Automated data acquisition and report generation system for battery test laboratories

Abstract

The application discloses a kind of battery test laboratory automation data acquisition and report generation system, it is related to battery test technical field;For solving the problem that data acquisition and compression strategy cannot perceive the change of data value in test process in prior art;The system acquires voltage, current, temperature parameter sequence by data acquisition module;Fluctuation intensity analysis module calculates window fluctuation intensity;State weight acquisition module queries the preset state weight corresponding to the current test state;Importance index fusion module carries out nonlinear coupling to window fluctuation intensity and state weight;Parameter feedback adjustment module dynamically adjusts the debouncing parameter for state transition confirmation in finite state machine according to comprehensive importance index and updates state machine;The application realizes data value perception, adaptive matching of compression strategy and dynamic optimization of test process control parameter, while guaranteeing the integrity of key information, significantly reduces storage pressure.

Description

Technical Field

[0001] This invention belongs to the field of battery testing technology, specifically an automated data acquisition and report generation system for battery testing laboratories. Background Technology

[0002] In the field of battery research and development and manufacturing, laboratory testing is the core link to verify battery performance, evaluate cycle life and ensure safety. With the rapid development of new energy vehicles and energy storage industry, battery testing cycles are getting longer and the number of samples is increasing. The resulting massive amount of test data puts higher demands on the storage efficiency and data processing capabilities of the acquisition system. Automated data acquisition and report generation systems have become an indispensable infrastructure for modern battery testing laboratories.

[0003] Existing technologies typically use a fixed sampling frequency to periodically collect parameters such as voltage, current, and temperature when processing battery test data, and apply a uniform compression algorithm in the data storage stage to reduce storage usage. At the same time, the test process is controlled by a preset finite state machine. The state machine switches between test stages according to fixed transition conditions and debouncing time based on the comparison results between the measured values ​​of voltage and current and the set values. This linearly decoupled acquisition and control architecture can meet the basic requirements in conventional test scenarios.

[0004] However, various complex operating conditions exist in actual testing. When the battery enters the critical region of constant current to constant voltage, the rate of change of voltage and current changes drastically, and the information density of the data increases sharply. During long periods of rest, the parameters change extremely slowly, and the data exhibits high redundancy. Existing technologies employ a uniform acquisition and compression strategy, resulting in high-value critical transient data being treated the same as low-value stable data. High-value information is easily smoothed or buried by compression algorithms. More seriously, as the battery ages or the ambient temperature changes, dynamic characteristics such as voltage rebound rate will shift. The fixed de-jittering time on which the state machine relies cannot be adaptively adjusted, and frequent state oscillations are prone to occur near the critical state. This not only generates a large number of invalid state switching records but also further contaminates the original data.

[0005] Therefore, how to ensure storage efficiency while enabling data acquisition and compression strategies to perceive changes in data value during testing, and how to make the behavior of the state machine adaptively match the dynamic response characteristics of the battery, has become a core problem that urgently needs to be solved in the field of battery testing technology. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an automated data acquisition and report generation system for battery testing laboratories.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows: An automated data acquisition and report generation system for battery testing laboratories, including: Data acquisition module: Acquires a sequence of battery state parameters continuously collected during the battery test, including voltage, current and temperature; Fluctuation intensity analysis module: Calculates the window fluctuation intensity to characterize the severity of data fluctuations based on the battery state parameter sequence; State weight acquisition module: Based on the current test state of the finite state machine in the test process, query the preset state weight corresponding to the test state; Importance index fusion module: Nonlinearly couples the window fluctuation intensity with the state weight to generate a comprehensive importance index for measuring the importance of the current data; Adaptive compression module: Based on the comprehensive importance index, determines a target compression level from at least two preset compression levels, and performs compression storage on the battery state parameter sequence according to the target compression level; Parameter feedback adjustment module: Based on the comprehensive importance index, dynamically adjust the debouncing parameters used for state transition confirmation in the finite state machine, and update the finite state machine with the adjusted debouncing parameters.

[0008] The present invention has the following beneficial effects: 1. By nonlinearly coupling window fluctuation intensity and state weights, an objective quantitative perception of data value is achieved. This solution breaks through the limitations of traditional acquisition systems that only focus on the absolute value or simple rate of change of data. It introduces window fluctuation intensity to characterize the activity of data on a micro time scale, and at the same time, combines the preset state weights of the test process to give macro semantic value. A comprehensive importance index is generated through multiplicative nonlinear fusion. This index can accurately identify high-value transient events such as the tiny transition of the battery in the constant current to constant pressure critical zone and the puncture pulses caused by gas evolution. At the same time, it marks the stable data in the long-term static stage as low-value fragments, fundamentally solving the problem of high-value information being submerged by massive redundant data, and enabling the system to have a "data value judgment ability" comparable to that of test engineers. 2. A tiered compression strategy based on a comprehensive importance index significantly reduces storage pressure while ensuring zero loss of critical information. By setting a first and a second preset threshold, data windows are divided into three value levels: high, medium, and low, and differentiated compression is performed: high-value windows use lossless compression to retain all original details, medium-value windows use a rotating door compression algorithm to retain key morphological features and extract fluctuation pattern labels, and low-value windows only store statistical summaries. This mechanism allows storage resources to be dynamically allocated according to data value. For cyclic lifetime tests lasting months or even years, storage usage can be reduced by more than 60%, while ensuring a complete record of critical events such as thermal runaway precursors and internal resistance mutations, providing a complete data foundation for subsequent lifetime prediction and fault diagnosis. 3. By utilizing the comprehensive importance index to dynamically correct the state machine's debouncing parameters, adaptive optimization of the test process control is achieved. Traditional state machines use fixed debouncing parameters, which easily lead to frequent oscillations near the critical state when dynamic response shifts are caused by battery aging or environmental changes, generating a large amount of invalid data and polluting the original records. This solution uses first-order low-pass filtering on the comprehensive importance index and establishes an inverse proportional modulation rule to calculate new debouncing points in real time, making the state machine more sensitive to capture key turning points when data importance is high and more robust to suppress noise interference when data importance is low. This closed-loop feedback mechanism ensures that the behavior of the state machine always matches the actual dynamic characteristics of the battery, effectively avoiding misjudgments and data pollution caused by rigid control parameters. 4. Through multi-dimensional indexing and fluctuation pattern mining, a highly efficient retrieval and analysis capability oriented towards data value has been built. The storage and indexing module associates compressed data blocks with comprehensive importance index, compression level identifier, and start timestamp for storage, and establishes a dual index structure with timestamp and importance index as retrieval fields. This enables engineers to quickly locate key time periods according to importance thresholds and trace back abnormal events such as thermal runaway and gas evolution from massive test data in seconds. At the same time, for the data of the second compression level, by analyzing the variance of the key point time interval retained by the rotating door compression, "continuous fluctuation pattern" and "instant fluctuation pattern" are identified and marked, providing deeper feature clues for subsequent fault type identification. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a system block diagram of Embodiment 1 of the present invention; Figure 2 This is a flowchart of the method in Embodiment 2 of the present invention. Detailed Implementation

[0011] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] In the field of battery research and development and manufacturing, laboratory testing is a core component for verifying battery performance, evaluating cycle life, and ensuring safety. With the rapid development of new energy vehicles and the energy storage industry, battery testing cycles are becoming longer and the number of samples is increasing. This generates massive amounts of test data, placing higher demands on the storage efficiency and data processing capabilities of the acquisition system. Automated data acquisition and report generation systems have become an indispensable infrastructure in modern battery testing laboratories. However, the acquisition of high-quality test data essentially depends on the acquisition system's ability to accurately capture the dynamic response of the battery—that is, to completely record the changes in physical quantities such as voltage, current, and temperature over time through a fixed sampling strategy, thereby providing a reliable basis for subsequent performance analysis.

