new change and the corresponding sampling identifier advances as the sampling boundary; Cross constraints are established for the sampling boundaries associated with the same current excitation event in different source protocols. The corresponding time windows of each sampling boundary are compressed according to their order. The sampling time windows of multiple source ends associated with the same physical event are arranged in a unified time order and do not overlap in the opposite direction.

[0011] Preferably, the generation of the time offset sequence includes: continuously selecting multiple current excitation events along the unified physical time axis, forming candidate offset nodes from the candidate voltage response positions corresponding to each current excitation event, and limiting the time offset change direction of adjacent candidate offset nodes by the message update order of the source protocol. For locations where there is no effective voltage response between adjacent current excitation events, segmented time delay trajectories are constructed based on the candidate offset nodes before and after and their corresponding source sampling time windows. Adjacent segmented time delay trajectories are connected only within the range allowed by the source sampling time window to form a dynamic time delay sequence that changes continuously with the protocol data stream.

[0012] Preferably, the screening of the candidate correspondence further includes: performing forward matching from the current excitation position to the voltage response position and reverse matching from the voltage response position back to the current excitation position for the same current excitation event, and retaining only the candidate correspondences where the forward matching and the reverse matching point to the same source sampling time window; When multiple voltage responses simultaneously satisfy bidirectional matching, the candidate range is limited according to the established time offset continuity relationship of the adjacent current excitation events, and voltage responses that cannot maintain the continuity of offset order are retained as undetermined corresponding items, and undetermined corresponding items are not written into the time offset sequence.

[0013] Preferably, the establishment of the temperature-related residuals includes: associating the stable measurement segments between adjacent current excitation events with the corresponding temperature ranges according to the unified physical time axis, and extracting residual components that recur in different temperature ranges and do not change synchronously with the protocol time base correction; Based on the continuous migration sequence of the temperature measurement data, the residual components of adjacent temperature ranges are connected to form a temperature deviation trajectory that is continuously updated with temperature changes. When a temperature range changes, the temperature deviation trajectory is updated only using the residual components of the stable measurement segment that exist before and after the change. Temperature ranges lacking the stable measurement segment are marked as intervals to be supplemented, and the corresponding deviation status is maintained without direct replacement across intervals.

[0014] Preferably, after the dynamic delay sequence is formed, the dynamic delay sequence is reverse-mapped to the refresh events of each source protocol, and the correction position of each refresh event on the unified physical time axis is re-determined. Based on the corrected refresh event position, the corresponding source sampling time window is recompressed, and candidate offset nodes are generated again with the recompressed source sampling time window. The refresh event correction position and candidate offset node generated in two consecutive rounds are continuously compared. Only the correction results that simultaneously satisfy the source protocol update order, current excitation event order and the source end sampling time window inclusion relationship are retained. The correction results are written back to the dynamic delay sequence to form a closed-loop iterative constraint across protocol time bases.

[0015] Preferably, the update of the temperature deviation trajectory further includes: for each temperature range with the stable measurement segment, extracting the temperature-related residual difference before and after the protocol time base residual update, and verifying the consistency of the difference with the residual migration direction of adjacent temperature ranges; Only the residual components that continue to migrate after the protocol time base correction and repeat in adjacent stable measurement segments are written into the temperature deviation trajectory, while the residual components that disappear synchronously with the dynamic time delay are transferred back to the protocol time base residual. After reclassification, the voltage, current, and temperature correction data, as well as the measurement uncertainty residuals, on the unified physical time axis are recalculated according to the updated temperature deviation trajectory and dynamic delay.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By establishing a time correction mechanism based on measurement semantic records, current excitation events, and voltage response relationships, data output from different communication protocols can be mapped to a unified physical time axis, restoring the corresponding actual change relationships between voltage, current, and temperature data. This scheme utilizes the electrical response generated by current changes as the basis for time correlation, determining dynamic delay and sampling phase deviation according to the response position in the data streams of different protocols, avoiding state misalignment caused by data arrangement solely based on message reception time. Simultaneously, by classifying the corrected data residuals, changes related to time offset and changes related to temperature migration are identified separately, ensuring that data deviations caused by protocol transmission are not mixed into the temperature compensation process. The time-reconstructed measurement data maintains the temporal consistency between current excitation, voltage response, and temperature changes, enabling the intelligent sensing system to obtain a unified data benchmark during cross-protocol integration and reducing electrical parameter correlation errors caused by asynchronous sampling.

[0017] 2. By jointly and iteratively correcting protocol time base deviation and temperature-related deviation, a stable correlation can be maintained between measurement data from different sources during continuous operation. During dynamic delay correction, the system continuously adjusts the time position of the protocol data based on new current excitation events and voltage response relationships, ensuring the time correction results adapt to communication refresh changes. During temperature deviation correction, the system establishes a continuous deviation trajectory based on residual changes during temperature range migration, allowing measurement offsets caused by temperature changes to be distinguished from protocol timing changes. Since the protocol time base correction results provide a unified time reference for temperature deviation analysis, and the temperature deviation analysis results reduce interference from non-communication factors in time correction, the two form a mutually constraining data processing procedure. This approach reduces the accumulation of measurement deviations between different temperature environments and data from different protocol sources, ensuring the continuity and consistency of the output voltage, current, and temperature data. Attached Figure Description

[0018] Figure 1 A flowchart of the full-temperature-range cross-protocol time-base reconstruction and sensing process for lithium batteries based on an intelligent sensing system; Figure 2 The graph shows the objective function for dynamic time delay. Figure 3 Flowchart for determining cross-protocol measurement event chains, sampling time windows, and dynamic delays; Figure 4 The flowchart shows the joint decoupling and closed-loop correction process for protocol time base residuals and temperature-dependent residuals. Figure 5 This is a joint decomposition diagram of the voltage residuals; Figure 6 This is a diagram illustrating the closed-loop convergence criterion. Detailed Implementation

