Calibration control method and device for battery test channel

Abstract

The application provides a calibration control method and device for a battery test channel, the method comprising: obtaining electrical test data generated by a test channel on a plurality of measured batteries and extracting statistical features; performing state classification based on the statistical features to continuously monitor the channel state and obtain state monitoring information; when the state monitoring information meets a preset early warning trigger condition, predicting the precision deviation trend of the test channel in a future time interval to obtain a deviation prediction result; triggering calibration verification when the deviation prediction result meets a preset calibration trigger condition; in response to the calibration verification, performing an investigation of external abnormalities of the test channel; obtaining the historical calibration frequency of the test channel after excluding the external abnormalities; and issuing a calibration instruction to the test channel when the historical calibration frequency does not exceed a preset frequency threshold. The application can perceive the precision deviation in advance and exclude external interference, avoid false calibration and invalid calibration, and ensure the accuracy and reliability of the calibration control.

Description

Technical Field

[0001] This application relates to the field of battery testing technology, and in particular to a calibration control method and apparatus for a battery testing channel. Background Technology

[0002] During battery production testing, test channels may experience hardware aging or parameter drift under long-term continuous operation, leading to a decrease in electrical testing accuracy. Existing technologies typically employ periodic shutdowns for static calibration or calibration triggered by test anomalies.

[0003] However, in the process of developing this application, the inventors discovered the following defects in the prior art:

[0004] Periodic calibration requires periodic interruptions to test operations to connect to a standard source for calibration, which not only affects the continuous operation of the test production line, but also cannot cope with sudden or gradual accuracy drift that occurs between two calibration cycles.

[0005] Calibration triggered by test anomalies is passively activated only after electrical test data has exceeded the accuracy tolerance range. It fails to detect the degradation process of channel accuracy in advance, leading to inaccurate battery data during periods of accuracy deviation. Furthermore, when electrical test data fluctuates, it cannot distinguish whether the fluctuation is caused by differences in the incoming material characteristics of the tested battery batch, abnormal contact of the test fixture, or hardware drift of the test channel itself. This makes it easy for external interference factors to mistakenly perform calibration on normal test channels, requiring work interruption and impacting operational efficiency.

[0006] In addition, when there are hardware defects such as component damage or poor soldering in the test channel, the existing technology still continuously sends calibration commands to compensate for parameters, which cannot restore the defects of the channel and will cover up the hardware fault status of the channel. Summary of the Invention

[0007] The purpose of this application is to provide a calibration control method and apparatus for battery test channels that improves the accuracy and reliability of calibration control.

[0008] To achieve the above objectives, one embodiment of this application provides a calibration control method for a battery test channel, comprising the following steps: Acquire electrical test data generated by the test channel for testing several batteries under test, and extract the statistical characteristics of the electrical test data; Based on the statistical characteristics, state classification is performed to continuously monitor the channel state of the test channel and obtain state monitoring information; When the status monitoring information meets the preset early warning triggering conditions, the accuracy deviation trend of the test channel in the future time interval is predicted to obtain the deviation prediction result. When the deviation prediction result meets the preset calibration trigger condition, calibration verification is triggered; In response to the calibration verification, troubleshooting for external anomalies related to the test channel is performed; After ruling out the external anomalies, obtain the historical calibration frequency of the test channel; When the historical calibration frequency does not exceed the preset frequency threshold, a calibration command is sent to the test channel to calibrate the test channel.

[0009] As a further improvement to this application, the extraction of statistical features from the electrical test data includes: The electrical parameters generated by the test channel testing a preset number of the tested batteries within a sliding data window are obtained. The electrical parameters include at least one of input current, output current, input voltage, output voltage, and internal resistance. Calculate at least one of the mean, variance, and range of each electrical parameter within the sliding data window, and use the calculation result as the statistical feature.