[0013] Existing technologies for processing battery test data typically employ a fixed sampling frequency to periodically collect parameters such as voltage, current, and temperature, and use a uniform compression algorithm in the data storage stage to reduce storage footprint. Simultaneously, the test process is controlled by a pre-defined finite state machine. Based on the comparison between measured values ​​of voltage and current and set values, the state machine switches between test phases according to fixed transition conditions and debouncing times. This linearly decoupled acquisition and control architecture can meet basic requirements in conventional testing scenarios. However, existing technologies lack the ability to perceive the dynamic changes in data value during testing, and also lack a verification mechanism for the degree of matching between state machine behavior and the actual response characteristics of the battery. This results in the inability to accurately identify high-value segments in the test data and the adaptive deviation of the state machine parameters.

[0014] Specifically, battery testing involves various complex conditions: when a battery enters the critical region of constant current to constant voltage transition, the rate of change of voltage and current changes drastically, leading to a sharp increase in data information density and revealing key mechanistic characteristics such as battery gas evolution and polarization. Conversely, during prolonged static periods, parameters change extremely slowly, resulting in highly redundant data. Existing technologies employ uniform acquisition and compression strategies, treating high-value transient data and low-value stable data equally, making high-value information easily smoothed or buried by compression algorithms. More seriously, as batteries age or ambient temperature changes, dynamic characteristics such as voltage rebound rate shift, and the fixed debouncing time relied upon by the state machine cannot adaptively adjust, easily generating frequent state oscillations near critical states. For example, in the later stages of battery aging, the rebound after the voltage reaches a set value slows down, and the fixed debouncing time may cause the state machine to repeatedly enter and exit near the constant current to constant voltage critical point, generating not only a large number of invalid state transition records but also further contaminating the original data, making subsequent analysis difficult to distinguish between genuine battery behavior and artifacts introduced by system control.

[0015] For example, in power battery cycle life testing, engineers often focus on the inflection point of battery capacity decay and the moment of sudden increase in internal resistance. However, after hundreds of cycles, the battery's dynamic response slows down, and the original fixed de-jitter parameters cause the state machine to frequently misjudge during the constant voltage phase. The system records a large amount of invalid data generated by state oscillations, while the data on the key inflection points that truly reflect capacity decay are submerged in a sea of ​​redundant information. Furthermore, when engineers attempt to retrospectively analyze the precursors of thermal runaway, they find that early voltage fluctuations are discarded as noise by a unified compression algorithm, making it impossible to pinpoint the true origin of the fault. Specifically, the system misjudges state oscillations as normal battery behavior and continues to record invalid data, or misclassifies early gas evolution fluctuations as measurement noise and discards them. As a result, the test report cannot accurately reflect the battery's true degradation trajectory, and the R&D decisions made by engineers based on erroneous data deviate from the correct direction.

[0016] If the above problems are not addressed, the testing system will continuously lose its ability to objectively assess the value of data. Specifically, the failure to effectively retain high-value transient data will prevent subsequent analysis from capturing key mechanistic characteristics of the battery, leading to distorted input data for the lifespan prediction model. Simultaneously, the lack of correction for rigid state machine parameters will cause a disconnect between the control logic and the actual response characteristics of the battery, resulting in a lack of stability in the testing process and ultimately causing the test report to lose its objective reflection of the battery's true performance. Therefore, inaccuracies in data acquisition and process control will systematically hinder engineers from accurately assessing battery status, impacting the achievement of battery R&D and quality testing goals.

[0017] Example 1: As Figure 1As shown, the automated data acquisition and report generation system for the battery testing laboratory includes: Data acquisition module: Acquires the sequence of battery state parameters continuously collected during the test, including voltage, current and temperature; In practical engineering implementation, the data acquisition module constructed in this embodiment is the data source of the entire system. Its core responsibility is to capture the instantaneous response of the battery under complex test conditions in a high-fidelity and highly synchronous manner. This module does not simply read sensor values, but is designed with an acquisition architecture that has raw data caching and real-time preprocessing capabilities, providing a reliable data foundation for subsequent fluctuation intensity analysis and adaptive compression.

[0018] Specifically, the data acquisition module is connected to the voltage, current, and temperature sensors of the battery under test. The voltage sensor uses a high-precision Hall effect element or an isolation amplifier, with a range set from 0 to 5V, covering the voltage range of common single cells and series battery packs. The current sensor is a zero-flux current sensor based on fluxgate technology, with a range configured according to the maximum current of the test channel, such as ±100A or ±500A. The temperature sensor is a surface-mount Pt100 platinum resistance thermometer, attached to the surface of the battery casing, with a measurement range of -20℃ to 80℃. The analog signals from all sensors are filtered for anti-aliasing before being sent to a high-precision synchronous sampling ADC converter to ensure that the three physical quantities are acquired at the same time, thereby ensuring that the phase relationship between voltage, current, and temperature is not distorted.

[0019] To ensure the capture of transient events during battery charging and discharging, this embodiment sets the base sampling rate of the data acquisition module to 10Hz. This means that 10 sampling points are generated per second for each physical quantity. The sampled data enters the module's internal dual-buffer structure as a continuous data stream. The dual-buffer refers to maintaining two independent circular buffers for each physical quantity: one for storing the original sampled values, and the other for storing the smoothed values ​​after moving average filtering. The original data buffer fully preserves the sensor's original output, including any white noise or power frequency interference, which may contain very early signs of faults in specific analysis scenarios; the smoothed data buffer is used to remove high-frequency noise and extract the main trend of signal variation.

[0020] In actual operation, whenever the ADC completes a conversion and generates a new sampling point, this point is first pushed into the corresponding raw data circular buffer. This buffer is 100 units long, precisely accommodating 10 seconds of historical raw data. Next, the module calls a moving average filter to process the new data. The filter window length is set to 5 sampling points, which is the arithmetic mean of the current sampling point and its four preceding historical sampling points. This process can be represented by the following formula: For current and The temperature smoothing process is exactly the same. The reason for choosing a 5-point moving average is that, at a sampling rate of 10Hz, this filter can effectively suppress 50Hz power frequency interference and its higher harmonics, while preserving the dynamic characteristics of the battery's electrochemical reaction, such as the inflection point when switching from constant current to constant voltage.