[0019] In one embodiment, please refer to Figure 1A precise full-temperature-range sensing method for lithium batteries based on an intelligent sensing system includes: acquiring voltage, current, and temperature measurement data output by the battery management system via different communication protocols; establishing a measurement semantic record for each measurement data, including source protocol identifier, sampling identifier, reception time, update sequence, and valid state; extracting current excitation events from the current measurement data; determining the dynamic delay and sampling phase deviation of each source protocol data relative to a unified physical time axis based on the correspondence between the current excitation events and voltage response and continuous temperature changes; reconstructing the time position of each measurement data according to the determined dynamic delay and sampling phase deviation; and forming a system for processing the reconstructed voltage, current, and temperature data. The residuals are jointly decomposed to obtain the true trajectory of electrical parameters, protocol time base deviation, and temperature-related measurement deviation. The reconstructed results are then iteratively corrected using the protocol time base deviation and temperature-related measurement deviation. In engineering implementation, the intelligent sensing system saves each parsed measurement data as an event object with source, time, update status, and validity. For consecutive events from the same source, a fixed order is maintained. The receiving time is not directly equated with the actual sampling time. Instead, the estimated interval of the source sampling time is retained, allowing for correction space in subsequent time base reconstruction. At the same time, a unified physical quantity identifier is used for voltage, current, and temperature mapped by different protocols to avoid the impact of differences in protocol field names on subsequent correlation processing.

[0020] In this embodiment, the measurement semantic record is generated jointly by the protocol parsing result and measurement behavior information. The source protocol identifier is used to distinguish the communication source to which the data belongs. The sampling identifier is used to reflect the source sampling count, frame count, or sequence information that can indicate the sampling order. The receiving time records the local time when the intelligent sensing system completes the message reception. The update sequence records the refresh order of the same measurement quantity in consecutive messages. The valid status records the data verification result, missing status, and status identifier given by the source. When a certain type of protocol does not explicitly provide the actual sampling time, the system does not manually generate a fixed sampling time based on the message arrival time. Instead, it forms a candidate sampling time interval based on adjacent update sequences, receiving times, and refresh events. When a certain type of protocol can provide the source sampling count relationship, the continuity of the sampling count is used to narrow the candidate sampling time interval. The system continues to correct the data by using current change events that occur in common in different protocol data. This transforms the data from different sources from format layer unification to measurement semantic layer unification, while maintaining the source evidence and time evidence corresponding to each measurement value.

[0021] Specifically, after receiving data from different protocols, the intelligent sensing system maintains voltage event queues, current event queues, and temperature event queues according to the source protocol identifier. Each event is associated with at least the parsed measured value, reception time, sampling identifier, update sequence, reference to the previous event, and current valid state. A state relationship is then established between adjacent events, allowing for maintenance, updating, missing, or recovery. When the same message carries multiple measurements, although each measurement shares the message reception time, it still maintains its own update state, without assuming that all measurements were sampled simultaneously by the source. When different messages transmit voltage, current, and temperature separately, the system maintains their respective candidate time positions based on the sampling identifier and update event relationship, enabling the intelligent sensing system to determine whether data near the same reception time truly corresponds to the same battery operating state during subsequent electrical event association processes.

[0022] In this embodiment, the intelligent sensing system extracts current excitation events that can form identifiable terminal voltage responses from continuous current sequences. Event determination is based on the starting point, direction, duration, and relatively stable segments before and after the change in the local current sequence, without directly defining any single-point fluctuation as a current excitation. For sequences where the current changes from a stable state to a continuously rising, continuously falling, or changing direction of charging / discharging, or from a large load to a relatively stable state, event records are established according to the local change boundaries. The event records store the starting position, ending position, direction marker, and adjacent event relationships of the current change. The system then searches for the starting position of voltage changes adjacent to the current excitation in the candidate sampling time windows of each source protocol voltage sequence, while simultaneously reading the temperature migration state within the same candidate interval. This ensures that the timing of the voltage response is constrained by both the physical excitation of the current and the continuity of temperature, rather than simply pairing based on the order of message arrival.

[0023] When establishing a correlation between current excitation events and voltage responses, the system generates a candidate response set for each current event. The candidate response set only includes voltage events that are located within the corresponding source sampling time window and can maintain the continuity of local voltage changes. For voltage jumps that occur at the protocol refresh boundary but lack corresponding changes in adjacent current sequences, they are retained as refresh-related candidates and not immediately written into the electrical response relationship. When the temperature sequence remains continuously and slowly changing during the voltage change, the temperature information serves as the background constraint for matching the current event. When the temperature itself undergoes continuous migration and may change the voltage baseline, the system records the temperature migration position and delays determining the deviation of some voltage changes. Thus, the distinguishing conditions between protocol delay, actual electrical response, and temperature migration are preserved during the time alignment stage.

[0024] For the construction of a unified physical time axis, the system uses a continuous local timing sequence as the carrying reference, but does not require the measurement data of each source protocol to fall directly at the message reception time. Instead, it maintains an independent dynamic delay state and sampling phase state for each source protocol. The dynamic delay state represents the time offset that needs to be corrected within the current data segment between the formation of the source measurement and the system reception. The sampling phase state represents the local misalignment caused by the periodic or asynchronous sampling of the source protocol relative to the common current excitation event. When multiple consecutive current excitations form voltage responses with the same time relationship in a certain source, the system uses the relative positions of these events to update the dynamic delay. When a certain data segment lacks sufficient current excitation, the intermediate segment is constrained by the previously confirmed candidate offset nodes, and the intermediate delay trajectory is not allowed to break through the corresponding source sampling time window, thus avoiding the generation of time positions without source basis by unconstrained interpolation.

[0025] The dynamic delay can be quantified according to the event correspondence between current excitation and candidate voltage response. The candidate dynamic delay for source protocol p is determined by the following expression: ; in, This represents the candidate dynamic delay of the source protocol p relative to the unified physical time axis. This represents the time offset searched within the current source sampling time window. Let K represent the searchable offset set defined by all candidate sampling time windows from the current source, and let K represent the number of current excitation events participating in the current calculation. This indicates the time position of the k-th candidate voltage response in the source protocol p. Indicates the location of the corresponding current excitation event. and These represent the markings used to describe the direction of voltage response and the direction of current excitation, respectively. This indicates the amount of conflict generated by the current temperature migration for the corresponding event. , , Non-negative normalized weights are used to maintain their meaning within the same implementation. For example, if a candidate offset ensures that the voltage responses corresponding to three consecutive current excitations are all within their respective allowed source-side sampling time windows, the current events and voltage responses maintain the same directional relationship, and there is no rapid migration of temperature overlapping with the response, then this candidate offset will form a smaller objective function value. Another candidate offset, even if it is closer to the message reception time, will not be used as the current dynamic delay if it causes any response to exceed the allowed time window or disrupt the event directional relationship.

