Lithium battery state data calculation processing system driven by ac-dc power supply

Abstract

This invention relates to the field of computer data processing technology, specifically a lithium battery state data calculation and processing system driven by AC / DC power supply. The system includes a data access processing link, a power supply condition fingerprint generation link, a credibility allocation link, a state data reconstruction link, and a state calculation cache link. It receives voltage, current, temperature, charge / discharge flags, timestamps, and power supply mode information of the lithium battery during AC / DC power supply operation. Based on voltage differential, current slope, ripple period, temperature hysteresis, and sampling interval within adjacent time windows, it forms a condition fingerprint and configures data credibility weights accordingly. Then, it combines these fingerprints with constraints such as charge conservation, charge / discharge direction consistency, temperature rise hysteresis, and voltage recovery monotonicity to generate a reconstructed state data sequence. This system can distinguish between AC / DC switching disturbances and changes in the lithium battery's intrinsic state, providing a more stable data foundation for subsequent calculations of state of charge, state of health, and state of available power.

Description

Technical Field

[0001] This invention relates to the field of computer data processing technology, specifically to a lithium battery state data calculation and processing system driven by AC / DC power supply. Background Technology

[0002] Lithium-ion battery state data processing systems driven by AC / DC power supplies typically receive voltage, current, temperature, charge / discharge flags, timestamps, and power supply mode information during lithium-ion battery operation. These data are then preprocessed, state estimated, and buffered for output. Existing systems generally employ a unified data access link, merging state data generated during AC power supply, DC power supply, and AC / DC switching into continuous time-series data. This data is then processed using methods such as sliding windowing, mean filtering, median filtering, outlier removal, interpolation, and model estimation to generate a data sequence suitable for subsequent calculations. Such solutions assume that data disturbances of the same lithium-ion battery under different power supply modes have similar statistical characteristics and typically do not separately identify the impact of power supply mode changes on voltage derivatives, current slope, temperature hysteresis, and sampling interval stability.

[0003] In existing lithium battery state-of-the-art (SOA) calculations, results such as state of charge (SOC), state of health (SOC), and state of available power (SOP) are mostly obtained by directly inputting raw collected data or data after general filtering into the SOA estimation model. Conventional solutions typically smooth voltage, current, and temperature data within a fixed-length window to suppress sampling noise, and delete or replace data exceeding preset amplitude or slope ranges as outliers. For systems with known power supply modes, AC and DC power supply segments are labeled separately, and different calculation models or compensation parameters are called according to the segmented results. While this approach can handle some random noise, its processing logic mainly relies on numerical deviations and fixed segmentation, failing to create a condition fingerprint for AC / DC switching disturbances and failing to transfer the reliability of data from different windows to subsequent SOA calculations.

[0004] Under AC / DC power supply driving conditions, the data change mechanisms in the AC stable segment, DC stable segment, AC / DC switching segment, and recovery segment are not consistent. The AC driving process may generate periodic ripples and short-time residuals, while the DC driving process is more likely to reflect current steps, voltage plateaus, and recovery trends. The AC / DC switching process also introduces short-time slope abrupt changes, temperature response lag, and sampling interval offsets. If existing technologies use unified filtering, they will mix the transient disturbances generated by AC / DC switching with the actual state changes of the lithium battery in the same data sequence. If simple segmentation is made based solely on the power supply mode, it is difficult to handle the overlapping of disturbances and actual state changes in the recovery segment, resulting in unstable data boundaries on which the state calculation is based.

[0005] Therefore, the main technical problem of existing AC / DC power-driven lithium battery state data calculation and processing systems is that ripple, slope abrupt changes, recovery lag, and sampling offset caused by changes in AC / DC power supply mode can easily enter the lithium battery state calculation link and be mistakenly processed as changes in the battery's own state by existing unified filtering, anomaly removal, or simple segmentation methods. This results in the reconstructed state data sequence lacking credibility differentiation and physical consistency constraints that match the power supply conditions, leading to drift, jumps, or false alarms in subsequent calculations of state of charge, state of health, and available power state before and after AC / DC switching. Summary of the Invention

[0006] The purpose of this invention is to provide a lithium battery state data calculation and processing system driven by AC / DC power supply, which can solve the problems in the background art mentioned above.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A lithium battery state data calculation and processing system driven by AC / DC power supply includes a data access and processing link, a power supply condition fingerprint generation link, a credibility allocation link, a state data reconstruction link, and a state calculation cache link. The data access and processing link receives information on the voltage, current, temperature, charge / discharge flags, timestamps, and power supply mode of the lithium battery during AC / DC power supply operation. The power supply condition fingerprint generation link forms a condition fingerprint based on the voltage differential, current slope, ripple period, temperature hysteresis and sampling interval within adjacent time windows. The credibility allocation link configures data credibility weights for the state credibility observation window, the power supply disturbance dominant window, and the mixed uncertainty window based on the operating condition fingerprint. The state data reconstruction link generates a reconstructed state data sequence based on the data credibility weight, charge conservation constraint, charging and discharging direction consistency constraint, temperature rise hysteresis constraint, and voltage recovery monotonicity constraint. The state calculation cache link is associated with storing the original data, reconstructed state data sequence, operating condition fingerprint, and data credibility weight.

[0008] Preferably, the power supply condition fingerprint generation link includes window segmentation processing, disturbance feature extraction processing, and fingerprint fusion processing; The window segmentation process forms overlapping time windows based on the continuity of timestamps and the changing positions of charge / discharge markers. The disturbance feature extraction process extracts the voltage differential sequence, current slope sequence, ripple period sequence, temperature hysteresis sequence, and sampling interval offset sequence from the overlapping time window, respectively. The fingerprint fusion process synchronizes and aligns the voltage differential sequence with the current slope sequence, and combines the synchronization and alignment result with the ripple period sequence, the temperature hysteresis sequence and the sampling interval offset sequence to form a power supply condition fingerprint.

[0009] Preferably, the credibility allocation link includes working condition segmentation processing, credibility initialization processing, and weight correction processing; The operating condition segmentation process divides the lithium battery status data into an AC stable segment, a DC stable segment, an AC / DC switching segment, and a recovery segment based on the operating condition fingerprint. The reliability initialization process configures the state reliability observation window according to the AC stable segment and the DC stable segment; The power supply disturbance dominance window is configured according to the AC / DC switching section; Configure a hybrid uncertainty window based on the characteristic overlap relationship between the recovery segment and the adjacent stable segment; The weight correction process continuously corrects the data reliability weights based on the cumulative charge deviation and voltage recovery trend of adjacent windows.