[0010] As a further improvement to this application, the step of classifying the status based on the statistical characteristics to continuously monitor the channel status of the test channel and obtain status monitoring information includes: The statistical features are classified in real time using a classification model, and the classification results are output to indicate whether the test channel is currently in a normal state, a warning state, or an out-of-tolerance state. The normal state indicates that the electrical test data has no abnormal fluctuations, the warning state indicates that the test channel has a potential accuracy drift trend and the test has no defects, and the out-of-tolerance state indicates that the test channel is close to or exceeds the accuracy specification tolerance range. The classification results are continuously recorded to generate the status monitoring information.

[0011] As a further improvement to this application, the classification model is a random forest classification model; The preset warning triggering condition is: all the statistical features extracted from the preset number of groups in the status monitoring information are classified as the warning status, wherein the preset number of groups is at least two groups; The calibration control method for the battery test channel further includes: When the classification model outputs the out-of-tolerance state, the probability of the out-of-tolerance state occurring in the preset number of analysis groups is statistically analyzed. Based on the probability range in which the occurrence probability falls, a corresponding graded handling strategy is executed for the test channel. The graded handling strategy includes at least one of logging, outputting early warning signals, and executing channel blocking.

[0012] As a further improvement to this application, the step of predicting the accuracy deviation trend of the test channel in a future time interval and obtaining the deviation prediction result when the status monitoring information meets the preset early warning triggering conditions includes: In response to the state monitoring information satisfying the preset early warning triggering condition, multiple statistical features within the sliding time window are input into the long short-term memory network model; The long short-term memory network model is used to calculate the probability of the test channel's accuracy deviating from the accuracy specification tolerance range within a future preset period, and the deviation probability is used as the deviation prediction result.

[0013] As a further improvement to this application, the step of triggering calibration verification when the deviation prediction result meets the preset calibration trigger condition includes: Determine whether the deviation probability in the deviation prediction result is greater than a preset probability threshold; If the deviation probability is greater than the preset probability threshold, then the preset calibration trigger condition is met, and calibration verification is triggered. If the deviation probability is less than or equal to the preset probability threshold, then new statistical features are obtained by scrolling through the sliding time window, the data input to the long short-term memory network model is updated, and the process returns to the step of calculating the deviation probability.

[0014] As a further improvement to this application, The external anomaly includes at least one of the batch anomaly corresponding to the plurality of batteries under test and the vehicle anomaly corresponding to the test vehicle carrying the plurality of batteries under test. The process of troubleshooting external anomalies for the test channel includes: Determine whether there are any batch anomalies based on the batch test data associated with the battery under test; If no batch anomaly is found, then the presence of a vehicle anomaly is determined based on the historical test data of the test vehicle carrying the battery under test. The step of determining whether there is a batch anomaly based on the batch test data associated with the tested battery includes: Obtain the deviation prediction results corresponding to the testing of the batches of the several tested batteries by multiple test channels; Compare the deviation prediction results for each of the test channels: If the deviation prediction results of multiple test channels all show the same deviation characteristics, it is determined that there is an abnormality in the batch, the current calibration verification is terminated and the test channel continues to operate. If only the deviation prediction result of the current test channel shows deviation characteristics, then it is determined that there is no batch anomaly.

[0015] As a further improvement to this application, the determination of whether there is a vehicle anomaly based on the historical test data of the test vehicle carrying the battery under test includes: Retrieve historical test data of the test vehicle within a preset historical time window; Calculate the average vehicle deviation corresponding to the historical test data; Obtain the overall average deviation and standard deviation of all test vehicles within the preset historical time window; If the difference between the average deviation of the vehicle and the overall average deviation is greater than a preset outlier threshold, it is determined that there is an abnormality in the vehicle, the current calibration verification is terminated and the test channel continues to operate, wherein the preset outlier threshold is 3 times the standard deviation of the deviation; If the difference between the vehicle's average deviation and the overall average deviation is less than or equal to the preset outlier threshold, then it is determined that there is no vehicle anomaly.

[0016] As a further improvement to this application, after obtaining the historical calibration frequency of the test channel, the method further includes: If the historical calibration frequency exceeds the preset frequency threshold, the calibration of the test channel is terminated, and the test channel is shielded and isolated.