[0026] Please refer to Figure 2 , Figure 2The horizontal axis represents the relative position of the candidate offset within the source sampling time window using Chinese partitions, gradually transitioning from significantly ahead to significantly behind. The vertical axis represents the objective function value composed of time distance, directional consistency, and temperature migration conflict. Figure 2 The fast sampling bus, industrial control bus, and serial link are displayed in three independent coordinate areas to avoid overlapping curves from multiple sources within the same area, which would obscure their respective optimal offset changes. The lowest neighborhood of the curve in each subplot corresponds to... The candidate value direction is indicated by the gray dashed line, which represents the selected dynamic delay position. When the curve drops rapidly in a certain Chinese partition and forms a trough in a neighboring partition, it indicates a current excitation event of the protocol. With candidate voltage response It has a better time correspondence within this offset range.

[0027] from Figure 2 It can be seen that the objective function values ​​of both the fast sampling bus and the serial link form low value regions on the lag side, indicating that the measurement formation time of these two types of sources has a need for backward correction relative to the reception time; the trough of the industrial control bus is located on the advance side, reflecting that there is a time base compensation relationship in opposite directions between its refresh boundary and the unified physical time axis. Figure 2 The overall objective changes corresponding to different source protocols for different line types are not determined by a single time distance, but are simultaneously influenced by... , , , and The common constraints. If a candidate offset, although close to the receiving time, causes inconsistencies in direction marking or increased temperature migration conflicts, the corresponding curve will not enter the low-value region.

[0028] After completing dynamic delay and sampling phase correction, the system projects voltage, current, and temperature data from different sources onto a unified physical time axis according to the corrected time position. At the same physical time position, it allows the storage of candidate measurement values, source protocol identifiers, and time uncertainty ranges corresponding to multiple sources, without using direct averaging to cover source differences. For voltage data, the system forms local response segments with adjacent current excitation events as boundaries. For current data, it maintains the original direction and time boundary of the excitation events. For temperature data, it forms temperature state trajectories according to the continuous migration sequence. Then, it combines the data from each source to construct a cross-protocol reconstruction trajectory. The cross-protocol reconstruction trajectory is used to represent the electrical parameter change relationship that can be supported by multiple sources under the current time correction conditions, and serves as a reference for subsequent residual classification, rather than replacing the original measurement records of each source.

[0029] When the system first starts running or a source protocol reconnects before sufficient current excitation events have been generated, the intelligent sensing system first performs measurement semantic recording and source sampling time window construction. It does not directly calculate the dynamic delay without the basis of electrical events. At this time, it maintains an initial time base state with range attributes based on the sampling identifier, update sequence, and reception time, and saves the corresponding data as measurements to be corrected. As subsequent current excitation events arrive, the system gradually establishes candidate correspondences between current excitation and voltage response and narrows the dynamic delay range. Only when a new source forms a continuous and reliable event correspondence is the corresponding measurement value written to a determined position on a unified physical time axis. Therefore, the same time base establishment logic is used for system initialization, protocol recovery, and reconnection after operation interruption, without the need to use fixed communication delay as a long-term unchanging correction condition.

[0030] In a preferred embodiment, please refer to Figure 3 The establishment of measurement semantic records further includes parsing the measurement value update relationship in continuous messages according to the source protocol, marking messages that continuously maintain the same measurement value and whose update sequence advances as hold events, and marking messages whose measurement value changes and whose update sequence changes synchronously as refresh events. The sampling identifier continuity, reception time interval, and valid state change are associated with hold events and refresh events to form a source-end measurement event chain corresponding to each voltage, current, and temperature measurement data. The source-end sampling time window that the measurement value may correspond to is defined according to the hold interval between adjacent refresh events, and the source-end sampling time window is written into the corresponding measurement semantic record. In specific processing, continuous messages from the same source are first restored to their original order according to the update sequence, and then the values, sampling identifiers, and valid states of each measurement quantity before and after are compared respectively, so that the message sending behavior and the actual measurement value refresh behavior are kept separate.

[0031] Furthermore, when a voltage value remains unchanged in a series of messages while the update sequence continues to advance, the system does not interpret each message as a new voltage measurement. Instead, it merges the corresponding message into the current hold event and records the earliest and latest reception positions covered by the hold event. When a voltage value subsequently changes and the corresponding sampling marker advances, the position is defined as a new refresh event. At the same time, the source-end sampling time window is established using the hold range between the refresh event and the previous refresh event. If the current measurement value is refreshed earlier within the same message interval, the current event chain has an independent time window boundary. If the temperature value changes only within a longer data interval, the temperature event chain maintains its own hold interval. This allows the system to express the asynchronous updates of different measurements within the same communication source.

[0032] In data storage, the source-end measurement event chain connects hold events, refresh events, invalidation events, and recovery events according to their chronological order. When data verification fails or the source-end valid status indicates that a certain measurement is unavailable, an invalidation event is generated at the corresponding position, interrupting the current refresh relationship, but the preceding and following events are not deleted. When the data is restored to validity, the recovery event and the subsequent first refresh event jointly determine a new candidate sampling range, so that the invalid interval is not misinterpreted as a long-term hold. In the case of repeated message transmission, if the reception time changes but the measurement value, sampling identifier, and update sequence remain unchanged, the repeated message only supplements the communication reception evidence and does not generate new measurement events, so that the source-end event chain reflects the measurement behavior rather than simply the number of communication messages.

[0033] In a preferred embodiment, the limitation of the source-side sampling time window further includes comparing the joint change order of voltage, current, and temperature measurements before and after the refresh for adjacent refresh events within the same source protocol, merging segments where only message transmission occurs but the measurement values ​​are not synchronously updated into the previous hold interval, and determining segments where at least one electrical measurement value shows a new change and the corresponding sampling flag advances as sampling boundaries; establishing cross constraints for sampling boundaries associated with the same current excitation event in different source protocols, compressing their corresponding time windows according to the order of each sampling boundary, and limiting multiple source-side sampling time windows associated with the same physical event to be arranged in a unified time order and without reverse overlap; during the processing, the intersection of time intervals and event order constraints are used to jointly correct candidate boundaries, and boundaries confirmed by the advancement of sampling flags take precedence over boundaries inferred only from the message reception time.