[0010] Preferably, the state data reconstruction link includes stable segment preservation processing, disturbance segment replacement processing, hybrid segment constraint solving processing, and boundary splicing processing; The stable segment maintains the voltage, current, and temperature change trends within the reliable observation window of the retained state. The disturbance segment replacement process generates a short-term replacement sequence based on the state evolution relationship between adjacent stable segments before and after the AC / DC switching segment; The hybrid segment constraint solution process generates candidate sequences under the constraints of charge conservation, charging and discharging direction consistency, temperature rise hysteresis, and voltage recovery monotonicity. The boundary splicing process forms a reconstructed state data sequence based on the slope consistency of the overlapping window regions.

[0011] Preferably, the window segmentation process includes a two-layer segmentation of a coarse-grained operating condition window and a fine-grained disturbance window; The coarse-grained operating condition window uses power supply mode information, the change position of charge / discharge indicator, and the continuity of sampling interval as boundaries; The fine-grained disturbance window is embedded within the coarse-grained operating condition window and slides overlappingly along the time axis; The disturbance feature extraction process generates power supply mode background features in the coarse-grained operating condition window; Local mutation features are generated in the fine-grained perturbation window. The fingerprint fusion process uses the power supply mode background features as index constraints and the local mutation features as perturbation descriptors to form a hierarchical power supply condition fingerprint.

[0012] Preferably, the fingerprint fusion process includes phase alignment processing and hysteresis matching processing; The phase alignment process uses the direction change point of the current slope sequence as a reference to perform time offset correction on the voltage differential sequence and the ripple period sequence. The hysteresis matching process constructs a thermal response matching segment based on the delay relationship between the temperature hysteresis sequence and the current slope sequence. The fingerprint fusion process encodes the voltage differential sequence and ripple period sequence that have completed time offset correction, together with the thermal response matching segment, into AC / DC switching segment identification features, and forms a power supply condition fingerprint together with the AC stable segment identification features and the DC stable segment identification features.

[0013] Preferably, the weight correction process includes residual attribution processing and cross-window propagation processing; The residual attribution processing generates power supply disturbance residual labels and state change residual labels based on the temporal sequence relationship between voltage residual, current residual, temperature hysteresis residual and charge cumulative deviation within the same time window. The cross-window propagation process restricts the propagation of the power supply disturbance residual mark to the AC / DC switching section and its adjacent recovery section, and propagates the state change residual mark along a continuous window consistent with the charging and discharging direction. The credibility allocation link adjusts the data credibility weight of the corresponding window based on the residual label after propagation.

[0014] Preferably, the mixed segment constraint solving process includes candidate sequence generation, constraint conflict detection, and candidate sequence filtering. The candidate sequence generation process generates multiple candidate sequences based on the state slope of the preceding stable segment and the state slope of the following recovery segment of the AC / DC switching segment. The constraint conflict detection process detects the conflict positions of each candidate sequence under the constraints of charge conservation, charge / discharge direction consistency, temperature rise hysteresis, and voltage recovery monotonicity. The candidate sequence screening process removes candidate sequences whose conflict positions are concentrated at the boundary of the stable segment, and determines the reconstructed state data sequence based on the consistency of the boundary slope of the remaining candidate sequences.

[0015] Preferably, the layered power supply condition fingerprint includes a stable background layer, a transient disturbance layer, and a recovery delay layer; The stable background layer is generated from the power supply mode background features in the coarse-grained operating condition window, and the transient disturbance layer is generated from the local abrupt change features in the fine-grained disturbance window. The recovery delay layer is generated by the delay correspondence between the temperature hysteresis sequence, voltage recovery sequence, and current slope sequence after the AC / DC switching segment; The power supply condition fingerprint generation link uses the stable background layer as the basis for window type determination, the transient disturbance layer as the basis for disturbance boundary determination, and the recovery delay layer as the basis for recovery segment end position determination.

[0016] Preferably, the state data reconstruction link further includes reconstruction result write-back processing; The reconstruction result write-back process establishes an index relationship between the reconstructed state data sequence and the corresponding original data residual, candidate sequence conflict position, boundary slope consistency result and data confidence weight, and writes it into the state calculation cache link; The state calculation cache link generates searchable cache records based on timestamps, power supply condition fingerprints, and window types. The retrieveable cache records retain the reference relationships between the state-credible observation window, the power supply disturbance-dominant window, and the mixed uncertainty window.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Through the coordinated processing of data access processing links, power supply condition fingerprint generation links, credibility allocation links, state data reconstruction links, and state calculation cache links, the original state data during the AC / DC power supply drive process is first identified as time windows with different disturbance sources before entering the reconstruction calculation. The power supply condition fingerprint is formed by features such as voltage differential, current slope, ripple period, temperature hysteresis, and sampling interval, providing data processing basis for the AC stable segment, DC stable segment, AC / DC switching segment, and recovery segment. Data credibility weights differentiate the degree of influence of different windows on subsequent calculations. Constraints such as charge conservation, charging and discharging direction consistency, temperature rise hysteresis, and voltage recovery monotonicity ensure that the reconstructed state data sequence remains consistent with the operating law of the lithium battery. With the above technical means combined, the situation where AC / DC switching disturbances are mistakenly included in the calculation as battery state changes can be reduced, and the data stored in the state calculation cache has a more stable operating condition correspondence and temporal continuity.

[0018] 2. By employing hierarchical power supply condition fingerprinting, residual attribution, cross-window propagation, candidate sequence screening, and reconstruction result write-back, the state data processing process gains clearer data source records and intermediate calculation basis. The dual-layer segmentation of coarse-grained condition windows and fine-grained disturbance windows preserves both the background characteristics and local abrupt change characteristics of the power supply mode; phase alignment and hysteresis matching clarify the time correspondence between voltage, current, ripple, and temperature; residual attribution processing separately labels power supply disturbance residuals and state change residuals, preventing unconstrained diffusion of similar residuals within different windows; and reconstruction result write-back establishes an index relationship between the reconstructed state data sequence, original data residuals, conflict locations, boundary slope consistency results, and data credibility weights, enabling subsequent state estimation to call data records with condition labels and credibility information, facilitating the review of state drift, jumps, and false alarms. Attached Figure Description

[0019] Figure 1 The overall processing flowchart of the lithium battery state data calculation and processing system driven by AC / DC power supply; Figure 2 Flowchart for generating fingerprints for power supply conditions; Figure 3 Flowchart for data credibility weight allocation and residual attribution; Figure 4 Flowchart for state data reconstruction and cache write-back. Detailed Implementation