[0021] In practical engineering applications, the data acquisition module needs to have the ability to identify and tolerate abnormal data. Specifically, after each reading of the ADC conversion result, the module first performs a validity check: if the voltage, current, or temperature value exceeds the effective range of the corresponding sensor (e.g., voltage exceeds 0~5V, current exceeds ±100A, temperature exceeds -20~80℃), or no new data is received for three consecutive sampling cycles, the current sampling point is determined to be invalid data. For invalid data, the module uses the following strategy to handle it: If the anomaly is a single, sporadic point, linear interpolation should be used to fill the gap. The prerequisite is that both the preceding and following points are valid. If there are multiple consecutive anomalies or sensor disconnections, data acquisition for that channel will be paused, an alarm signal will be sent to the host computer, and the data for that period will be marked as "invalid" in the data block to prevent subsequent modules from analyzing based on erroneous data.

[0022] In addition, to prevent the moving average filter from failing due to sensor noise, the module will not output the smoothed value if there are fewer than 3 valid data points in the window when calculating the smoothed value. Instead, the original value will be used directly, and a warning log will be recorded.

[0023] After smoothing, the newly obtained smoothed values ​​are pushed into another smoothed data circular buffer of the same length (100). At this point, for each sampling moment, the system simultaneously holds both the original data sequence and the smoothed data sequence within the current time window. These two sequences will serve as input to the subsequent fluctuation intensity analysis module. It is worth noting that to ensure data time alignment, the two buffers use the same pointer indexing mechanism, ensuring a one-to-one correspondence between the original point and the smoothed point at any given time. Through this method, the data acquisition module not only completes the basic data collection function but also provides the system with data views exhibiting different noise characteristics through preprocessing. This allows subsequent fluctuation intensity calculations to obtain robust trend indicators based on the smoothed data and to backtrack to the original data when needed to analyze weak anomalous pulses, laying a solid data foundation for the intelligent decision-making of the entire system.

[0024] Fluctuation Intensity Analysis Module: Calculates the window fluctuation intensity to characterize the severity of data fluctuations based on the battery state parameter sequence; The calculation process for the intensity of window fluctuations specifically includes: The battery state parameter sequence is divided into voltage, current and temperature subsequences for the current window by using a fixed time window. Calculate the sum of the absolute values ​​of the differences between adjacent sampling points within each subsequence to obtain the cumulative change of each subsequence; The normalized fluctuation intensity of each subsequence is obtained by dividing the cumulative change of each subsequence by the product of the range of the corresponding sensor and the window duration. The maximum value among the normalized fluctuation intensities of the three subsequences is determined as the window fluctuation intensity of the current window.

[0025] After real-time acquisition and preprocessing by the data acquisition module, the resulting smoothed data sequence will serve as input to the fluctuation intensity analysis module. The core function of this module is to quantify abstract data fluctuations into a specific, comparable numerical indicator (window fluctuation intensity), thereby providing an objective basis for subsequent data value assessment.

[0026] The fluctuation intensity analysis module first needs to divide the continuous data stream into time windows. To achieve continuous coverage of historical data while processing real-time data, this embodiment uses a sliding window mechanism. The window length is set to a fixed value, meaning the number of sampling points in each window is consistent with the length of the circular buffer in the data acquisition module. Because the base sampling rate is Therefore, the time span corresponding to each window is Every second, after a new sampling point is preprocessed and stored in the smoothed data buffer, the system uses the current time as the right boundary of the window and truncates the latest 100 smoothed data points to form a complete current window. This means that every 0.1 seconds, the system generates a new current window that overlaps with the previous window by 99 points, ensuring the real-time nature and continuity of the analysis.

[0027] Before performing the fluctuation intensity calculation, the module first verifies the window length. and sampling interval The validity of. If If there are fewer than two points within the window, then the difference between adjacent points cannot be calculated, and the fluctuation intensity of the current window is directly set. And skip subsequent calculations; if If the default value of 0.1s is used, configuration exception warnings will be recorded.

[0028] For each newly generated current window, the fluctuation intensity analysis module extracts three independent subsequences: the voltage subsequence. Current subsequence and temperature subsequence , where index From the earliest point within the window Arranged to the latest point Next, the module will perform a preliminary calculation of the fluctuation intensity for each subsequence.

[0029] Taking a voltage subsequence as an example, the first step in calculating its fluctuation intensity is to calculate the sum of the absolute differences between all adjacent sampling points within the window, i.e., the cumulative change. The physical meaning of this step is to measure the total length of the voltage signal's trajectory within a 10-second window. Regardless of whether the voltage is rising or falling, as long as the changes are frequent or large, this value will increase accordingly. Its mathematical expression is: Similarly, the values ​​for current and temperature can be calculated separately. and ; However, , , These three values ​​have different physical units and dimensions, making direct comparison impossible. For example, voltage changes might be in the millivolt range, while current changes might be in the ampere range. To measure the fluctuations of these three channels on the same scale, normalization is necessary. This embodiment introduces the concept of normalized fluctuation intensity, the core idea of ​​which is to divide the cumulative change by the maximum theoretical change that the channel might produce within a window time.

[0030] Specifically, for the voltage channel, its maximum possible range of variation is determined by the sensor's range. In this embodiment, it is decided that... Window duration is And sampling interval Therefore, the number of changes within the window is In the most extreme case, the voltage jumps back and forth between the two ends of the range with maximum amplitude in each sampling interval. Theoretically, the maximum value is ( )× To calculate the proportion of the average rate of change per second to the range, the maximum theoretical change is multiplied by... Divide by This can be simplified, and the final normalized formula is: ;because =1 / ,and This formula essentially measures the percentage of the average change per second within the window relative to full scale. In this way, It is mapped to a dimensionless interval [0, 1]. When the voltage is perfectly stationary, When the voltage fluctuates drastically at its maximum amplitude, Approaching 1; the exact same calculation method is used for both current and temperature channels. Among them, the range of the current sensor Based on the test channel configuration, this embodiment takes Temperature sensor range Pick .

[0031] To understand this more intuitively, consider a simple scenario: within a 10-second window, the voltage linearly rises from 3.000V to 3.005V, with a smooth and stable increase. At this point... Approximately 0.005V, substituting into the formula This is a very small value, indicating extremely low fluctuation. If the voltage experiences five rapid fluctuations of 0.1V within 10 seconds, then... It will be much greater than 0.005. This will also increase accordingly, thus accurately reflecting the existence of fluctuations.

[0032] The normalized fluctuation intensity of the three channels, voltage, current, and temperature, was calculated separately. , , Finally, the module faces a final problem: how to use a single metric to characterize the overall fluctuation level of the entire window. This embodiment adopts the strategy of taking the maximum value among the three. This is because a drastic fluctuation in any physical quantity may indicate a significant event or anomaly within the battery; if the average value is taken, the information about the drastic fluctuation might be overwhelmed by the stability of the other two channels. Therefore, the window fluctuation intensity... The final expression is Thus, the fluctuation intensity analysis module has successfully condensed a multivariate time series containing 100 points into a dimensionless numerical value ranging from 0 to 1. This value accurately depicts the intensity of change in the dimension of battery voltage, current, or temperature that experiences the most dramatic fluctuation within the current 10-second window. The values ​​will serve as key inputs, along with the state weights generated in subsequent steps, and will be fed into the importance index fusion module to generate a higher-level comprehensive importance index. Through this refined feature extraction, the system can accurately perceive the "active" parts of the data, laying the foundation for subsequent differentiated processing.