[0034] Specifically, when a voltage refresh from source A is limited to the first candidate time window, and source B forms another candidate time window under the same current excitation, and both sources indicate that the voltage response is before the next current event, the system cross-contracts the time ranges of the two sources according to the order of the common physical events. If there is partial reverse overlap in the original time windows, the system checks the advancement position of each sampling identifier and the order of adjacent refreshes, and removes the overlapping parts that do not conform to the internal event order of the source protocol from the candidate range. If the candidate intervals of the two sources still cannot form a consistent order, they are not forcibly merged into a single time point, but the current time uncertainty is retained and the system waits for subsequent current excitations or refresh events to continue to constrain the process, so that the determination process of the unified time axis always has traceable source-end evidence.

[0035] In a specific operational scenario, a certain source protocol continuously receives five messages containing the same voltage value, and the update sequence continuously changes. The sixth message presents a new voltage value and is accompanied by a sampling marker. The intelligent sensing system classifies the first five messages into the same hold event and uses the sixth message to form a refresh boundary, rather than forming six equally spaced voltage sampling points. If the current event chain within the same time range indicates that the current excitation occurs in the latter half of the hold interval, the sampling time window that the sixth voltage value may correspond to is further compressed according to the causal order of the current excitation and the voltage response. When another protocol source generates an identifiable voltage response near the same current event, the aforementioned time window can be narrowed by utilizing the event sequence across sources. This allows the time uncertainty caused by communication repetition, buffered transmission, and asynchronous sampling at the source to be gradually transformed into calculable dynamic delay information.

[0036] In a preferred embodiment, determining the dynamic delay and sampling phase deviation further includes generating current excitation event markers for amplitude change segments and direction change segments in the current measurement data; extracting the voltage change direction, change start position, and local change sequence adjacent to the current excitation event from the voltage measurement data of each source protocol; and simultaneously extracting the continuous change interval of the corresponding temperature measurement data. Based on the current excitation event, establishing candidate correspondences between current excitation and voltage response within the sampling time window of the candidate source end of each source protocol, and eliminating candidate correspondences that conflict with the continuous temperature change interval, to obtain the time offset sequence of each source protocol. The amplitude change segment is identified by the difference between the local current baseline and the continuous change sequence, while the direction change segment retains the turning point where the current changes from increasing to decreasing or from decreasing to increasing, so that voltage response matching can use event shape rather than a single value.

[0037] In this embodiment, the system establishes a forward search interval and a backward verification interval for each current excitation event. The forward search interval covers the candidate sampling time window corresponding to the current source protocol and the voltage event that matches the dynamic delay allowable range. The backward verification interval is used to check whether there are other current events that can explain the response after the candidate voltage response. The voltage local change sequence is composed of the change direction relationship of several consecutive events after the change start position. Its length is determined by the event boundary instead of using a fixed number of hardware samples. The temperature continuous change interval is determined according to the time range in which adjacent temperature events maintain the same migration direction. When the candidate voltage response must cross a time boundary that is incompatible with the current temperature migration in order to match the current excitation, the system marks the candidate relationship as conflict and does not write it into the confirmed time offset sequence.

[0038] In a preferred embodiment, the generation of the time offset sequence includes continuously selecting multiple current excitation events along a unified physical time axis, forming candidate offset nodes from the candidate voltage response positions corresponding to each current excitation event, and defining the time offset change direction of adjacent candidate offset nodes by the message update order of the source protocol; for positions where there is no effective voltage response between adjacent current excitation events, segmented time delay trajectories are constructed based on the preceding and following candidate offset nodes and their corresponding source sampling time windows, and adjacent segmented time delay trajectories are connected only within the range allowed by the source sampling time window to form a dynamic time delay sequence that changes continuously with the protocol data stream; each candidate offset node simultaneously saves the current event identifier, voltage response event identifier, candidate time window, and current matching status so that the basis for node formation can be restored during subsequent reverse verification of refresh events.

[0039] When there is a lack of available voltage response between two confirmed candidate offset nodes, the system constructs intermediate candidate trajectories based on the dynamic delay change trends of the preceding and following nodes. However, the intermediate trajectories only define the allowed time mapping relationships and do not generate new measurement values. For example, if the preceding offset node corresponds to a larger source delay and the following offset node corresponds to a smaller source delay, the intermediate data can adjust its time position along the continuous direction within its respective sampling time window. However, no intermediate position may cross the boundary between adjacent refresh events. When a new current excitation occurs and forms an identifiable voltage response in the intermediate segment, the new offset node replaces part of the original candidate trajectory. The dynamic delay change is then determined by the internal sequence check of the source protocol, thereby avoiding unfounded jumps during the communication delay change process.

[0040] Preferably, the candidate correspondence is further filtered using a bidirectional matching method. For the same current excitation event, forward matching from the current excitation position to the voltage response position and reverse matching from the voltage response position back to the current excitation position are performed respectively. Only candidate correspondences that point to the same source sampling time window are retained. When multiple voltage responses simultaneously satisfy bidirectional matching, the candidate range is limited according to the already determined time offset continuity of adjacent current excitation events. Voltage responses that cannot maintain the offset order continuity are retained as undetermined correspondences and are not written into the time offset sequence. During forward matching, the starting position of voltage change is selected first within the allowable time relationship of the current event. During reverse matching, the current event that can produce the response is traced back from the candidate voltage response. Only when the same event pair is formed by the two searches is a definite association established.

[0041] Specifically, when voltage response A and voltage response B exist near the same current excitation event, the system calculates the candidate time base relationship between each of them and the current current event. If voltage response A is selected by the current current event in the forward search, and the current current event is also returned when searching backward from voltage response A, and its time offset is continuous with the adjacent confirmed offset node, then voltage response A is determined as a candidate offset node. If voltage response B is closer to the next current event in the backward search, or can only be paired with the current event if it exceeds its source sampling time window, then voltage response B is saved as an undetermined correspondence. Undetermined correspondences are not deleted. When subsequent refresh order, temperature migration, or new current events provide more constraints, they can re-enter the bidirectional matching process, thereby avoiding a one-time change of the entire dynamic time delay trajectory due to local communication refresh changes.