[0017] To achieve one of the above objectives, one embodiment of this application provides a calibration control device for a battery testing channel, comprising: The feature extraction module is used to acquire electrical test data generated by the test channel for testing several batteries under test, and to extract the statistical features of the electrical test data; The status monitoring module is used to classify the status based on the statistical characteristics to continuously monitor the channel status of the test channel and obtain status monitoring information. The trend prediction module is used to predict the accuracy deviation trend of the test channel in a future time interval when the status monitoring information meets the preset early warning triggering conditions, and obtain the deviation prediction result. The verification trigger module is used to trigger calibration verification when the deviation prediction result meets the preset calibration trigger conditions; An anomaly detection module is used to perform anomaly detection for the test channel in response to the calibration verification. The frequency acquisition module is used to acquire the historical calibration frequency of the test channel after excluding the external anomalies. The calibration control module is used to send a calibration command to the test channel to calibrate the test channel when the historical calibration frequency does not exceed a preset frequency threshold.

[0018] Compared with commonly used technologies, this application has the following advantages: This application can trigger calibration verification in advance before the accuracy deviation occurs, avoiding test data inaccuracy caused by the actual deviation of the test channel; at the same time, it can identify and eliminate external anomalies during calibration verification, avoiding miscalibration of normal test channels; moreover, it only issues calibration instructions when the historical calibration frequency does not exceed the preset frequency threshold, preventing continuous invalid calibration of frequently inaccurate test channels, and ensuring the accuracy and reliability of test channel calibration control. Attached Figure Description

[0019] Figure 1 This is a flowchart of a calibration control method for a battery test channel according to an embodiment of this application.

[0020] Figure 2 This is a schematic diagram of a calibration control device for a battery test channel according to an embodiment of this application. Detailed Implementation

[0021] The present application will now be described in detail with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present application, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the scope of protection of this application.

[0022] One embodiment of this application provides a calibration control method and apparatus for a battery test channel to improve the accuracy and reliability of test channel calibration control.

[0023] The method provided in this embodiment can be applied to battery manufacturing and formation / capacity testing systems. In the battery manufacturing process, multiple testing devices and numerous testing channels are typically configured to perform electrical performance tests on mass-produced batteries, including voltage, current, internal resistance, and charge / discharge capacity.

[0024] During long-term continuous operation of the test channel, the sampling circuit, amplification components, and reference source inside the channel are prone to component aging or temperature drift, which can lead to deviations in the electrical measurement accuracy of the test channel. Existing control methods typically employ fixed-period offline calibration shutdowns, or passively trigger calibration shutdowns after the data measured by the test channel has exceeded the accuracy specification tolerance range.

[0025] However, fixed-cycle calibration cannot cope with sudden accuracy drift during cycle operation, while passively triggered calibration after actual deviation has obvious lag, which can easily cause batch test data to be inaccurate during channel deviation. At the same time, in the actual production environment, abnormal fluctuations in electrical test data are often caused by complex reasons. In addition to hardware drift of the test channel itself, external factors such as material and process fluctuations of the batch of batteries under test and poor contact of the test carrier probes carrying the batteries can all cause test data to deviate.

[0026] Existing technologies lack a multi-dimensional attribution and troubleshooting mechanism when data fluctuations occur, which can easily lead to incorrect parameter adjustments being made to normal test channels. Furthermore, for high-frequency inaccurate channels with physical damage to their internal hardware, continuous parameter compensation not only fails to restore the hardware defects but also masks the true fault state of the channel.

[0027] To address the aforementioned issues, this specific embodiment provides a calibration control method for a battery testing channel. This method extracts statistical features to provide early warnings and predict future trends for the channel status. After triggering calibration verification, it performs decoupled troubleshooting of batch anomalies and vehicle anomalies. Finally, it issues calibration commands based on historical calibration frequencies, thereby achieving precise and reliable calibration control of the testing channel accuracy.