[0020] In one embodiment, reference Figure 1A lithium battery state data calculation and processing system driven by AC / DC power supply includes a data access and processing link, a power supply condition fingerprint generation link, a credibility allocation link, a state data reconstruction link, and a state calculation cache link. The data access and processing link receives voltage, current, temperature, charge / discharge flags, timestamps, and power supply mode information of the lithium battery during AC / DC power supply operation. The power supply condition fingerprint generation link generates a condition fingerprint based on voltage differential, current slope, ripple period, temperature hysteresis, and sampling interval within adjacent time windows. The credibility allocation link configures data credibility weights for the state credibility observation window, the power supply disturbance dominant window, and the mixed uncertainty window based on the condition fingerprint. The state data reconstruction link generates a reconstructed state data sequence based on the data credibility weights, charge conservation constraints, charge / discharge direction consistency constraints, temperature rise hysteresis constraints, and voltage recovery monotonicity constraints. The state calculation cache link associates and stores the original data, the reconstructed state data sequence, the condition fingerprint, and the data credibility. In specific implementation, the data access processing link uniformly converts records from different sampling sources into data frames arranged in ascending order of timestamps. Each data frame contains at least voltage, current, temperature, charging / discharging direction marker, power supply mode marker, and sampling time. The power supply condition fingerprint generation link does not directly use the power supply mode marker as the sole criterion for judgment, but instead uses multiple disturbance features within adjacent time windows to form a calculable condition description. The credibility allocation link classifies the original data into three processing objects based on the condition description: reliable observation, disturbance-dominated, and mixed uncertainty. The state data reconstruction link adopts different data retention and replacement strategies among the three processing objects and writes the reconstructed voltage, current, temperature, and corresponding window labels into the state calculation cache link. In this embodiment, the data disturbance caused by AC / DC switching is distinguished from its source at the state calculation entry point, avoiding the original data from directly entering the calculation process of state of charge, state of health, and state of available power without condition identification.

[0021] In this embodiment, the data access processing link organizes voltage, current, temperature, charge / discharge identifiers, timestamps, and power supply mode information under the same time reference. Specifically, the data access processing link constructs a sequential cache with timestamps as the primary index. When the timestamp difference between adjacent records is near the same sampling period, the adjacent records are grouped into a continuous sampling segment. When the timestamp difference between adjacent records jumps or the power supply mode information changes, the jump position is marked as a candidate operating condition boundary. Voltage, current, and temperature are not directly deleted during the access stage, but rather each record is marked with a sampling continuity mark, a power supply mode availability mark, and a charge / discharge direction mark. Missing power supply mode information is inferred from voltage ripple, DC platform changes, and current direction continuity. The inference result is used as a candidate mark in subsequent calculations without overwriting the original record. When reading the above sequential cache, the power supply condition fingerprint generation link can simultaneously obtain the original sampled value and additional marks related to sampling quality, so that subsequent window segmentation does not depend on a single field. This embodiment retains the original disturbance information and sampling boundary information during the AC / DC power supply driving process, providing the same data foundation for subsequent operating condition fingerprints, confidence weights, and state reconstruction.

[0022] In this embodiment, the power supply condition fingerprint is generated according to the following formula: ; in, This represents the power supply condition fingerprint for the k-th time window. This indicates the normalized variation intensity of the voltage differential sequence within the time window. This represents the intensity of the change in the current slope sequence after normalization. This represents the ripple period stability extracted from the voltage residual signal. This indicates the degree of lag between temperature changes and current changes. The square brackets indicate that the sampling interval offset is represented by a vector composed of features of different dimensions in a fixed order. When the voltage differential change intensity is 0.2, the current slope change intensity is 0.1, the ripple period stability is 0.8, the temperature hysteresis is 0.3, and the sampling interval offset is 0.1 within a certain AC stability window, the power supply condition fingerprint forms a stable AC feature with the ripple period as the main identification quantity. When the voltage differential change intensity is 0.9, the current slope change intensity is 0.7, the ripple period stability is 0.4, the temperature hysteresis is 0.6, and the sampling interval offset is 0.2 within a certain AC / DC switching window, the power supply condition fingerprint forms a switching feature with slope abrupt change and hysteresis response as the main identification quantities. In this embodiment, multiple AC / DC disturbances are unified into the same vector structure, which facilitates repeatable window determination in subsequent confidence allocation links.

[0023] Preferred, Reference Figure 2The power supply condition fingerprint generation link includes window segmentation, disturbance feature extraction, and fingerprint fusion. The window segmentation process forms overlapping time windows based on timestamp continuity and the changing positions of charge / discharge markers. The disturbance feature extraction process extracts voltage differential sequences, current slope sequences, ripple period sequences, temperature hysteresis sequences, and sampling interval offset sequences from the overlapping time windows. The fingerprint fusion process synchronizes and aligns the voltage differential sequences with the current slope sequences, and combines the synchronization and alignment results with the ripple period sequences, temperature hysteresis sequences, and sampling interval offset sequences to form the power supply condition fingerprint. In specific implementation, the window segmentation process uses a fixed length combined with event boundaries to form overlapping time windows, and within continuous sampling segments, the overlapping time windows are generated at predetermined intervals. A basic window is formed by span sliding, and boundary windows are set at the locations where the power supply mode mark changes, the charging / discharging mark flips, and the sampling interval offset. The boundary windows retain the overlapping area with the adjacent basic windows. The disturbance feature extraction process performs adjacent difference on the voltage sequence in each window to obtain the voltage differential sequence, performs adjacent difference on the current sequence and normalizes it according to the time interval to obtain the current slope sequence, performs periodic peak-valley interval statistics on the voltage residual to obtain the ripple period sequence, statistically analyzes the time difference between the peak value of temperature change and the peak value of current change to obtain the temperature lag sequence, and statistically analyzes the difference between the actual sampling interval and the median sampling interval within the segment to obtain the sampling interval offset sequence. In this embodiment, the overlapping time window retains the continuous information before and after the AC / DC switching boundary, so that the switching disturbance will not lose the boundary shape due to hard segmentation.