[0042] In a preferred embodiment, please refer to Figure 4 Based on the reconstructed voltage, current, and temperature data, a cross-protocol reconstructed trajectory is generated. The temporal residual of the voltage measurement trajectory relative to the cross-protocol reconstructed trajectory is calculated. The residual that changes synchronously with the boundary of the current excitation event and changes with time offset correction is established as the protocol time base residual. The residual that continuously migrates with temperature range across multiple current excitation events is established as the temperature-dependent residual. The part not explained by the two types of residuals is retained as the measurement uncertainty residual. The dynamic delay is updated according to the protocol time base residual, and the temperature deviation trajectory is updated according to the temperature-dependent residual. The cross-protocol reconstructed trajectory is then regenerated with the updated dynamic delay and temperature deviation trajectory. The cross-protocol reconstructed trajectory adopts the joint constraint of continuity within the same source and consistency of electrical events between different sources, and does not permanently set a certain protocol source as an error-free reference.

[0043] The voltage timing residual can be decomposed according to the following expression: ; in, This represents the voltage timing residual of the j-th data source at the n-th unified time position. This represents the time base offset increment of the j-th source relative to the current cross-protocol reconstruction trajectory. The local response coefficients represent the mapping of the time base offset increment to the voltage residual. This represents the temperature state transition between adjacent time locations. This represents the residual change coefficient corresponding to the temperature state transition. This refers to measurement uncertainty residuals that cannot be explained by time base changes and temperature migration stability; for example, a residual in the correction... If the current excitation boundary decreases synchronously while the current temperature state remains continuous, then the corresponding part is classified as... If a certain residual continues to change along the temperature migration direction among multiple current excitations, then the corresponding part is classified as... The remaining quantity that cannot satisfy both relationships is retained as .

[0044] Please refer to Figure 5 The horizontal axis represents the correction process segments on the unified physical time axis, covering the initial segment, the first segment, the transition segment, the middle segment, the verification segment, the last segment, and the final segment in sequence; the vertical axis represents the amplitude change of the voltage residual in the current reconstruction trajectory. Figure 5 The solid line represents the time base mapping term, the dashed line represents the temperature migration term, the dotted line represents the uncertainty term, and the dotted-dash line represents the total residual. Figure 5 around The meaning of the decomposition is expanded, but the complete formula is not repeated. Instead, the meaning is expanded. , and The contributions of the three sources in the same time process are separated, so that whether each type of residual has a continuous physical interpretation can be directly observed.

[0045] Figure 5 The data shows that the time base mapping term oscillates slightly across multiple segments, indicating that the protocol time offset increment... The effect on the voltage residual is localized and varies with the current excitation boundary; the temperature migration term changes more gradually, and its positive and negative changes are related to the temperature state migration. The continuity is consistent, making it suitable for identifying temperature-related deviations that persist across multiple current excitations. Uncertainties exhibit significant fluctuations near the transition and verification sections, driving peak-valley variations in the total residual. This indicates that time-base correction or temperature migration alone cannot fully explain measurement differences at these locations, and these uncertainties need to be retained. We await further verification based on subsequent refresh sequences, stable temperature zones, or new current events.

[0046] Residual classification is accomplished through cross-event repeated verification, rather than directly determining the deviation category based on a single data segment. When a residual changes only near a current excitation, the system first saves its candidate category and then checks whether the residual continues to change in the same way with dynamic delay correction after subsequent current excitations occur. For temperature-related candidate residuals, they are required to remain continuous in the temperature migration order in adjacent temperature intervals and still have the same migration relationship after protocol time base recalibration. If the category of a candidate residual changes after new data enters, the system updates the source label and classification status corresponding to the residual and regenerates the cross-protocol reconstruction trajectory within the relevant time range, so that the separation process of protocol time base deviation and temperature deviation is constrained by continuous physical evidence.

[0047] In this embodiment, the protocol time base residual and temperature-related residual are processed by alternating updates. In each round of processing, the current temperature deviation trajectory is fixed, the dynamic delay is recorrected by the residual change near the current excitation boundary, and a new dynamic delay is fixed. The residual component that changes continuously with temperature migration is extracted from the adjacent stable measurement segment. Then, the cross-protocol reconstruction trajectory is regenerated based on the updated two types of deviations. If a residual disappears synchronously after the dynamic delay is recalculated, it is removed from the temperature-related candidate set and included in the protocol time base related part. If a residual continues to migrate with temperature state under different dynamic delay states, it continues to be retained as the temperature-related part. If a residual does not satisfy either of the two types of rules, it is written into the measurement uncertainty residual sequence and the original source information is retained.

[0048] In a preferred embodiment, the establishment of temperature-related residuals includes associating stable measurement segments between adjacent current excitation events with corresponding temperature intervals according to a unified physical time axis; extracting residual components that recur in different temperature intervals and do not change synchronously with the protocol time base correction; connecting the residual components of adjacent temperature intervals according to the continuous migration sequence of temperature measurement data to form a temperature deviation trajectory that is continuously updated with temperature changes; when a temperature interval changes, only the residual components with stable measurement segments before and after the change are used to update the temperature deviation trajectory, and the temperature intervals lacking stable measurement segments are marked as intervals to be supplemented, maintaining the corresponding deviation state without direct replacement across intervals; the stable measurement segments are jointly determined based on the conditions that the current change is relatively smooth, there are no new protocol refresh conflicts in the current time range, and the measurement semantics are continuous, and their purpose is to provide a data range that can distinguish between dynamic load response and temperature deviation migration.

[0049] Instead of switching between temperature ranges using discrete compensation tables that are not interconnected, the system establishes adjacency relationships based on the continuous migration sequence in the temperature event chain. When the temperature moves from one range to an adjacent range, the system checks whether there are stable measurement segments available for residual comparison before and after the transition. Only when there is corresponding residual evidence on both sides is the new residual node connected to the existing temperature deviation trajectory. If there is a strong current change during the transition and a stable measurement segment cannot be formed, the system retains the previously confirmed temperature deviation state and writes the current temperature position and the range to be supplemented into the status record. The system does not directly cover the current range with deviations from other temperature ranges that are far away, thus avoiding the mistaken writing of the true voltage response into the temperature deviation relationship due to the simultaneous existence of load changes and temperature migration.