[0028] The following is combined Figure 1 This application describes a calibration control method for a battery test channel according to an embodiment. Although this application provides method operation steps as shown in the following embodiments or flowcharts, the execution order of steps that do not have a necessary causal relationship in logic is not limited to the execution order provided in the embodiments of this application.

[0029] Specifically, the calibration control method for a battery test channel in this embodiment includes the following steps: Step S10: Obtain electrical test data generated by the test channel for testing several batteries under test, and extract the statistical characteristics of the electrical test data.

[0030] Step S20: Based on the statistical characteristics, perform state classification to continuously monitor the channel status of the test channel and obtain state monitoring information.

[0031] Step S30: When the status monitoring information meets the preset early warning triggering conditions, predict the accuracy deviation trend of the test channel in the future time interval and obtain the deviation prediction result.

[0032] Step S40: When the deviation prediction result meets the preset calibration trigger condition, the calibration verification is triggered.

[0033] Step S50: In response to the calibration verification, perform troubleshooting for external anomalies in the test channel.

[0034] Step S60: After ruling out the external anomalies, obtain the historical calibration frequency of the test channel.

[0035] Step S70: When the historical calibration frequency does not exceed the preset frequency threshold, a calibration command is sent to the test channel to calibrate the test channel.

[0036] In step S10, during the electrical testing of several batteries on the production line, the test channel continuously collects electrical test data output by the test channel while the batteries are mounted on a carrier and electrically connected to the carrier. The electrical test data includes raw sampled values ​​characterizing the battery's electrical properties, such as input current, output current, input voltage, output voltage, internal resistance, and capacity during the test. Since single-test data for a single battery is easily affected by occasional minor disturbances during the test, this embodiment extracts statistical features from the dataset corresponding to multiple batteries under test. These statistical features can be numerical indicators characterizing the central tendency, dispersion, or distribution pattern of the dataset, such as one or more combinations of mean, variance, standard deviation, range, or median.

[0037] By extracting statistical features, random transient noise during single-cell testing was filtered out, enabling an objective characterization of the data distribution pattern of the test channel within a certain time window, thus providing a stable data foundation for accurately judging the actual operating status of the channel in the future.

[0038] In step S20, a pre-configured classification algorithm or pattern recognition model is used to perform real-time classification calculations on the statistical features extracted in step S10 to determine the current operational health level of the test channel and generate time-series status monitoring information. The channel status can be divided into a normal state, characterized by stable and drift-free data; an early warning state, characterized by a slight drift trend in data but no actual deviation; and an out-of-tolerance state, characterized by test data approaching or exceeding the tolerance limit.

[0039] Status monitoring information can be the current channel status, or a representation of several channel statuses recorded based on preset rules, such as a summary of several consecutively recorded channel statuses.

[0040] In step S30, while continuously monitoring the status information, corresponding preset early warning trigger conditions are set. For example, if the channel status is detected as being in an early warning state for multiple consecutive evaluation cycles, once the status monitoring information meets the preset early warning trigger conditions, it indicates that the test channel has a potential gradual degradation trend. The prediction mechanism is then activated to extrapolate the accuracy deviation trend of the test channel in the future time interval based on previous historical data sequences, and a deviation prediction result is generated. The deviation prediction result can be characterized by the probability value of test data exceeding the accuracy specification tolerance range in the future time interval, a deviation prediction curve, or the predicted deviation magnitude.

[0041] For example, the preset warning trigger condition is that two consecutive sets of state classification results are in a warning state. Each statistical feature is based on data obtained from 40 tested batteries. That is, if two warning states are obtained based on test data from 80 consecutive tested batteries, then the prediction of the deviation from the trend will be triggered.

[0042] This embodiment predicts the deviation trend of the future time interval after the early warning is triggered, which changes the lag defect of the traditional control method that only passively detects the deviation after the channel has actually exceeded the tolerance. It uses the time evolution law to assess the deviation risk of the channel in the future in advance, and buys time for subsequent production decisions.