[0033] State weight acquisition module: Based on the current test state of the finite state machine in the test process, query the preset state weight corresponding to the test state; The test states include at least constant current charging state, constant voltage charging state, discharging state and resting state. Each state corresponds to a different preset state weight, and the preset state weights are preset based on test experience.

[0034] After the volatility analysis module quantifies the data volatility, the system needs to evaluate the potential value of the current data from another dimension (the macro-level stage of the test process). This function is implemented by the state weight acquisition module. The core task of this module is to transform the abstract operational stage into a concrete numerical weight, which relies on deep collaboration with the main control unit of the test system and pre-modeling of the test process.

[0035] The state weight acquisition module first needs to establish real-time communication with the finite state machine running on the system's main control unit. This finite state machine is the command center of the entire test process. Based on the preset test steps and the comparison results of the real-time collected voltage and current values ​​with the set values, it drives the test process to switch orderly between multiple predefined test states. In this embodiment, the test states managed by the finite state machine include at least four basic types: constant current charging state, constant voltage charging state, discharging state, and resting state. Each state corresponds to a specific set of control objectives and boundary conditions.

[0036] Specifically, the finite state machine maintains a current state register, the value of which reflects the current stage of the system. Determining state transitions relies on threshold monitoring of measured data. For example, when the system is in a constant current charging state, the main control unit continuously compares the battery terminal voltage. Compared with the preset constant pressure target value When detected ≥ And current When the downward trend begins, the state machine does not switch immediately, but instead starts a debouncing counter. This counter requires the above conditions to be met continuously. After sampling several points, it was confirmed that the voltage had indeed reached the inflection point, and the value in the current status register was updated from "constant current charging state" to "constant voltage charging state". The number of debouncing points here... It is not fixed, but dynamically provided by the subsequent parameter feedback adjustment module, which reflects the closed-loop coupling between the various modules of the system.

[0037] The state weight acquisition module reads the current state register of the finite state machine at a fixed frequency (e.g., once per second) to obtain the current state identifier. However, state identifiers alone are insufficient for direct computation; they must be mapped to numerical weights. Therefore, this embodiment pre-constructs a "test state-weight mapping table," which serves as the knowledge base for the state-weight acquisition module to function.

[0038] This mapping table is not arbitrarily set, but rather calibrated based on a deep understanding of battery testing processes. In constructing this table, the system referenced common battery testing standards and extensive historical testing experience. Its core principle is to assign higher weights to stages where battery electrochemical reactions are active, parameters change drastically, or state transitions are prone to occur; and lower weights to stages where parameters are stable and information redundancy is high. Based on this principle, the mapping relationship set in this embodiment is as follows: For constant current charging, this is the main stage of battery energy storage. Although it is generally stable, polarization may be present towards the end of the charging process, hence it is assigned a basic weight. ; During constant-voltage charging, the current gradually decreases, and the internal polarization of the battery reaches its peak. This is a critical period for analyzing the battery's acceptability and gas evolution risk, and therefore it is given high weight. ; During the discharge phase, the battery performs work, and key performance indicators such as voltage plateau and internal resistance changes are concentrated and weighted at this stage. 7; In the resting state, the battery is in an open circuit state, and the voltage recovers slowly. Although it contains polarization decay information, the rate of change is extremely slow, and the information density is low. Therefore, it is given low weight to save resources. 2.

[0039] To ensure the scientific validity and universality of the preset state weights, this embodiment uses the following method for weight calibration: Expert experience method: Invite experts in the field of battery testing to compare and score the importance of each testing stage in pairs, and use the analytic hierarchy process (AHP) to calculate the initial weight values.

[0040] Historical data statistical method: Retrospectively analyze the key events marked in the historical test data (such as the gas evolution initiation point and the precursor to thermal runaway) and calculate the frequency of each state before the occurrence of the key event. The higher the frequency, the greater the weight.

[0041] Iterative optimization method: In actual test runs, record the key time periods manually marked by users, statistically analyze the state distribution corresponding to these time periods, and continuously calibrate the preset weights using the Bayesian update method.

[0042] By combining the above methods, the final determined weight values ​​can better reflect the actual information value of each testing phase, and can be optimized and adjusted through offline calculation during system upgrades.

[0043] The above weight values ​​form a simple key-value pair storage structure, where the "key" is the state identifier and the "value" is the corresponding weight. The state weight acquisition module reads the current state. After that, Use this as an index to perform a lookup operation in the mapping table to quickly retrieve the corresponding... .

[0044] It is worth noting that, in addition to the four basic states mentioned above, there are also some brief transition phases during the test, such as the tens of seconds before the transition from constant current to constant voltage. Although the state machine still reports a constant current state before the formal transition, the voltage is already close to the set value at this time, and the rate of change begins to slow down. Its information value is far higher than that of the ordinary constant current phase. To characterize this phenomenon more precisely, this embodiment allows fine-tuning of the basic weights through external rules. For example, an additional rule can be configured: when the state machine is in a constant current charging state, and the measured voltage... When this happens, the system determines that it has entered the "critical transition zone" and retrieves the weight from the mapping table. It has been temporarily upgraded to version 1.0. This dynamic adjustment logic further improves the rationality of weight allocation.

[0045] It should be noted that the aforementioned temporary weight increase based on the critical transition zone is a fine-grained correction, applied only to the current window, and the corrected weight is still subject to the amplitude control of the importance index fusion module. This correction and the dynamic adjustment of the debouncing parameters in the parameter feedback adjustment module are two independent dimensions of optimization: the weight correction affects the static evaluation of data importance, while the debouncing parameter adjustment affects the dynamic response speed of the state machine to transition conditions. They are physically and logically mutually exclusive but can coexist, jointly improving the overall adaptability of the system. To avoid over-correction, the magnitude of the temporary weight increase should not exceed the preset maximum weight (e.g., 1.0), and only one correction rule is allowed to be applied within the same window.

[0046] Finally, the state weight acquisition module will use the values ​​obtained after querying (possibly with dynamic adjustments) As output, this value, along with the windowed wave intensity previously calculated by the wave intensity analysis module, is included. Together, these constitute two orthogonal dimensions of quantitative indicators, which are simultaneously fed into the importance index fusion module. Among them, It characterizes the activity level of the data at micro-timescales, while This characterizes the potential value of data in the macro-level testing process, and the combination of the two provides complete input information for the subsequent generation of a comprehensive importance index. Through this design, the system can deeply integrate pure physical quantity fluctuations with domain knowledge, laying the logical foundation for intelligent data management.

[0047] Importance index fusion module: Nonlinearly couples the window fluctuation intensity with the state weight to generate a comprehensive importance index for measuring the importance of the current data; The nonlinear coupling in the importance index fusion module is implemented according to the following formula: ,in, As a comprehensive importance index, For state weights, For window fluctuation intensity, The amplification factor is preset; and the overall importance index is limited to a preset value range.