[0050] Furthermore, the updating of the temperature deviation trajectory also includes extracting the temperature-related residual difference before and after the protocol time base residual update for each temperature range with a stable measurement segment, and verifying the consistency of the difference with the residual migration direction of adjacent temperature ranges; only the residual components that continue to migrate after the protocol time base correction and repeatedly appear in adjacent stable measurement segments are written into the temperature deviation trajectory, while the residual components that disappear synchronously with the change of dynamic time delay are transferred back to the protocol time base residual; after the reclassification is completed, the voltage, current and temperature correction data and measurement uncertainty residuals on the unified physical time axis are recalculated according to the updated temperature deviation trajectory and dynamic time delay; this reclassification process is performed after each new stable temperature segment or new time base correction result is formed to keep the temperature deviation trajectory consistent with the current cross-protocol time relationship.

[0051] In a continuous operation scenario across the entire temperature range, as the battery gradually moves from a lower temperature range to an intermediate temperature range and then continues into a higher temperature range, the system records the range migration relationship along the temperature event chain without directly switching independent compensation values. If stable measurement data exists between the lower and intermediate temperature ranges, adjacent temperature deviation nodes are established using the repeated residuals within the two ranges after time-base correction. If dense current excitation continues during the migration from the intermediate temperature range to the higher temperature range, the transition range is marked as to be supplemented. The corresponding deviation node is then supplemented when the corresponding temperature range is passed again and a stable measurement segment is formed during subsequent operation. Through this process, each expansion of the temperature deviation relationship corresponds to traceable stable measurement data and a corrected protocol time-base state.

[0052] In a preferred embodiment, after forming the dynamic delay sequence, the dynamic delay sequence is back-mapped to the refresh events of each source protocol to redetermine the correction position of each refresh event on a unified physical time axis; the corresponding source sampling time window is recompressed based on the corrected refresh event position, and candidate offset nodes are generated again with the recompressed source sampling time window; the correction positions of refresh events and candidate offset nodes generated in two adjacent rounds are continuously compared, and only the correction results that simultaneously satisfy the source protocol update order, the current excitation event order, and the source sampling time window inclusion relationship are retained, and the correction results are written back to the dynamic delay sequence to form a closed-loop iterative constraint across protocol time bases; the back mapping does not modify the original message, but adds the corrected unified time position to the measurement semantic record, so that the original reception time and the correction time are retained simultaneously.

[0053] The closed-loop iterative process can determine whether the time base needs to be reconstructed in the current time segment based on the following comprehensive changes: ; in, This represents the overall change after the q-th round of closed-loop correction. This represents the dynamic delay obtained by the source protocol p in round q. This indicates the dynamic latency of the previous round from the same source. This represents the number of conflict events in round q that violate the source protocol update order, the current excitation event order, or the source-side sampling time window inclusion relationship. This represents a fixed non-negative coefficient that maps the number of conflicting events to the overall change; for example, when the dynamic delay changes of each source in two consecutive rounds are already small, but there is still a refresh event that is mapped outside its source sampling time window. The time base state remains non-zero and continues to trigger time window compression and candidate offset node reconstruction. When the dynamic latency change remains stable and the number of conflict events is 0, the current time base state is used for the subsequent reconstruction of the data segment.

[0054] Please refer to Figure 6 The horizontal axis represents the iterative stages of the closed-loop correction, displayed in Chinese order from the first to the last round; the left vertical axis represents the overall change, and the right vertical axis represents the number of conflict events. The black line corresponds to... The changes in each iteration are represented by the gray dashed line (convergence threshold) and the gray bars (symbols). Contribution to the current round of conflict. Figure 6 The two vertical axes in the diagram correspond to two criteria: continuous time delay variation and structural conflict, respectively, making... and Differences, as well as conflicts in the source protocol update order or sampling time window inclusion relationship, can be judged simultaneously on the same graph.

[0055] from Figure 6 As can be seen, the overall change gradually decreases with each iteration, with a larger decrease in the first to fourth rounds, indicating that the dynamic delay sequence still requires frequent adjustments in the initial correction phase. After the fifth round, the linear decrease slows down, indicating that the candidate offset nodes of each source protocol have gradually stabilized. The number of collisions in the bar chart gradually decreases from a relatively high level in the earlier rounds to zero, reflecting... right The additional effects are constantly weakening. When the broken line approaches and falls below the convergence threshold and the bars no longer appear, it indicates... The item no longer triggers additional penalties, and the current data segment meets the conditions for continuing to write back the dynamic delay sequence and the unified physical time axis.

[0056] In this embodiment, the refresh events after reverse mapping are reordered and checked according to the original source protocol update sequence. If the correction position of a certain refresh event is earlier than the previous refresh event, it indicates that the current dynamic delay trajectory is inconsistent with the source end update relationship. The system cancels the candidate offset node that caused the conflict and restores its undetermined state. If the refresh event sequence is correct but the correction position exceeds the original source end sampling time window, the candidate range is recompressed by using the time constraints of related current excitation events, adjacent refresh events and other sources on the same physical event. After the new candidate range is formed, bidirectional matching from current excitation to voltage response is performed, thereby forming a cyclic verification relationship between dynamic delay estimation, refresh event correction, time window reconstruction and candidate offset node regeneration.

[0057] In another preferred embodiment, the intelligent sensing system saves the iterative state according to continuous data segments. At the end of each data segment, it retains the dynamic delay end state, sampling phase state, temperature deviation trajectory boundary, undetermined corresponding items, temperature interval to be supplemented, and measurement uncertainty residual corresponding to each source protocol. After the start of a new data segment, the state is used as a continuous constraint instead of a fixed calibration value. When the source protocol refresh rule changes, the distribution of the hold events and refresh events in the source event chain changes accordingly. The system regenerates the candidate sampling time window and triggers dynamic delay correction accordingly. When the communication relationship remains stable but the temperature continues to migrate, the system maintains the current time base state and continues to update the temperature deviation trajectory at the location with a stable measurement segment. When both changes occur at the same time, the corresponding change source is distinguished by residual classification.