[0043] Step S40 compares the deviation prediction result output in step S30 with the preset calibration trigger conditions. When the deviation prediction result reaches a preset severity level, such as the predicted probability of future deviations reaching a preset probability threshold, or the predicted deviation magnitude reaching a tolerance threshold, the preset calibration trigger conditions are determined to be met, and a calibration verification request is generated, triggering the calibration verification process. Alternatively, if the preset calibration trigger conditions are not met, it indicates that the current trend does not yet pose a risk, and routine monitoring and prediction updates can be maintained.

[0044] This embodiment can filter out temporary, self-converging minor fluctuations, and only initiates the verification process when the deviation in the future has a high degree of certainty, thus avoiding unnecessary frequent calibration overhead.

[0045] In step S50, after the calibration verification is triggered, a calibration command is not directly issued to the test channel. Instead, a troubleshooting step for external interference sources is performed first. In the battery testing production line, abnormal fluctuations and accuracy deviations in electrical test data may be caused by external interference factors independent of the test channel's own hardware. These external anomalies include, but are not limited to: batch anomalies caused by fluctuations in raw materials or upstream processes of the battery under test; carrier anomalies caused by wear or increased contact impedance of the test fixture probes carrying the battery under test; and common environmental interference caused by changes in temperature and humidity in the test environment or fluctuations in workshop power supply.

[0046] In response to calibration verification, this embodiment retrieves multi-dimensional reference data or production line operation data associated with the test channel to check for the existence of the external anomaly. If the external anomaly is detected, it is determined that the deviation in the prediction is caused by external interference factors and is not a real drift of the measurement hardware of the test channel itself. The calibration verification is terminated and the channel continues to operate, preventing miscalibration caused by attributing external problems to the channel hardware.

[0047] In step S60, if step S50 confirms that there are no external anomalies, it means that the root cause of the data fluctuation has been ruled out as interference from external products and tooling fixtures, and is caused by the drift of internal hardware parameters of the test channel. At this time, the database is further accessed to retrieve and count the historical calibration frequency of the test channel within the preset historical statistical period.

[0048] In step S70, the obtained historical calibration frequency is compared with the preset frequency threshold. If the historical calibration frequency does not exceed the preset frequency threshold, it indicates that the accuracy drift of the test channel is an occasional, slight drift caused by normal aging of components or environmental temperature drift, which is within the recoverable range. A calibration command is sent to the test channel to restore the test accuracy by adjusting the bias compensation register or calibration parameters inside the channel.

[0049] If the historical calibration frequency exceeds the preset frequency threshold, it indicates that the channel is frequently inaccurate in a short period of time, and there are irreversible defects such as poor soldering of components, breakdown or hardware damage, which will inhibit the issuance of calibration commands.

[0050] By setting a calibration frequency threshold, continuous and ineffective parameter compensation for channels with damaged hardware is avoided, effectively preventing the risk of using software calibration to cover up hardware failures and ensuring the accuracy and reliability of test channel calibration control.

[0051] This embodiment monitors the channel status by extracting statistical features from electrical test data, and predicts the accuracy deviation trend of the test channel in the future time interval when the preset early warning trigger conditions are met. This changes the lag of the traditional control method, which passively handles the channel after it actually exceeds the tolerance, and enters the intervention process in advance. On this basis, by performing a check on external anomalies after triggering calibration verification, it avoids the miscalibration action caused by external interference factors to the test channel.

[0052] In addition, by combining historical calibration frequency with preset frequency thresholds after excluding external anomalies, and issuing calibration commands only when the threshold is not exceeded, the system prevents continuous and ineffective parameter compensation for frequently inaccurate damaged channels, avoids the risk of masking channel hardware failure status, and ensures the objectivity, accuracy, and reliability of the battery test channel calibration control process and production line operation.

[0053] In one embodiment, step S10 extracts statistical features from the electrical test data, including: Step S11: Obtain the electrical parameters generated by the test channel testing a preset number of the tested batteries within the sliding data window. The electrical parameters include at least one of the following: input current, output current, input voltage, output voltage, and internal resistance.