[0024] Furthermore, the window segmentation process includes a two-layer segmentation of a coarse-grained operating condition window and a fine-grained disturbance window. The coarse-grained operating condition window uses power supply mode information, the position of charge / discharge indicator changes, and the continuity of sampling intervals as boundaries. The fine-grained disturbance window is embedded within the coarse-grained operating condition window and overlaps and slides along the time axis. The disturbance feature extraction process generates power supply mode background features in the coarse-grained operating condition window and generates local mutation features in the fine-grained disturbance window. The fingerprint fusion process uses the power supply mode background features as index constraints and the local mutation features as disturbance descriptors to form a hierarchical power supply operating condition fingerprint. Specifically, the coarse-grained operating condition window is used to describe the AC stable segment, DC stable segment, AC / DC switching segment, and recovery segment. The large-scale state of the segment is described by a fine-grained perturbation window, which is used to describe local changes such as voltage differential spikes, current slope reversals, ripple period compression, and temperature response delay. The power supply mode background features are composed of the power supply mode proportion, charging and discharging direction stability, and sampling interval continuity in the coarse-grained operating condition window. The local mutation features are composed of the differential peak position, slope inflection point position, and residual peak and valley position in the fine-grained perturbation window. When forming the hierarchical power supply operating condition fingerprint, the fingerprint fusion process writes the coarse-grained operating condition window number into the local mutation feature index, so that each local mutation feature can be traced back to the corresponding power supply background. This embodiment distinguishes between stable power supply background and transient perturbation morphology, so that the relationship between local perturbations in the recovery segment and the stable segment background can be preserved.

[0025] Furthermore, the fingerprint fusion process includes phase alignment processing and hysteresis matching processing. The phase alignment processing uses the direction change point of the current slope sequence as a reference to perform time offset correction on the voltage differential sequence and ripple period sequence. The hysteresis matching processing constructs a thermal response matching segment based on the delay relationship between the temperature hysteresis sequence and the current slope sequence. The fingerprint fusion process encodes the time-off corrected voltage differential sequence, ripple period sequence, and the thermal response matching segment together as AC / DC switching segment identification features, and forms a power supply condition fingerprint together with the AC stable segment identification features and the DC stable segment identification features. In specific implementation, the phase alignment processing searches in the current slope sequence... The direction change point from positive to negative or from negative to positive is used as the alignment reference. Within the allowed window overlap range, the voltage differential sequence is translated and the distance between the change peaks is calculated. The smaller translation amount is used as the voltage differential time offset. The ripple periodic sequence is corrected by the offset of the periodic peak and valley center relative to the direction change point. The hysteresis matching process finds the temperature change segment in the temperature sequence that corresponds to the current change direction and records the delay of the temperature change segment's start point, peak point, and fall point relative to the current direction change point. In this embodiment, the electrical fast response and thermal hysteresis response are established on the time axis to avoid data with different response speeds being mistakenly merged into the same instantaneous state.

[0026] In this embodiment, reference Figure 3 The credibility allocation link includes operating condition segmentation processing, credibility initialization processing, and weight correction processing. The operating condition segmentation processing divides the lithium battery state data into an AC stable segment, a DC stable segment, an AC / DC switching segment, and a recovery segment based on the operating condition fingerprint. The credibility initialization processing configures a state credibility observation window based on the AC and DC stable segments, a power supply disturbance-dominated window based on the AC / DC switching segment, and a mixed uncertainty window based on the feature overlap relationship between the recovery segment and adjacent stable segments. The weight correction processing continuously corrects the data credibility weights based on the charge accumulation deviation and voltage recovery trend of adjacent windows. Specifically, the operating condition segmentation processing reads the power supply operating condition fingerprint. The system uses information on ripple period stability, slope abrupt change location, and recovery delay to classify AC power supply windows with stable ripple periods and no abrupt changes in current slope into the AC stable segment, DC power supply windows with weak ripple periods and continuous voltage plateaus into the DC stable segment, windows with synchronous abrupt changes in voltage differential and current slope into the AC-DC switching segment, and windows that still exhibit temperature hysteresis and voltage recovery but whose current slope tends to stabilize after the switching segment into the recovery segment. The reliability initialization process establishes different initial weight intervals for the stable segment, switching segment, and recovery segment. The weight correction process does not change the window category but corrects the degree of participation of the window in subsequent calculations. In this embodiment, the data differences caused by the power supply mode are converted into data reliability differences that can participate in the calculation.

[0027] In this embodiment, the data credibility weight is generated and corrected according to the following formula: ; in, This represents the data reliability weight for the k-th time window, with a value between 0 and 1. Represents a stable and consistent quantity. Indicates the amount of sudden change in operating conditions. Indicates the cumulative charge deviation. This indicates the deviation of the voltage recovery. to The calculation coefficients are fixed with the system calibration file. When the stability consistency is 0.8, the operating condition change is 0.1, the charge cumulative deviation is 0.1, and the voltage recovery deviation is 0.1, and all coefficients are configured according to the positive stability constraint, the data credibility weight tends to take a higher value and the corresponding window participates in the reconstruction as a state credibility observation window. When the operating condition change is 0.9, the charge cumulative deviation is 0.6, and the voltage recovery deviation is 0.5, the data credibility weight tends to take a lower value and the corresponding window participates in the reconstruction in the manner of power supply disturbance-dominated window. In this embodiment, the window category and the degree of data participation are connected through the same continuous calculation formula to avoid discrete jumps between the stable segment and the switching segment.

[0028] Furthermore, the weight correction process includes residual attribution processing and cross-window propagation processing. The residual attribution processing generates power supply disturbance residual markers and state change residual markers based on the temporal sequence of voltage residuals, current residuals, temperature hysteresis residuals, and charge accumulation deviations within the same time window. The cross-window propagation processing restricts the propagation of the power supply disturbance residual markers to the AC / DC switching section and its adjacent recovery section, and propagates the state change residual markers along a continuous window consistent with the charging and discharging direction. The reliability allocation link corrects the data reliability weights of the corresponding window based on the propagated residual markers. Specifically, the voltage residual is calculated by combining the original voltage with the predicted voltage of the adjacent stable section. The difference between the two is formed by the current residual, which is formed by the difference between the actual current slope and the smooth slope within the window; the temperature hysteresis residual is formed by the delay deviation of the temperature change segment relative to the current change segment; and the charge accumulation deviation is formed by the deviation between the current integral within the window and the state change estimate. When the voltage residual and current residual appear before the temperature hysteresis residual and are concentrated at the power supply mode switching boundary, a power supply disturbance residual mark is generated. When the charge accumulation deviation accumulates continuously along the same charging and discharging direction and has no fixed correspondence with the power supply mode switching boundary, a state change residual mark is generated. In this embodiment, the residual propagation range limits different error sources, so that the confidence weight correction has a basis for the operating condition boundary.