[0058] In multi-protocol system integration scenarios, each battery management node maintains measurement semantic records and dynamic delay states according to its own source protocol, but shares a unified physical time axis and a set of current excitation events that can be formed by a common current path. When the voltage data of multiple nodes correspond to the same current excitation, each node searches for the voltage response in its own source sampling time window, without directly replacing its own data with the voltage response of other nodes. If a node temporarily lacks voltage data, the corresponding dynamic delay state only continues within the established candidate range, and the data state during the missing period is written into the event chain. When the node reappears with voltage refresh and can form a bidirectional match with the new current excitation, the candidate offset node and time base closed-loop correction are restored to maintain the independence between the data from different nodes.

[0059] In the case of data recovery after data loss, the system checks whether the first refresh after the recovery event can maintain the source protocol update order with the last valid refresh before the loss. If the source sampling identifiers are continuous, the candidate time window after recovery and the end state of dynamic delay before the loss are used together for the search of new offset nodes. If the sampling identifiers are recounted or the sequence relationship is interrupted, the recovery position is regarded as the new time base connection boundary, and the original dynamic delay is directly inherited without crossing the boundary. The same logic is also used in the scenario of protocol switching or data source reinitialization, so that the time base continuity continues only when there is evidence of source event, and retains a new time uncertainty range when there is no evidence, so as to avoid the historical delay state from continuing to accumulate in the new communication state.

[0060] Specifically, the final calibration record on the unified physical time axis must at least maintain the correlation between calibration voltage, calibration current, calibration temperature, source protocol identifier, corresponding refresh event, dynamic delay state, temperature deviation trajectory position, and measurement uncertainty residual. The calibration voltage is mapped to the unified time axis according to the currently determined time position, and the corresponding temperature-related deviation is deducted based on the confirmed temperature deviation trajectory. The calibration current completes the time mapping according to the source time base state and continues to retain its original excitation boundary. The calibration temperature maintains the continuous migration order of the temperature event chain and serves as the index condition for the temperature deviation trajectory. When a record is still associated with an undetermined corresponding item or a temperature range to be supplemented, the record continues to retain the unexplained residual without using neighboring data for forced replacement, so that subsequent processing can distinguish based on the completeness of the data constraints.

[0061] In a continuous processing phase, the system can initially receive voltage, current, and temperature messages from source A and source B, respectively, and generate hold and refresh events. Subsequently, the current sequence changes continuously and generates a common current excitation event. The voltage response of source A is located in its first candidate sampling time window, and the voltage response of source B is located in another candidate time window. The system generates candidate offset nodes and obtains the dynamic time delays of the two sources through forward and reverse matching. After the unified time position is formed, if some voltage residuals of source B disappear with the re-correction of the time delay, the corresponding residuals are included in the protocol time base deviation. If source A still has residual deviations that change continuously with temperature migration, a temperature deviation node is generated in the subsequent stable measurement phase. The verification process is repeated after new current events and temperature ranges appear, so that the time base correction and temperature correction are continuously connected throughout the entire operation.

[0062] When current excitation, temperature migration, and protocol refresh changes occur simultaneously within a similar time range, the system does not directly determine the source of the residual based on a single time point. Instead, it saves candidate correlations between current events and completes classification through subsequent data. For example, if a voltage jump occurs simultaneously after current excitation and near the protocol refresh boundary, the system first checks whether the jump remains in the same position after dynamic time delay adjustment, and then checks whether it repeats in adjacent temperature ranges. If the jump and current excitation form a stable correspondence after adjusting the time offset, it is retained as the true voltage response. If it disappears with time delay changes, it is classified as a protocol time base residual. If it continues to exist along the temperature migration direction after the time base stabilizes, it enters the temperature-related residual trajectory. Multi-factor separation is completed based on the causal order and repeatability between events.

[0063] During the continuous operation of this embodiment, all intermediate states maintain a clear source and time boundary. Dynamic delay is only used for time mapping of the corresponding source protocol data. The source sampling time window is compressed only according to the source refresh evidence and shared physical event constraints. The temperature deviation trajectory is only updated within the data range that has been time-base verified and has continuous temperature migration evidence. Measurement uncertainty residuals are used to store remaining information that cannot be explained by the protocol time base and temperature migration temporarily. Undetermined correspondences are used to store candidate voltage responses that have not yet been confirmed by bidirectional event matching. Temperature intervals to be supplemented are used to mark the locations where stable measurement data is missing, so that each correction can be traced back to the corresponding measurement semantic record, refresh event, current excitation event, and temperature interval.

[0064] Furthermore, when new data provides new constraints on the existing state, the system only performs local reconstruction on data segments that are temporally related to the new evidence. For example, if a new current excitation can confirm a previously uncertain voltage response, the system recalculates the candidate offset nodes and dynamic delay trajectories adjacent to the uncertain corresponding item, without arbitrarily modifying data that is far from the current event and has already passed closed-loop verification. If a new stable temperature segment can fill in the previously missing temperature range, a continuous migration relationship is established from the adjacent confirmed temperature deviation nodes to the current range, and the protocol time base residual is checked again to see if it is affected. If the supplemented temperature trajectory changes the local residual classification, only the voltage correction data for the unified time interval involved is regenerated, thereby maintaining the temporal continuity and state traceability of long-term running data.

[0065] In the data chain formed by the above processing, the original message, measurement semantic record, source measurement event chain, source sampling time window, current excitation event, candidate voltage response, time offset sequence, dynamic delay trajectory, cross-protocol reconstruction trajectory, protocol time base residual, temperature deviation trajectory, and measurement uncertainty residual all retain the previous and subsequent reference relationships. When a certain correction data needs to be reprocessed, its source message and the corresponding event relationship used at that time can be restored along the reference relationship, without having to deduce the intermediate process from the final correction value. In low temperature, continuous temperature rise, continuous temperature fall, and high temperature operation data, the same set of measurement semantics and time base reconstruction mechanism are used. Temperature differences only enter the residual decomposition through the temperature migration evidence formed in the data, without changing the basic processing logic of cross-protocol event association, forming a continuous data processing flow for full temperature range sensing of lithium batteries.

Claims

1. A method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system, characterized in that, include: Acquire voltage, current and temperature measurement data output by BMS via different communication protocols, and establish measurement semantic records for each measurement data, including source protocol identifier, sampling identifier, reception time, update sequence and valid status; Extract current excitation events from current measurement data, and determine the dynamic delay and sampling phase deviation of each source protocol data relative to the unified physical time axis based on the correspondence between the current excitation events and voltage response and continuous temperature change. The time position of each measurement data is reconstructed according to the determined dynamic time delay and sampling phase deviation. The residuals formed by the reconstructed voltage, current and temperature data are jointly decomposed to obtain the true trajectory of electrical parameters, protocol time base deviation and temperature-related measurement deviation. The reconstruction results are then iteratively corrected using the protocol time base deviation and the temperature-related measurement deviation.