[0048] The fluctuation intensity is output in the fluctuation intensity analysis module and the state weight acquisition module, respectively. and state weights Next, these two quantitative indicators from different dimensions need to be effectively integrated to form a unified benchmark that can comprehensively measure the importance of the current data. This integration task is undertaken by the importance index integration module.

[0049] The design philosophy of the importance index fusion module is based on a core understanding: the importance of data is not determined by a single factor, but rather by the combined effect of its inherent activity level and external process semantics. (Window fluctuation intensity) It reflects the dramatic changes in data at a microscopic timescale, capturing the instantaneous dynamics of electrochemical reactions within the battery; while state weights... This reflects the prior value of data in the macro-level testing process, representing the domain experts' prediction of the importance of each testing stage. Simply adding or averaging the two cannot reflect the difference in their physical meaning: the information value contained in a drastic fluctuation occurring in a low-weight static stage (e.g., an unexpected disturbance) is completely different from that in a high-weight constant-pressure stage (e.g., a gas evolution reaction).

[0050] Therefore, this embodiment designs a nonlinear coupling mechanism, the core of which is a fusion formula with an amplification effect. This module receives the current window's... and Then, the following calculations are performed first. ;in, This is the overall importance index of the output, which is a dimensionless numerical value used to measure the importance of the current data; It is a preset magnification factor, which is taken as a value based on experience in this embodiment. The structure of this formula contains a clear physical logic: state weights. This constitutes the "base" of importance, while the intensity of fluctuations... It then acts as an "amplifier," multiplying the base. When When the data is completely stationary, At this point, its importance is entirely determined by the test state itself; when At that time, importance will increase on the basis of the base number. The part, and The larger the value, the more significant the increment. In particular, when... When (data fluctuations reach their maximum), 3 It achieved a maximum amplification effect of three times.

[0051] This multiplicative rather than additive coupling method ensures that high volatility can be significantly amplified in critical states, while in low-weight states, even if some volatility exists, its absolute importance will not be too high, thus aligning with domain knowledge. For example, consider two scenarios: Scenario A is a constant voltage charging state ( And slight fluctuations occurred. ), calculated Scene B is a static state. And the same fluctuations occur ( ), calculated It is evident that the same fluctuations are rated as highly important in critical states, but remain low in static states, which aligns with the intuition of test engineers.

[0052] To prevent damage caused by sensor malfunctions or extreme operating conditions If the value is too large, it will interfere with the subsequent determination of the compression level. This module calculates the initial value... After calculating the value, a limiting operation is performed. Specifically, the system presets a numerical range, such as [0, 3]. If the calculated result exceeds the upper limit of this range, then... If it is below the lower limit, then let In practical applications, due to and ,exist hour The theoretical maximum value is 3, so the limiting operation is mainly to prevent overflow that may occur due to abnormal coefficients or future parameter adjustments.

[0053] After the threshold is set, this module will calculate the final overall importance index. As a real-time updated numerical output, this value encapsulates information from both microscopic fluctuations and macroscopic semantics, becoming the cornerstone of all subsequent differentiated processing decisions. The adaptive compression module will then... The parameter feedback adjustment module determines which compression level to use. As a feedback signal, the debouncing parameters of the state machine are dynamically optimized. In this way, the importance index fusion module plays a key role in information fusion and decision-making in the system, ensuring that the entire system can dynamically adjust its behavior based on the intrinsic value of the data.

[0054] Adaptive compression module: Based on the comprehensive importance index, it determines the target compression level from at least two preset compression levels, and performs compression storage on the battery state parameter sequence according to the target compression level; The specific process of determining the target compression level based on the comprehensive importance index in the adaptive compression module includes: When the comprehensive importance index is greater than or equal to the first preset threshold, the target compression level is determined as the first compression level, and lossless compression is performed on the battery state parameter sequence. When the comprehensive importance index is greater than or equal to the second preset threshold and less than the first preset threshold, the target compression level is determined to be the second compression level, and the rotating door compression algorithm is used to perform lossy compression on the battery state parameter sequence to retain the key points in the sequence. When the comprehensive importance index is less than the second preset threshold, the target compression level is determined to be the third compression level, and the statistical features of the battery state parameter sequence are extracted and stored. The statistical features include at least one of the following: mean, variance, maximum value, minimum value, starting value, ending value, and rate of change.

[0055] A comprehensive importance index is generated in the importance index fusion module. Then, the adaptive compression module starts up, and its core task is to... Depending on the numerical value, differentiated compression and storage strategies are applied to the battery state parameter sequence within the current window. This module is designed to "use resources wisely"—preserving as much original detail as possible for periods containing high-value information, while significantly compressing redundant information to save storage space, while ensuring that key features are not lost.

[0056] The adaptive compression module first retrieves the data corresponding to the current window from the double buffer of the data acquisition module. Specifically, it needs to acquire the original data sequence. and smooth data sequences ,in Simultaneously, this module receives the current window comprehensive importance index from the importance index fusion module. ;based on The module executes hierarchical compression decisions based on the given values.

[0057] This embodiment presets two importance thresholds for classifying compression levels: the first preset threshold... The second preset threshold These two thresholds divide the importance index into three intervals, corresponding to three different compression levels. The specific determination and execution process for the compression level is as follows: when When the current window is identified as a high-value data window, the system selects the first compression level. At this level, the primary goal of compression is to ensure zero information loss. Therefore, the module directly performs lossless compression on the original data sequence within the current window. Specifically, differential encoding can be performed on the data first to reduce the dynamic range of the values, followed by Huffman coding or directly calling a general compression algorithm library (such as zlib) for compression. The compressed data can be completely decompressed to recover the original sampled values ​​without introducing any distortion. This strategy ensures that key features such as minute transitions from constant current to constant pressure and puncture pulses caused by gas evolution are fully preserved. when When the current window is determined to be a medium-value data window, the system selects the second compression level. At this level, a balance needs to be struck between information fidelity and compression ratio. This embodiment uses the widely used rotating door compression algorithm in industry to perform lossy compression on smooth data sequences. The core idea of ​​rotating door compression is that, within a given compression tolerance... Within the range, approximate the original curve with a polyline, retaining only the key points that change the direction of the polyline; Before implementing revolving door compression, an independent compression tolerance needs to be set for each physical channel. In this embodiment, the compression tolerance of the voltage channel... Set to 0.1% of the sensor's range, i.e. Compression tolerance of the current channel Set to 0.5% of the range, that is Compression tolerance of the temperature channel The tolerance value is set to 0.1℃. These tolerance values ​​represent the maximum reconstruction error allowed by the compression algorithm and can be adjusted according to the requirements of the test accuracy.

[0058] Before performing the revolving door compression, the module checks the number of valid data points in the current window. If the number of valid points... If the compression fails, the point cannot be compressed and is directly retained as a keypoint, with the compression level marked as the second compression level but without a keypoint sequence; if If both points are retained as key points, no threshold judgment will be performed.