[0054] Step S12: Calculate at least one of the average, variance, and range of each electrical parameter within the sliding data window, and use the calculation result as the statistical feature.

[0055] In step S11, during the electrical testing of several batteries being transported on the production line in the testing channel, a sliding data window of a preset length is constructed in the time domain. The preset number represents the number of battery samples the window can hold, and can be pre-configured according to production cycle and testing requirements, for example, set to include data from 10 to 50 batteries undergoing continuous testing. As the testing process progresses, whenever the testing channel completes testing a new battery, the newly generated electrical parameters enter the sliding data window, while the oldest data that entered the window first is removed, ensuring that the window always maintains a fixed preset number of the latest test data.

[0056] Regarding the specific acquisition of electrical parameters, depending on the different battery testing steps, the electrical parameters may include at least one of the key electrical physical quantities such as the input current and input voltage measured by the test channel during the charging phase, the output current and output voltage measured during the discharging phase, and the battery internal resistance measured through the excitation response.

[0057] The uses of the various statistical calculations in step S12 are as follows: By calculating the average value, the central reference position of the measurement data within the time window can be objectively quantified, which can be used to characterize whether there is an overall translation of the channel measurement reference.

[0058] By calculating the variance or standard deviation, the dispersion of measurement data relative to the mean can be quantified, which can be used to characterize the repeatability and stability of the channel measurement output.

[0059] By calculating the range, such as the difference between the maximum and minimum values ​​within a sliding data window, the maximum local fluctuation range of the test data can be reflected, which can be used to capture transient disturbances during the test process.

[0060] One or more combinations of the calculated mean, variance, and range are used to construct a statistical feature vector characterizing the distribution of the test window.

[0061] Through the above steps S11 and S12, the statistical characteristics of electrical parameters are extracted using a sliding data window, which smooths out the random noise caused by the slight differences of individual batteries or the occasional jitter at the moment of contact of the test clamp, and extracts the characteristics that can stably reflect the measurement reference offset and dispersion changes of the test channel.

[0062] In one embodiment, step S20 performs state classification based on the statistical characteristics to continuously monitor the channel state of the test channel and obtain state monitoring information, including: Step S21: Use a classification model to classify the statistical features in real time and output the classification results representing whether the test channel is currently in a normal state, a warning state, or an out-of-tolerance state; wherein, the normal state indicates that the electrical test data has no abnormal fluctuations, the warning state indicates that the test channel has a potential accuracy drift trend and the test has no defects, and the out-of-tolerance state indicates that the test channel is close to or exceeds the accuracy specification tolerance range.

[0063] Step S22: Continuously record the classification results to generate the status monitoring information.

[0064] Step S21 can perform classification using a classification model, which can be implemented using classification algorithms such as decision trees, ensemble learning models, support vector machines, or neural networks. During operation, whenever a new set of statistical features is generated in step S12, these statistical features are input into the classification model for online classification inference. The classification model outputs the corresponding classification result based on the distribution of the input features in the multidimensional feature space.

[0065] If the output is in a normal state (OK state), it indicates that the electrical test data in the current sliding data window is concentrated and the benchmark is stable, with no abnormal fluctuations, and the test channel is within the controlled nominal accuracy range.

[0066] If the output is a warning state, it indicates that the statistical characteristics of the electrical test data have shown a slight unidirectional deviation, which indicates that there is a potential accuracy drift trend in the test channel due to early temperature drift or slight aging of components. However, at this time, the test results of the battery under test are still within the qualified specification range.

[0067] If the output is out of tolerance (NG), it indicates that the statistical characteristics of the electrical test data have deviated significantly, and the measurement accuracy of the test channel is extremely close to or even exceeds the preset accuracy specification tolerance range, which poses a direct risk of inaccurate battery test judgment.

[0068] Step S22 continuously adds the classification results output by the classification model at each time point to the status monitoring cache or database in chronological order, forming time-series status monitoring information that is updated as the test process progresses. The status monitoring information reflects the trajectory of the channel status change during the test channel's operation.