[0029] In this embodiment, reference Figure 4The state data reconstruction link includes stable segment preservation processing, disturbance segment replacement processing, hybrid segment constraint solving processing, and boundary splicing processing. The stable segment preservation processing retains the voltage, current, and temperature change trends within the reliable state observation window. The disturbance segment replacement processing generates a short-time replacement sequence based on the state evolution relationship between adjacent stable segments before and after the AC / DC switching segment. The hybrid segment constraint solving processing generates candidate sequences under constraints of charge conservation, charging / discharging direction consistency, temperature rise hysteresis, and voltage recovery monotonicity. The boundary splicing processing forms a reconstructed state data sequence based on the slope consistency of the overlapping window regions. Specifically, the stable segment preservation processing does not enforce strict changes on the original change trends within the reliable state observation window. The system employs a flattening mechanism, using interpolation of adjacent samples to fill isolated sampling gaps. The disturbance segment substitution process reads the slope of the stable segment before the AC / DC switching segment and the trend before the recovery segment reaches stability, generating short-time substitution sequences that conform to the boundaries at both ends within the switching segment. The hybrid segment constraint solution process uses samples in the recovery segment that simultaneously contain disturbance attenuation and real state changes as a candidate set. Constraint conflict detection removes candidate sequences that do not satisfy the relationships of charge, direction, thermal response, and voltage recovery. The boundary splicing process compares the slopes of voltage, current, and temperature in the overlapping window region, ensuring that adjacent reconstructed segments form a continuous time sequence. This embodiment uses different reconstruction methods for the stable segment, switching segment, and recovery segment to avoid weakening of real state changes due to uniform filtering.

[0030] In this embodiment, the hybrid segment constraint solution determines the reconstructed state data sequence according to the following formula: ; in, This represents the reconstructed state data sequence from time a to time b. Let represent the t-th reconstructed state vector to be determined. Let t represent the observation vector of the t-th original state. This represents the data reliability weight corresponding to the t-th time window. This represents the cost of the charge conservation constraint. This represents the cost of maintaining consistency in the charging and discharging directions. This indicates the cost of the temperature rise lag constraint. This represents the cost of restoring the monotonicity constraint of voltage recovery. , , , The weight coefficients represent the costs of each constraint. When the data confidence weight in a certain mixed window is 0.3, the constraint of the original observation vector on the reconstruction result is weakened, while the constraints of charge conservation and voltage recovery monotonicity on the reconstruction result are strengthened. When the data confidence weight in adjacent stable windows is 0.9, more of the original observation vector is preserved. In this embodiment, the original data confidence and battery operation constraints are merged into the same solution process by unifying the optimization objective.

[0031] Furthermore, the hybrid segment constraint solution process includes candidate sequence generation, constraint conflict detection, and candidate sequence filtering. The candidate sequence generation process generates multiple candidate sequences based on the state slope of the preceding stable segment and the following recovery segment of the AC / DC switching segment. The constraint conflict detection process detects the conflict positions of each candidate sequence under the constraints of charge conservation, charge / discharge direction consistency, temperature rise hysteresis, and voltage recovery monotonicity. The candidate sequence filtering process deletes candidate sequences with conflict positions concentrated at the stable segment boundary and determines the reconstructed state data sequence based on the boundary slope consistency of the remaining candidate sequences. In specific implementation, the candidate sequence generation process uses a cutting... The voltage slope, current slope, and temperature slope at the end of the previous stable segment are used as entry conditions, and the voltage slope, current slope, and temperature slope at the end of the subsequent recovery segment are used as exit conditions. Linear transition sequences, segmented slope transition sequences, and hysteresis recovery transition sequences are generated respectively. The constraint conflict detection process marks each candidate sequence point by point with the position where the charge accumulation is not closed, the position where the charging and discharging direction is inconsistent with the current sign, the position where the temperature change is earlier than the current change, and the position where the voltage recovery direction is reversed. The candidate sequence screening process prioritizes retaining candidate sequences with dispersed conflict positions and continuous boundary slopes. This embodiment enables the reconstruction of the recovery segment to have complete logic of candidate generation, constraint detection, and boundary screening.

[0032] In a preferred embodiment, candidate sequence screening is calculated according to the following formula: ; in, This represents the screening cost of the m-th candidate sequence. This indicates the degree of inconsistency between the slope of the candidate sequence and that of the adjacent window at the boundary. This represents the quantity of charge conservation conflict. Indicates the amount of conflict in the charging and discharging directions. This indicates the amount of temperature rise lag conflict. This indicates the voltage recovery monotonicity conflict quantity. to The screening cost coefficient is used to select candidate sequences with lower screening costs and whose conflict positions are not concentrated at the boundary of the stable segment as reconstruction sources. When the boundary slope inconsistency of the first candidate sequence is 0.4, the charge conservation conflict is 0.1, the charge-discharge direction conflict is 0, the temperature rise hysteresis conflict is 0.2, and the voltage recovery monotonicity conflict is 0.3, while the corresponding values ​​of the second candidate sequence are 0.1, 0.1, 0, 0.2, and 0.1, the second candidate sequence is selected as the priority reconstruction source. In this embodiment, the selection of candidate sequences is limited to the comparison of calculable conflict costs to avoid determining the reconstruction sequence based solely on the fitting result of a single boundary.

[0033] Furthermore, the layered power supply condition fingerprint includes a stable background layer, a transient disturbance layer, and a recovery delay layer. The stable background layer is generated from the power supply mode background features in the coarse-grained operating condition window. The transient disturbance layer is generated from the local abrupt change features in the fine-grained disturbance window. The recovery delay layer is generated from the delay correspondence between the temperature hysteresis sequence, voltage recovery sequence, and current slope sequence after the AC / DC switching segment. The power supply condition fingerprint generation link uses the stable background layer as the window type determination basis, the transient disturbance layer as the disturbance boundary determination basis, and the recovery delay layer as the recovery segment end position determination basis. In specific implementation, the stable background layer records... The coarse-grained operating condition window records the continuity of power supply mode information, the stability of charge and discharge indicators, and the uniformity of sampling intervals. The transient disturbance layer records the voltage differential peak, current slope peak, and the start and end positions of the ripple period compression segment within the fine-grained disturbance window. The recovery delay layer records the time point when the temperature lag sequence enters the smooth interval, the time point when the monotonic trend of the voltage recovery sequence ends, and the time point when the current slope sequence returns to the stable range. The window type determination, disturbance boundary determination, and recovery segment end position determination are referenced sequentially among the three levels without overlapping each other. This embodiment enables the data boundaries before, during, and after AC / DC switching to be determined by different level features.