[0059] The specific execution process of revolving door compression is as follows: Starting from the first point of the window, this is taken as the first base point. The slope of the door is then considered. and the slope of the lower door Construct two virtual doors respectively, where For each subsequent new point, it is determined whether the point falls within the sector area defined by the current base point and the upper and lower doors. If it falls within the area, the next point is processed, and the upper and lower boundaries of the doors are dynamically adjusted based on the new point (actually, the maximum value of the slope of the upper and lower doors is updated). If a point falls outside the area, the previous point is identified as a key point and retained, and the upper and lower doors are reinitialized using the current new point as the new base point, and the above process is repeated. In this way, all data points within the window are compressed into a set of key point sequences, each key point containing its timestamp and value. For the voltage, current, and temperature channels, the above compression process needs to be performed independently.

[0060] When the target compression level is the second compression level, it also includes: Extract the timestamps of key points retained after compression; Based on the timestamps of key points, determine the time interval sequence between adjacent key points; Calculate the variance of the time interval sequence as a volatility density index; The fluctuation pattern of the current window data is determined based on the comparison between the fluctuation density index and the preset fluctuation threshold.

[0061] The process of determining the fluctuation pattern of the current window data specifically includes: If the variance is less than the first preset threshold, the fluctuation pattern is determined to be a continuous fluctuation pattern. If the variance is greater than the second preset threshold, the fluctuation mode is determined to be an instantaneous fluctuation mode. The fluctuation pattern is stored as an additional attribute associated with the compressed data block.

[0062] After completing the revolving door compression and obtaining the key points of each channel, this embodiment further performs secondary analysis on these key points to uncover richer fluctuation pattern information. Specifically, the module merges the timestamps of the key points retained from all channels into a single timestamp set, removes duplicates, and sorts them by time. Then, the time intervals between adjacent key points are calculated to obtain a time interval sequence. ,in This represents the total number of keypoints after deduplication. Next, the variance of this time interval sequence is calculated. : ;in This represents the mean of the time intervals. The variance reflects the uniformity of the keypoints' distribution along the time axis. A smaller variance indicates a more uniform distribution of keypoints, suggesting that data fluctuations may be a continuous, stable oscillation; a larger variance indicates that keypoints are concentrated in a few time periods, suggesting that data fluctuations may be caused by sudden spikes.

[0063] To quantify this fluctuation pattern into a storable tag, this module presets two thresholds: the first preset threshold... 2 Second preset threshold 2 The judgment rules are as follows: like If so, the fluctuation pattern of the current window data is determined to be a continuous fluctuation pattern; like If so, the fluctuation pattern is determined to be an instantaneous fluctuation pattern; If it falls between the two, it will not be specially marked, or it will be marked as a general fluctuation pattern.

[0064] when When the current window is deemed a low-value data window, the system selects the third compression level. At this level, maximum compression is achieved, extracting only the statistical features of the data within the window for storage, discarding all specific time-series points. The extracted statistical features include: the mean, variance, maximum, minimum, starting value, ending value, and rate of change for each channel. Taking the voltage channel as an example, its mean... ,variance Maximum value Minimum value Initial value ,Finish and rate of change All data were calculated and stored. The current and temperature channels were processed in the same way. These statistical characteristics are sufficient to reflect the basic shape and trend of the data within the window, while the amount of data is significantly reduced compared to the original sequence.

[0065] Finally, the adaptive compression module encapsulates the compression results into standard data blocks. Each data block contains the following fields: window start timestamp. Window duration (Fixed to 10 seconds), Overall Importance Index The data blocks are labeled with a compression level identifier (1, 2, or 3) and the compressed data content. For the second compression level, an additional fluctuation pattern label (such as "continuous fluctuation" or "instant fluctuation") is added to the data blocks. These complete data blocks are then passed to the storage and indexing module for persistent storage and the creation of a retrieval index. Through this sophisticated hierarchical compression strategy, the adaptive compression module significantly reduces the overall data storage pressure while ensuring that key information is not lost. Furthermore, the discovery of fluctuation patterns provides deeper clues for subsequent data analysis.

[0066] Parameter feedback adjustment module: Based on the comprehensive importance index, dynamically adjust the debouncing parameters used for state transition confirmation in the finite state machine, and update the finite state machine with the adjusted debouncing parameters.

[0067] The specific execution process of the parameter feedback adjustment module includes: A first-order low-pass filter is applied to the overall importance index to obtain a smoothed importance index; Based on the smoothed importance index, the new de-jitter points are calculated according to the following modulation rules: ;in, For the new anti-shake points, The preset minimum number of debouncing points, Based on the de-jitter points, This is the sensitivity coefficient. This is the smoothed importance index.

[0068] The importance index fusion module outputs a comprehensive importance index. Then, the parameter feedback adjustment module is activated, and its core task is to adjust the parameters. As a feedback signal reflecting the value of data and the complexity of operating conditions, the behavioral parameters of the finite state machine are dynamically optimized. This module is designed to break down the barriers between process control and data acquisition in traditional testing systems, enabling the state machine's response sensitivity to adaptively adjust according to the actual information content of the data. This allows for rapid response during critical events and enhanced anti-interference capabilities during stable phases.

[0069] The parameter feedback adjustment module is triggered once at the end of each data window processing cycle, that is, it performs an update every 10 seconds. Its input is the overall importance index of the current window. The output is the updated debouncing points. This value will be written into the internal register of the finite state machine for the confirmation process of all subsequent state transitions.

[0070] Finite state machines are the core unit of the control test process. Their state transitions depend on the confirmation that the transition conditions are met at multiple consecutive sampling points. The number of consecutive sampling points required for this confirmation is the number of de-jittering points. In traditional design, Typically, a fixed debouncing time is used, which cannot adapt to dynamic characteristic shifts caused by battery aging or changes in operating conditions. For example, in the later stages of battery aging, voltage rebound slows down, and a fixed debouncing time may cause the state machine to oscillate frequently near the critical point; while in the event of transient events such as internal micro-short circuits, an excessively long debouncing time will delay the response to the fault. This module is introduced precisely to resolve this contradiction.

[0071] To avoid The instantaneous jump causes drastic fluctuations in the debouncing parameters, resulting in instability in the state machine's behavior. The parameter feedback adjustment module first adjusts the overall importance index of the input. Smoothing is performed. Specifically, a first-order low-pass filter (exponentially weighted moving average) is used to smooth the image. The recursive formula for filtering is as follows: ;in, The current window's index. This is the original comprehensive importance index calculated for the current window. The smoothing importance index after filtering in the current window. These are the filter coefficients, and their value range is... In this embodiment, α = 0.3 is chosen, meaning that the new smoothing value mainly relies on historical information (weight 0.7) while retaining a certain proportion of current information (weight 0.3), thereby effectively suppressing the influence of sudden noise and making the adjustment of the de-jitter parameters smoother and more gradual. Initially, let... .