[0069] Steps S21 and S22 above use a classification model to quantify and map statistical features into specific discrete health states, thereby realizing online hierarchical monitoring of the test channel's measurement performance.

[0070] In one embodiment, the classification model is a random forest classification model. The random forest classification model is composed of multiple decision trees. During the training phase, different subsets are constructed based on a bootstrap resampling method, and random feature subsets are partitioned. During real-time classification, the statistical features extracted in step S12 are input into the random forest classification model composed of multiple decision trees. Each decision tree makes a decision on the statistical feature, and finally, the corresponding classification result is output through a multi-tree voting mechanism. The random forest classification model has a strong tolerance for local abnormal noise in multi-dimensional statistical features, effectively reducing the probability of misclassification caused by transient perturbations of a single feature, thereby ensuring the stability and accuracy of channel state classification decisions.

[0071] The preset warning triggering condition for step S30 is: all the statistical features extracted from the preset number of groups in the status monitoring information are classified as the warning status, wherein the preset number of groups is at least two groups.

[0072] The preset number of groups can be set to 2, 3 or 5 consecutive groups. When the classification model outputs the warning state only once, it may be an occasional transient perturbation. However, when the statistical characteristics of the preset number of groups recorded by the status monitoring information are all judged as warning states, or for example, when two out of three classifications are judged as warning states, it indicates that a deterministic one-way benchmark drift or dispersion deterioration trend has indeed formed inside the test channel, and only then is the preset warning trigger condition satisfied.

[0073] This embodiment filters out false triggers caused by occasional minor disturbances by setting a time series confirmation mechanism with a preset number of groups, avoids frequent calls to computing resources to perform subsequent predictions, and improves the anti-interference capability of the control logic.

[0074] The calibration control method for the battery test channel further includes: Step S23: When the classification model outputs the out-of-tolerance state, the probability of the out-of-tolerance state occurring in the preset number of analysis groups is calculated.

[0075] Step S24: Based on the probability range in which the occurrence probability is located, execute the corresponding graded handling strategy for the test channel. The graded handling strategy includes at least one of logging, outputting early warning signals, and executing channel blocking.

[0076] In step S23, when the random forest classification model outputs an out-of-tolerance state for the current input statistical features, it indicates that the test channel has approached or reached the preset accuracy specification boundary. To evaluate whether this out-of-tolerance phenomenon is a sudden, accidental event or a persistent, serious fault, an analysis window containing a preset number of groups (e.g., 10 to 20 consecutive groups of feature data) is set. The frequency of being classified as out-of-tolerance in the preset number of analysis groups is counted out, and the probability of the out-of-tolerance state is calculated.

[0077] In step S24, different probability intervals and corresponding graded handling actions are pre-defined, as illustrated below: When the calculated probability of occurrence is in the lower first probability range (e.g., the probability of occurrence is less than or equal to 10%), it indicates that the current out-of-range state is an occasional low-frequency fluctuation. The log handling strategy is executed, and the abnormal event and the test channel number are recorded in the log for tracking, so as to maintain the continued operation of the test channel.

[0078] When the calculated probability of occurrence is in the middle second probability range (e.g., the probability of occurrence is greater than 10% and less than or equal to 20%), it indicates that the out-of-tolerance phenomenon has a tendency to occur intermittently. The handling strategy of outputting a warning signal is implemented, and an audible and visual alarm or communication prompt message is sent to the monitoring terminal or upper management to remind the production line maintenance personnel to intervene and investigate, but the current test operation is not interrupted.

[0079] When the calculated probability of occurrence is in the high third probability range (e.g., the probability of occurrence is greater than 20%), it indicates that the test channel has experienced high-frequency or persistent serious inaccuracies. If the test channel continues to run, it will directly cause large-scale misjudgments of battery test results. The channel shielding strategy should be implemented by issuing hardware control commands to cut off the test stimulus output of the test channel or marking the channel as unavailable in the test scheduling software to prevent subsequent batteries from being allocated to the test channel.