[0034] Furthermore, the state data reconstruction link also includes a reconstruction result write-back process. This process establishes an index relationship between the reconstructed state data sequence and the corresponding original data residuals, candidate sequence conflict locations, boundary slope consistency results, and data reliability weights, and writes this index into the state calculation cache link. The state calculation cache link generates a searchable cache record based on timestamps, power supply condition fingerprints, and window types. This searchable cache record retains the reference relationships between the state reliability observation window, the power supply disturbance dominant window, and the mixed uncertainty window. Specifically, the reconstruction result write-back process generates a record identifier for each reconstructed state data sequence, recording... The system binds and stores the timestamp range of the identifier and the original data, the corresponding power supply condition fingerprint number, the window type, the original voltage residual, the current residual, the temperature hysteresis residual, the candidate sequence conflict position set, the boundary slope consistency result, and the data credibility weight. When the state calculation cache link receives subsequent requests for calculating the state of charge, health, or available power, it retrieves and reconstructs the state data sequence according to the timestamp range and synchronously returns the condition fingerprint and window type. If the calculation request needs to verify the source of a certain state transition, it traces back to the original data residual and the candidate sequence conflict position according to the record identifier. This embodiment ensures that the data used for state calculation and the data reconstruction basis remain searchable and related.

[0035] In this embodiment, the searchable cache records in the state calculation cache chain form the index key according to the following formula: ; in, This represents the index key of the nth cached record. This indicates the timestamp range covered by the record. Indicates the window type encoding. Indicates the source code of the power supply mode. This indicates the fingerprint corresponding to the power supply condition. The parentheses indicate that multiple index fields are arranged in a fixed order to form a composite key. When the timestamp interval of a cached record is a continuous segment before and after a certain AC / DC switch, the window type code corresponds to a mixed uncertain window, the power supply mode source code corresponds to the inferred power supply mode, and the data confidence weight is 0.4, the state calculation cache link will mark the record as a record that needs to be used together with the reconstruction basis when reading it in the future. When the window type code corresponds to the state confidence observation window and the data confidence weight is 0.9, the subsequent calculation can directly read the corresponding reconstructed state data sequence as a continuous observation source. In this embodiment, time, operating condition, window type and confidence are included in the index to avoid the state calculation cache only outputting data in time order and losing the semantics of the operating condition.

[0036] In a preferred embodiment, the voltage, current, and temperature change trends within the reliable state observation window are preserved through a stable segment retention process. This retention method does not simply copy all the original data; instead, it calculates local slope sign sequences and plateau continuation segments for continuous samples within the window. When the local slope sign sequence matches the charge / discharge indicator and the plateau continuation segment does not experience abrupt changes due to power supply mode switching, the original voltage, current, and temperature sequences are written into the reconstructed state data sequence. When isolated missing points or single-point sampling interval offsets exist within the window, the stable segment retention process uses adjacent samples with the same slope sign before and after the missing point for linear completion. The completed record is marked with a completion point and retains the original missing position. If continuous missing points cross power supply mode change positions, the corresponding interval does not enter the stable segment retention process but instead enters the hybrid segment constraint solution process. In this embodiment, the stable segment retention process uses a low-intervention approach for the reliable observation window and a constraint reconstruction approach for data crossing operating condition boundaries. This embodiment preserves the true battery state change trend while avoiding treating missing or abrupt changes at power supply mode boundaries as ordinary sampling gaps.

[0037] In a preferred embodiment, the disturbance segment replacement processing generates a short-term replacement sequence for the power supply disturbance-dominant window using the state evolution relationship of the preceding and following stable segments. Specifically, the state data reconstruction link reads the most recent reliable state observation window before the AC / DC switching segment as the forward reference window and the reliable state observation window after the AC / DC switching segment where the recovery delay layer determination has been completed as the backward reference window. The forward reference window provides the voltage slope, current slope, and temperature slope before entering the switching segment, and the backward reference window provides the voltage slope, current slope, and temperature slope after leaving the recovery segment. The disturbance segment replacement processing establishes a smooth transition segment between the two and applies a charging and discharging direction consistency constraint to the transition segment, so that the current sign of the replacement sequence matches the charging and discharging identifier, the voltage recovery direction matches the power supply mode relationship before and after the switching, and the temperature change does not precede the current change. In this embodiment, the short-term replacement sequence within the switching segment is jointly defined by adjacent stable segments, reducing boundary discontinuities caused by unilateral extrapolation.

[0038] In a preferred embodiment, the charge conservation constraint is calculated according to the following formula: ; in, This represents the cost of the charge conservation constraint. This represents the t-th reconstructed current value. This represents the t-th sampling interval. and These represent the state variables corresponding to the start and end points of the reconstructed interval, respectively. This represents the proportionality coefficient for converting state variable changes into charge changes. When the reconstructed current integral within the interval is 10 charge units and the charge change after state variable conversion is 9.8 charge units, the charge conservation constraint cost is 0.2 charge units, and the corresponding candidate sequence retains strong constraint consistency. When the difference between the two increases, the cost of the candidate sequence in solving the constraints of the mixed segment increases. This embodiment ensures that the reconstructed sequence maintains a quantitative correspondence with the current integral and state variable changes, reducing charge non-closure caused by excessive smoothing in the recovery segment.

[0039] In a preferred embodiment, the boundary splicing process performs slope consistency verification on the overlapping areas between adjacent reconstructed segments. Specifically, the state data reconstruction link calculates the local voltage slope, local current slope, and local temperature slope at the end of the previous segment and the beginning of the next segment in each overlapping area. The consistency of direction, magnitude proximity, and order of change of the three types of local slopes are used as the boundary slope consistency results. If the local voltage slopes are consistent in direction and the local current slope matches the charge / discharge indicator, and the local temperature slope appears after the local current slope, then adjacent segments are directly spliced. If the local voltage slopes are opposite in direction and the conflict positions are concentrated at the stable segment boundary, then the corresponding candidate sequence is deleted. If the conflict positions are concentrated within the power supply disturbance dominant window, then the corresponding segment re-enters the disturbance segment replacement process. In this embodiment, the boundary splicing process does not use the equality of single-point values ​​as the splicing condition, but rather uses the multivariate slope relationship and the operating condition boundary position to jointly limit the splicing relationship. This embodiment ensures that the reconstructed state data sequence maintains a continuous direction of change before and after AC / DC switching.