[0072] Importance index after smoothing Then, the module calculates a new number of dejitter points based on a preset modulation rule. The core idea of ​​this modulation rule is: the higher the data importance, the faster the state machine should respond, i.e., the fewer the number of dejitter points should be; the lower the data importance, the more conservative the state machine should be, i.e., the more dejitter points should be. To achieve this goal, this embodiment uses the following inverse proportional modulation formula. The parameters in the formula are defined as follows: The new number of de-jitter points is calculated, with the unit being the number of sampling points. The final result is rounded down and subject to a minimum value constraint to ensure that it is an integer and meets the minimum response requirements of the system. The baseline deshaking score represents the baseline deshaking score when the importance index is zero (i.e., the data is completely stationary and in the lowest weight state). In this embodiment, we take... The corresponding base de-shake time is 10 × 0.1s = 1 second; This is a sensitivity coefficient used to control the intensity of the influence of the importance index on the number of de-jittering points. The larger the value, the more significant the reduction in de-jitter points for the same importance index. This embodiment uses... ; The importance index after smoothing; This is the minimum allowed number of debouncing points, used to ensure the state machine has basic noise immunity under all circumstances, preventing false triggering caused by noise due to excessively short debouncing time. This embodiment uses... This corresponds to a minimum dejitter time of 0.2 seconds; The physical meaning of this formula is: denominator Follow Increasing the value leads to a decrease in the score, thus... Decrease. When When =0, ;when As it gradually increases, It then decreases until it reaches the lower limit. For example, suppose at a certain time Substituting into the formula, we get Rounded down, it is 6, which is the same as... After comparison, take the maximum value, and finally... This means that the dejittering time is reduced from 1 second to 0.6 seconds, and the state machine will be able to confirm the state transition conditions more quickly.

[0073] Calculation yields new Subsequently, the parameter feedback adjustment module immediately writes this value into the debouncing parameter register of the finite state machine, overwriting the original value. At each subsequent sampling point, when the state machine makes a transition condition judgment, the updated debouncing point count will be used for confirmation. This update mechanism is real-time and continuous, enabling the state machine's behavior to dynamically adjust as the importance of data changes during the test.

[0074] It is worth noting that the combined use of the smoothing filter and the modulation formula ensures that the changes in the dejittering parameters are gradual and bounded, preventing the dejittering time from instantly dropping to its minimum due to a sudden high importance index in a single window, thus avoiding oscillations in the control loop. Furthermore, since the number of dejittering points only affects the confirmation speed of state transitions without changing the threshold of the transition conditions themselves, the stability and security of the system are guaranteed.

[0075] Through the above design, the parameter feedback adjustment module achieves bidirectional closed-loop optimization between data acquisition and process control: on the one hand, the importance index of the data drives the response sensitivity of the state machine; on the other hand, the optimized state machine can more accurately capture key test stages and generate higher-value data. This adaptive mechanism significantly improves the test system's adaptability to complex operating conditions and avoids misjudgments or response delays caused by fixed parameters.

[0076] It also includes a storage and indexing module, used for: The compressed data blocks are associated with their corresponding comprehensive importance index, compression level identifier, start timestamp, and window duration and stored in the database; Create an index with timestamps and a comprehensive importance index as retrieval fields to enable quick data retrieval by importance in subsequent searches.

[0077] After the adaptive compression module completes the differential compression of the current window data and generates standardized data blocks containing all metadata and compressed content, these data blocks are sent to the storage and indexing module for persistent storage, laying the foundation for subsequent rapid retrieval and analysis. The core responsibility of the storage and indexing module is to build a storage structure that can efficiently write massive amounts of time-series data and support fast queries based on multi-dimensional conditions, ensuring that test engineers can backtrack to key data of any time period and any level of importance as needed.

[0078] The storage and indexing module first receives compressed data blocks from the adaptive compression module. Each data block is a structured data unit containing the following fields: Start timestamp : A Unix timestamp in milliseconds, precisely identifying the absolute moment when the data window began; Window duration : Fixed at 10 seconds, but the field is reserved for possible future window length adjustments; Overall Importance Index The dimensionless value calculated by the importance index fusion module ranges from [0,3] and represents the overall importance of the data in this window. Compression level label The value can be 1, 2 or 3, corresponding to lossless compression, rotating door compression (lossy) and statistical summary compression, respectively; Compressed data content: Depending on the compression level, the specific content contained in this field will vary. when At that time, the content is the original data stream processed by differential coding and general compression algorithm, which can be completely recovered by decompression; when At this time, the content is a sequence of key points preserved by the revolving door compression algorithm. Each key point includes a timestamp and the corresponding voltage, current, and temperature values. In addition, the data block of this compression level also includes an optional field "fluctuation mode", which can be "continuous fluctuation mode", "instantaneous fluctuation mode" or null. This label is calculated by the adaptive compression module based on the variance of the key point time interval. when At that time, the content is a set of statistical characteristic values ​​of each channel, including mean, variance, maximum value, minimum value, starting value, ending value, and rate of change; Upon receiving a data block, the storage and indexing module immediately initiates the write process. This embodiment employs a high-performance time-series database as the underlying storage engine, such as an LSM tree-based database, to support high-throughput continuous writes. The module parses each data block into a record, mapping the aforementioned fields to columns in the database table. Specifically, the database table design includes the following main columns: start_time (BIGINT), window_duration (INT), importance_index (FLOAT), comp_level (TINYINT), data_blob (BLOB), and the optional pattern (VARCHAR). The data_blob column stores compressed binary data, while the pattern column is only filled periodically when comp_level=2 and the fluctuation pattern is determined to be timed.

[0079] To support subsequent fast queries based on time and importance, the storage and indexing module maintains two types of core indexes during data writing: Primary index: Create a clustered index or B+ tree index with start_time as the first key. Since test data is naturally time-ordered, timestamp indexes can efficiently support slice queries by time range, such as retrieving all data windows within one hour after the start of an experiment; Secondary Index: An independent B+ tree index is built using importance_index as the secondary key. This index enables the system to quickly locate all data windows with a comprehensive importance index higher than a certain threshold without scanning the entire table; for example, when analyzing battery thermal runaway accidents, engineers can directly retrieve data windows with importance_index>=2.5, quickly focusing on the most critical time period.

[0080] In addition to the basic indexes mentioned above, this embodiment also supports combined query conditions. When a user query includes both a time range and an importance threshold, the database query optimizer can simultaneously perform an intersection operation on the two indexes, or select the optimal index path based on the data distribution, thereby further improving retrieval efficiency. For example, a user might want to retrieve "all windows with a comprehensive importance index greater than 2.0 on October 1, 2023". The system will first use the time index to quickly locate the data partition for that day, then combine it with the importance index to filter out records that meet the conditions, and finally return the corresponding compressed data block.

[0081] To further optimize storage efficiency, the storage and indexing module also introduces a data lifecycle management strategy. For windows with consistently low importance indices (such as statistical summary data generated during long periods of inactivity), the system can migrate them to lower-cost, high-capacity cold storage media, retaining only metadata pointers in the main database. Conversely, event data with high importance indices is always kept online at high speed, ensuring rapid access at any time. This tiered storage strategy effectively controls overall storage costs without sacrificing the performance of critical data access.