2. The method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 1, characterized in that, The establishment of the measurement semantic record includes: parsing the measurement value update relationship in the continuous messages according to the source protocol, marking the messages that continuously maintain the same measurement value and the update sequence advances as hold events, marking the messages whose measurement value changes and the update sequence changes synchronously as refresh events, and associating the sampling identifier continuity, the receiving time interval and the change of valid state with hold events and refresh events to form a source-end measurement event chain corresponding to each voltage, current and temperature measurement data; The source sampling time window that may correspond to the measurement value is defined by the hold interval between adjacent refresh events, and the source sampling time window is written into the corresponding measurement semantic record.

3. The method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 1, characterized in that, The determination of the dynamic delay and the sampling phase deviation includes: generating current excitation event markers for amplitude change segments and direction change segments in the current measurement data; extracting the voltage change direction, change start position and local change sequence adjacent to the current excitation event in the voltage measurement data of each source protocol; and extracting the continuous change interval of the corresponding temperature measurement data. Based on the current excitation event, candidate correspondences between current excitation and voltage response are established within the sampling time window of the candidate source end of each source protocol, and candidate correspondences that conflict with the continuous temperature change interval are eliminated to obtain the time offset sequence of each source protocol.

4. The method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 1, characterized in that, The joint decomposition of the residuals includes: generating a cross-protocol reconstruction trajectory based on the reconstructed voltage, current, and temperature data; calculating the temporal residual of the voltage measurement trajectory relative to the cross-protocol reconstruction trajectory; establishing the residual that changes synchronously with the boundary of the current excitation event and changes with time offset correction as the protocol time base residual; establishing the residual that continuously migrates with temperature range across multiple current excitation events as the temperature-related residual; and retaining the portion not explained by the two types of residuals as the measurement uncertainty residual. The dynamic delay is updated according to the protocol time base residual, the temperature deviation trajectory is updated according to the temperature-related residual, and then the cross-protocol reconstruction trajectory is regenerated with the updated dynamic delay and the temperature deviation trajectory.

5. The method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 2, characterized in that, The limitation of the source sampling time window includes: for adjacent refresh events within the same source protocol, comparing the order of joint changes in voltage, current and temperature measurements before and after refresh, merging the segment where only message transmission occurs but the measurement value is not synchronously updated into the previous holding interval, and determining the segment where at least one electrical measurement value shows a new change and the corresponding sampling identifier advances as the sampling boundary; Cross constraints are established for the sampling boundaries associated with the same current excitation event in different source protocols. The corresponding time windows of each sampling boundary are compressed according to their order. The sampling time windows of multiple source ends associated with the same physical event are arranged in a unified time order and do not overlap in the opposite direction.

6. The method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 3, characterized in that, The generation of the time offset sequence includes: continuously selecting multiple current excitation events along the unified physical time axis, forming candidate offset nodes from the candidate voltage response positions corresponding to each current excitation event, and limiting the time offset change direction of adjacent candidate offset nodes by the message update order of the source protocol. For locations where there is no effective voltage response between adjacent current excitation events, segmented time delay trajectories are constructed based on the candidate offset nodes before and after and their corresponding source sampling time windows. Adjacent segmented time delay trajectories are connected only within the range allowed by the source sampling time window to form a dynamic time delay sequence that changes continuously with the protocol data stream.

7. The method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 6, characterized in that, The screening of the candidate correspondence also includes: performing forward matching from the current excitation position to the voltage response position and reverse matching from the voltage response position back to the current excitation position for the same current excitation event, and retaining only the candidate correspondences where the forward matching and the reverse matching point to the same source sampling time window; When multiple voltage responses simultaneously satisfy bidirectional matching, the candidate range is limited according to the established time offset continuity relationship of the adjacent current excitation events, and voltage responses that cannot maintain the continuity of offset order are retained as undetermined corresponding items, and undetermined corresponding items are not written into the time offset sequence.

8. The method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 4, characterized in that, The establishment of the temperature-related residuals includes: associating the stable measurement segments between adjacent current excitation events with the corresponding temperature ranges according to the unified physical time axis, and extracting the residual components that recur in different temperature ranges and do not change synchronously with the protocol time base correction. Based on the continuous migration sequence of the temperature measurement data, the residual components of adjacent temperature ranges are connected to form a temperature deviation trajectory that is continuously updated with temperature changes. When a temperature range changes, the temperature deviation trajectory is updated only using the residual components of the stable measurement segment that exist before and after the change. Temperature ranges lacking the stable measurement segment are marked as intervals to be supplemented, and the corresponding deviation status is maintained without direct replacement across intervals.

9. A method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 6, characterized in that, After the dynamic delay sequence is formed, it is reverse-mapped to the refresh events of each source protocol to redetermine the corrected position of each refresh event on the unified physical time axis. Based on the corrected refresh event position, the corresponding source sampling time window is recompressed, and candidate offset nodes are generated again with the recompressed source sampling time window. The refresh event correction position and candidate offset node generated in two consecutive rounds are continuously compared. Only the correction results that simultaneously satisfy the source protocol update order, current excitation event order and the source end sampling time window inclusion relationship are retained. The correction results are written back to the dynamic delay sequence to form a closed-loop iterative constraint across protocol time bases.

10. A method for accurate full-temperature sensing of lithium batteries based on an intelligent sensing system according to claim 8, characterized in that, The update of the temperature deviation trajectory also includes: for each temperature range with the stable measurement segment, extracting the temperature-related residual difference before and after the protocol time base residual update, and verifying the consistency of the difference with the residual migration direction of adjacent temperature ranges; Only the residual components that continue to migrate after the protocol time base correction and repeat in adjacent stable measurement segments are written into the temperature deviation trajectory, while the residual components that disappear synchronously with the dynamic time delay are transferred back to the protocol time base residual. After reclassification, the voltage, current, and temperature correction data, as well as the measurement uncertainty residuals, on the unified physical time axis are recalculated according to the updated temperature deviation trajectory and dynamic delay.

Images