[0040] In a preferred embodiment, the state calculation cache link performs hierarchical storage on the retrieveable cache records. Specifically, the original data layer stores the voltage, current, temperature, charge / discharge identifier, timestamp, and power supply mode information after access processing; the fingerprint layer stores the power supply condition fingerprint, stable background layer, transient disturbance layer, and recovery delay layer; the reconstruction layer stores the reconstructed state data sequence, candidate sequence source, and constraint conflict location; the credibility layer stores the data credibility weight, power supply disturbance residual marker, and state change residual marker; and the reference relationship layer stores the sequential relationship between the state credibility observation window, the power supply disturbance dominant window, and the mixed uncertainty window. When subsequent state calculations read a certain time interval, the state calculation cache link retrieves the reconstruction layer according to the timestamp interval and synchronously retrieves the fingerprint layer and credibility layer according to the reference relationship layer. If subsequent calculations detect a state jump, it can return to the original data layer to verify the corresponding original residual. This embodiment forms a closed data link between data access, condition identification, credibility allocation, reconstruction calculation, and cache output.

[0041] In a preferred embodiment, when power supply mode information is missing or power supply mode marking is discontinuous, the power supply condition fingerprint generation link uses data features to infer candidate power supply modes. Specifically, if the ripple period sequence has a stable peak-valley interval and the current slope does not show a continuous step within a certain time window, then the time window is marked as an AC candidate window. If the ripple period sequence weakens within a certain time window and the voltage plateau changes continuously with the current direction, then the time window is marked as a DC candidate window. If adjacent windows change from AC candidate windows to DC candidate windows or from DC candidate windows to AC candidate windows, accompanied by the synchronous appearance of voltage differential peak and current slope peak, then the intermediate window is marked as an AC / DC switching candidate window. The candidate power supply mode inference result is written to the power supply condition fingerprint in parallel with the original power supply mode information, without overwriting the original power supply mode field. This embodiment can still maintain the continuous operation of the condition fingerprint generation link when the power supply mode record is incomplete, and retain the inference source for subsequent verification.

[0042] In a preferred embodiment, the temperature rise hysteresis constraint defines the temperature sequence in the recovery segment through thermal response matching segments. Specifically, the state data reconstruction link reads the current direction change point, temperature change start point, temperature change peak point, and temperature fall point from the hysteresis matching process to construct a temperature response sequence constraint. The temperature change in the candidate sequence of the recovery segment must not occur earlier than the corresponding current change, the temperature peak point must not occur before the voltage recovery trend stabilizes, and the temperature fall point must not overlap with the current continuously rising segment. If the candidate sequence violates the above order, the constraint conflict detection process marks the corresponding position as the temperature rise hysteresis conflict position and passes the conflict position to the candidate sequence screening process. If the same candidate sequence only has isolated temperature noise and both voltage recovery and charge conservation satisfy the constraints, the temperature position is locally replaced by adjacent thermal response segments. This embodiment makes the temperature data reconstruction conform to the hysteresis response relationship after electrical excitation, reducing the possibility that temperature hysteresis is misunderstood as a sudden change in battery state.

[0043] In a preferred embodiment, the charging and discharging direction consistency constraint is jointly defined by the charging and discharging identifier and the reconstructed current symbol. Specifically, after the candidate sequence is generated, the state data reconstruction link reads the charging and discharging identifier point by point. If the reconstructed current symbol corresponding to the charging identifier is inconsistent with the preset direction, or the reconstructed current symbol corresponding to the discharging identifier is inconsistent with the preset direction, the corresponding point is marked as a direction conflict position. If the direction conflict positions are concentrated inside the AC / DC switching section, the candidate sequence enters the disturbance section replacement processing and is regenerated. If the direction conflict positions continue to appear along the same charging and discharging direction, the residual attribution processing marks them as state change residual candidates and assigns them to the confidence allocation link correction weight. If the direction conflict position appears at the same time as the sampling interval offset, the corresponding original data record is marked with a sampling offset and the data confidence weight is reduced. This embodiment maintains logical consistency between current reconstruction, charging and discharging direction and sampling continuity.

[0044] In a preferred embodiment, voltage recovery monotonicity constraints are used to process the recovery segment after AC / DC switching. Specifically, the starting point of the recovery segment is determined by the end position of the disturbance boundary in the transient disturbance layer, and the ending point of the recovery segment is jointly determined by the voltage recovery sequence, temperature hysteresis sequence, and current slope sequence in the recovery delay layer. Within the recovery segment, the state data reconstruction link determines the voltage recovery direction according to the power supply mode switching direction and the charging / discharging identifier. If a candidate sequence shows a continuous reversal segment opposite to the recovery direction, the constraint conflict detection process marks the segment as a voltage recovery monotonicity conflict position. If the reversal segment coincides with the peak and valley positions of the ripple periodic sequence, the reversal segment is treated as a power supply disturbance residual. If the reversal segment does not coincide with the ripple periodic sequence and continues to span multiple fine-grained disturbance windows, the reversal segment is treated as a state change residual. In this embodiment, the source of voltage recovery anomalies is determined by the recovery direction, ripple position, and duration.

[0045] In one embodiment, the AC / DC power-driven lithium battery state data calculation and processing system forms a closed-loop processing flow consisting of access, fingerprinting, credibility, reconstruction, and caching during operation. The data access processing link retains the original state data and sampling tags. The power supply condition fingerprint generation link generates a hierarchical power supply condition fingerprint through a coarse-grained condition window and a fine-grained disturbance window. The credibility allocation link generates data credibility weights based on window type, residual attribution, and cross-window propagation. The state data reconstruction link generates a reconstructed state data sequence based on constraints such as weights and charge conservation, charging and discharging direction consistency, temperature rise hysteresis, and voltage recovery monotonicity. The state calculation caching link associates the reconstructed data, original residuals, candidate sequence conflict positions, boundary slope consistency results, condition fingerprints, and credibility weights into a searchable cache record. Subsequent calculations of state of charge, health, and available power do not directly face unprocessed AC / DC switching disturbances, but instead read data records with condition sources, credibility, and reconstruction basis. This embodiment enables the changes in AC / DC power supply mode and the changes in the state of the lithium battery itself to be expressed and constrained in the calculation link in a hierarchical manner.