[0082] When a user or upper-level analytics application initiates a data query request, the query interface receives the query conditions (such as time range, importance index threshold, compression level, fluctuation pattern, etc.) and converts them into a standard database query statement (e.g., SQL). The database engine uses pre-built indexes to quickly locate and return matching data block records. The application then uses the compression level identifier in the record to call the appropriate decompression algorithm to decode the data_blob, restoring analyzable waveform data or statistical features, which are finally presented to the user.

[0083] After completing the refined design of the above modules, this embodiment introduces a nonlinear coupling mechanism between window fluctuation intensity and state weights to deeply integrate the microscopic dynamic characteristics of battery test data with macroscopic process semantics, generating a comprehensive importance index. Based on this index, it achieves hierarchical adaptation of compression strategies and dynamic feedback correction of state machine debouncing parameters. This closed-loop optimization architecture breaks the inherent pattern of data acquisition and process control being separated in traditional testing systems. It enables the system to autonomously identify transient segments containing high-value information during the testing process and retain them with high fidelity. Simultaneously, it adaptively adjusts the response sensitivity of the state machine according to the actual information content of the data, effectively avoiding state oscillations and loss of key features caused by fixed parameters. Therefore, this embodiment, while significantly reducing storage pressure, ensures the holographic recording of key events such as battery gas evolution, polarization transition, and precursors to thermal runaway. This allows the test report to accurately reproduce the dynamic response characteristics of the battery, providing a reliable data foundation for subsequent lifespan prediction and fault diagnosis.

[0084] Example 2: Figure 2 As shown, the automated data acquisition and report generation method for a battery testing laboratory includes the following steps: Step S1: Obtain the sequence of battery state parameters continuously collected during the test. The battery state parameters include voltage, current and temperature. Step S2: Calculate the window fluctuation intensity, which characterizes the severity of data fluctuations, based on the battery state parameter sequence; Step S3: Based on the current test state of the finite state machine in the test process, query the preset state weight corresponding to the test state; Step S4: Nonlinearly couple the window fluctuation intensity with the state weight to generate a comprehensive importance index for measuring the importance of the current data; Step S5: Based on the comprehensive importance index, determine the target compression level from at least two preset compression levels, and perform compression storage on the battery state parameter sequence according to the target compression level; Step S6: Based on the comprehensive importance index, dynamically adjust the debouncing parameters used for state transition confirmation in the finite state machine, and update the finite state machine with the adjusted debouncing parameters.

[0085] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0086] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0087] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An automated data acquisition and report generation system for a battery testing laboratory, characterized in that, include: Data acquisition module: Acquires a sequence of battery state parameters continuously collected during the battery test, including voltage, current and temperature; Fluctuation intensity analysis module: Calculates the window fluctuation intensity to characterize the severity of data fluctuations based on the battery state parameter sequence; State weight acquisition module: Based on the current test state of the finite state machine in the test process, query the preset state weight corresponding to the test state; Importance index fusion module: Nonlinearly couples the window fluctuation intensity with the state weight to generate a comprehensive importance index for measuring the importance of the current data; Adaptive compression module: Based on the comprehensive importance index, determines a target compression level from at least two preset compression levels, and performs compression storage on the battery state parameter sequence according to the target compression level; Parameter feedback adjustment module: Based on the comprehensive importance index, dynamically adjust the debouncing parameters used for state transition confirmation in the finite state machine, and update the finite state machine with the adjusted debouncing parameters.

2. The automated data acquisition and report generation system for a battery testing laboratory according to claim 1, characterized in that, The calculation process for the window fluctuation intensity specifically includes: The battery state parameter sequence is divided into voltage subsequence, current subsequence and temperature subsequence for the current window by using a fixed time window; Calculate the sum of the absolute values ​​of the differences between adjacent sampling points within each subsequence to obtain the cumulative change of each subsequence; The normalized fluctuation intensity of each subsequence is obtained by dividing the cumulative change of each subsequence by the product of the range of the corresponding sensor and the window duration. The maximum value among the normalized fluctuation intensities of the three subsequences is determined as the window fluctuation intensity of the current window.

3. The automated data acquisition and report generation system for a battery testing laboratory according to claim 1, characterized in that, The test states include at least constant current charging state, constant voltage charging state, discharging state and resting state, each state corresponds to a different preset state weight, and the preset state weights are preset based on test experience.

4. The automated data acquisition and report generation system for a battery testing laboratory according to claim 1, characterized in that, The nonlinear coupling in the importance index fusion module is implemented according to the following formula: ,in, As a comprehensive importance index, For state weights, For window fluctuation intensity, The amplification factor is preset; and the comprehensive importance index is limited to a preset value range.

5. The automated data acquisition and report generation system for a battery testing laboratory according to claim 1, characterized in that, The specific process of determining the target compression level based on the comprehensive importance index in the adaptive compression module includes: When the comprehensive importance index is greater than or equal to the first preset threshold, the target compression level is determined as the first compression level, and lossless compression is performed on the battery state parameter sequence. When the comprehensive importance index is greater than or equal to the second preset threshold and less than the first preset threshold, the target compression level is determined to be the second compression level, and the rotating door compression algorithm is used to perform lossy compression on the battery state parameter sequence to retain the key points in the sequence. When the comprehensive importance index is less than the second preset threshold, the target compression level is determined to be the third compression level, and the statistical features of the battery state parameter sequence are extracted and stored. The statistical features include at least one of the following: mean, variance, maximum value, minimum value, starting value, ending value, and rate of change.

6. The automated data acquisition and report generation system for a battery testing laboratory according to claim 1, characterized in that, The specific execution process of the parameter feedback adjustment module includes: The comprehensive importance index is subjected to a first-order low-pass filter to obtain a smoothed importance index; Based on the smoothed importance index, the new number of de-jitter points is calculated according to the following modulation rules: ;in, For the new anti-shake points, The preset minimum number of debouncing points, Based on the de-jitter points, This is the sensitivity coefficient. This is the smoothed importance index.

7. The automated data acquisition and report generation system for a battery testing laboratory according to claim 1, characterized in that, It also includes a storage and indexing module, used for: The compressed data blocks are associated with their corresponding comprehensive importance index, compression level identifier, start timestamp, and window duration and stored in the database; Create an index with timestamps and a comprehensive importance index as retrieval fields to enable quick data retrieval by importance in subsequent searches.

8. The automated data acquisition and report generation system for a battery testing laboratory according to claim 5, characterized in that, When the target compression level is the second compression level, it also includes: Extract the timestamps of key points retained after compression; Based on the timestamps of the key points, determine the time interval sequence between adjacent key points; Calculate the variance of the time interval sequence as a fluctuation density index; The fluctuation pattern of the current window data is determined based on the comparison between the fluctuation density index and the preset fluctuation threshold.

9. The automated data acquisition and report generation system for a battery testing laboratory according to claim 8, characterized in that, The process of determining the fluctuation pattern of the current window data specifically includes: If the variance is less than the first preset threshold, then the fluctuation pattern is determined to be a continuous fluctuation pattern. If the variance is greater than the second preset threshold, then the fluctuation mode is determined to be an instantaneous fluctuation mode; The fluctuation pattern is stored as an additional attribute associated with the compressed data block.

Images