Claims

1. A lithium battery state data calculation and processing system driven by AC / DC power supply, characterized in that, This includes data access and processing links, power supply condition fingerprint generation links, trust allocation links, state data reconstruction links, and state calculation cache links; The data access and processing link receives information on the voltage, current, temperature, charge / discharge flags, timestamps, and power supply mode of the lithium battery during AC / DC power supply operation. The power supply condition fingerprint generation link forms a condition fingerprint based on the voltage differential, current slope, ripple period, temperature hysteresis and sampling interval within adjacent time windows. The credibility allocation link configures data credibility weights for the state credibility observation window, the power supply disturbance dominant window, and the mixed uncertainty window based on the operating condition fingerprint. The state data reconstruction link generates a reconstructed state data sequence based on the data credibility weight, charge conservation constraint, charging and discharging direction consistency constraint, temperature rise hysteresis constraint, and voltage recovery monotonicity constraint. The state calculation cache link is associated with storing the original data, reconstructed state data sequence, operating condition fingerprint, and data credibility weight.

2. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 1, characterized in that, The power supply condition fingerprint generation link includes window segmentation processing, disturbance feature extraction processing, and fingerprint fusion processing. The window segmentation process forms overlapping time windows based on the continuity of timestamps and the changing positions of charge / discharge markers. The disturbance feature extraction process extracts the voltage differential sequence, current slope sequence, ripple period sequence, temperature hysteresis sequence, and sampling interval offset sequence from the overlapping time window, respectively. The fingerprint fusion process synchronizes and aligns the voltage differential sequence with the current slope sequence, and combines the synchronization and alignment result with the ripple period sequence, the temperature hysteresis sequence and the sampling interval offset sequence to form a power supply condition fingerprint.

3. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 1, characterized in that, The credibility allocation link includes working condition segmentation processing, credibility initialization processing, and weight correction processing; The operating condition segmentation process divides the lithium battery status data into an AC stable segment, a DC stable segment, an AC / DC switching segment, and a recovery segment based on the operating condition fingerprint. The reliability initialization process configures the state reliability observation window according to the AC stable segment and the DC stable segment; The power supply disturbance dominance window is configured according to the AC / DC switching section; Configure a hybrid uncertainty window based on the characteristic overlap relationship between the recovery segment and the adjacent stable segment; The weight correction process continuously corrects the data reliability weights based on the cumulative charge deviation and voltage recovery trend of adjacent windows.

4. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 1, characterized in that, The state data reconstruction link includes stable segment preservation processing, disturbance segment replacement processing, hybrid segment constraint solving processing, and boundary splicing processing. The stable segment maintains the voltage, current, and temperature change trends within the reliable observation window of the retained state. The disturbance segment replacement process generates a short-term replacement sequence based on the state evolution relationship between adjacent stable segments before and after the AC / DC switching segment; The hybrid segment constraint solution process generates candidate sequences under the constraints of charge conservation, charging and discharging direction consistency, temperature rise hysteresis, and voltage recovery monotonicity. The boundary splicing process forms a reconstructed state data sequence based on the slope consistency of the overlapping window regions.

5. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 2, characterized in that, The window segmentation process includes a two-layer segmentation of coarse-grained operating condition windows and fine-grained disturbance windows; The coarse-grained operating condition window uses power supply mode information, the change position of charge / discharge indicator, and the continuity of sampling interval as boundaries; The fine-grained disturbance window is embedded within the coarse-grained operating condition window and slides overlappingly along the time axis; The disturbance feature extraction process generates power supply mode background features in the coarse-grained operating condition window; Local mutation features are generated in the fine-grained perturbation window. The fingerprint fusion process uses the power supply mode background features as index constraints and the local mutation features as perturbation descriptors to form a hierarchical power supply condition fingerprint.

6. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 2, characterized in that, The fingerprint fusion process includes phase alignment processing and hysteresis matching processing; The phase alignment process uses the direction change point of the current slope sequence as a reference to perform time offset correction on the voltage differential sequence and the ripple period sequence. The hysteresis matching process constructs a thermal response matching segment based on the delay relationship between the temperature hysteresis sequence and the current slope sequence. The fingerprint fusion process encodes the voltage differential sequence and ripple period sequence that have completed time offset correction, together with the thermal response matching segment, into AC / DC switching segment identification features, and forms a power supply condition fingerprint together with the AC stable segment identification features and the DC stable segment identification features.

7. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 3, characterized in that, The weight correction process includes residual attribution processing and cross-window propagation processing; The residual attribution processing generates power supply disturbance residual labels and state change residual labels based on the temporal sequence relationship between voltage residual, current residual, temperature hysteresis residual and charge cumulative deviation within the same time window. The cross-window propagation process restricts the propagation of the power supply disturbance residual mark to the AC / DC switching section and its adjacent recovery section, and propagates the state change residual mark along a continuous window consistent with the charging and discharging direction. The credibility allocation link adjusts the data credibility weight of the corresponding window based on the residual label after propagation.

8. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 4, characterized in that, The mixed segment constraint solving process includes candidate sequence generation, constraint conflict detection, and candidate sequence filtering. The candidate sequence generation process generates multiple candidate sequences based on the state slope of the preceding stable segment and the state slope of the following recovery segment of the AC / DC switching segment. The constraint conflict detection process detects the conflict positions of each candidate sequence under the constraints of charge conservation, charge / discharge direction consistency, temperature rise hysteresis, and voltage recovery monotonicity. The candidate sequence screening process removes candidate sequences whose conflict positions are concentrated at the boundary of the stable segment, and determines the reconstructed state data sequence based on the consistency of the boundary slope of the remaining candidate sequences.

9. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 5, characterized in that, The layered power supply condition fingerprint includes a stable background layer, a transient disturbance layer, and a recovery delay layer; The stable background layer is generated from the power supply mode background features in the coarse-grained operating condition window, and the transient disturbance layer is generated from the local abrupt change features in the fine-grained disturbance window. The recovery delay layer is generated by the delay correspondence between the temperature hysteresis sequence, voltage recovery sequence, and current slope sequence after the AC / DC switching segment; The power supply condition fingerprint generation link uses the stable background layer as the basis for window type determination, the transient disturbance layer as the basis for disturbance boundary determination, and the recovery delay layer as the basis for recovery segment end position determination.

10. The lithium battery state data calculation and processing system driven by AC / DC power supply according to claim 8, characterized in that, The state data reconstruction link also includes reconstruction result write-back processing; The reconstruction result write-back process establishes an index relationship between the reconstructed state data sequence and the corresponding original data residual, candidate sequence conflict position, boundary slope consistency result and data confidence weight, and writes it into the state calculation cache link; The state calculation cache link generates searchable cache records based on timestamps, power supply condition fingerprints, and window types. The retrieveable cache records retain the reference relationships between the state-credible observation window, the power supply disturbance-dominant window, and the mixed uncertainty window.

Images