[0080] Through the above steps S23 and S24, a dynamic hierarchical response mechanism based on the probability distribution of out-of-tolerance was established, realizing differentiated handling of channel faults of different severity. This not only avoids the production capacity loss caused by blindly shutting down the channel due to low-frequency occasional jitter, but also ensures that isolation can be quickly implemented when the channel experiences high-frequency severe inaccuracy, thus ensuring the safety and controllability of the battery testing process.

[0081] In one embodiment, step S30, when the status monitoring information meets a preset early warning triggering condition, predicts the accuracy deviation trend of the test channel within a future time interval to obtain a deviation prediction result, including: Step S31: In response to the state monitoring information satisfying the preset early warning triggering condition, input multiple statistical features within the sliding time window into the long short-term memory network model.

[0082] Step S32: Calculate the deviation probability of the test channel's accuracy deviating from the accuracy specification tolerance range within a future preset period using the long short-term memory network model, and use the deviation probability as the deviation prediction result.

[0083] In step S31, when the status monitoring information meets the preset early warning triggering condition (e.g., the classification results of two consecutive groups are both in an early warning state), it is determined that the preset early warning triggering condition is met, and the time series prediction process is then started. Historical data of the test channel within a preset sliding time window is extracted from the database or runtime cache, and multiple statistical features arranged chronologically within the time window are extracted, such as the mean sequence, variance sequence, and range sequence of each sampling period, and input into the pre-trained Long Short-Term Memory (LSTM) network model.

[0084] The Long Short-Term Memory (LSTM) network model has a gated recurrent unit consisting of an input gate, a forget gate, and an output gate. Through the gating mechanism, it regulates the degree of retention and forgetting of historical information, and can effectively capture the nonlinear degradation evolution pattern over a long period of time.

[0085] In step S32, after receiving the input temporal feature matrix, the Long Short-Term Memory Network Model performs multi-step cyclic deduction in the time domain to simulate the evolution trajectory of the measurement accuracy of the test channel within a preset period in the future, such as the evolution trajectory within the running time span of 3 to 10 days in the future, and calculates the deviation probability of the evolution trajectory exceeding the accuracy specification tolerance range, and outputs the calculated deviation probability as the deviation prediction result.

[0086] Through steps S31 and S32, the lightweight random forest classification state monitoring at the front end is combined with the deep long short-term memory network temporal regression prediction at the back end. The front-end model is responsible for filtering potential drift channels in real time, while the back-end model only starts to infer the future temporal evolution trend after the front end is triggered. On the one hand, this achieves efficient use of computing resources, and on the other hand, it utilizes the long short-term memory network's ability to model long-term sequence dependencies to map discrete historical statistical features into a probability index of channels exceeding tolerance in a specific future period, providing a basis for whether calibration verification needs to be performed.

[0087] In one embodiment, step S40, which triggers calibration verification when the deviation prediction result meets a preset calibration trigger condition, includes: Step S41: Determine whether the deviation probability in the deviation prediction result is greater than a preset probability threshold.

[0088] Step S42: If the deviation probability is greater than the preset probability threshold, it is determined that the preset calibration trigger condition is met, and calibration verification is triggered.

[0089] Step S43: If the deviation probability is less than or equal to the preset probability threshold, then new statistical features are obtained by scrolling through the sliding time window, the data input to the long short-term memory network model is updated, and the process returns to the step of calculating the deviation probability.

[0090] In step S41, a pre-configured preset probability threshold is retrieved, and the calculated deviation probability is compared with the preset probability threshold to quantify the degree of certainty that the test channel will actually exceed the tolerance within a preset period in the future.

[0091] In step S42, when the comparison result shows that the deviation probability is greater than a preset probability threshold, it indicates that the test channel has a deterministic trend of accuracy degradation in its temporal evolution. If no intervention is taken, the channel is likely to exceed the accuracy tolerance range in the future preset period. Then, it is determined that the preset calibration trigger condition is met, a calibration verification request is generated, and the control process is advanced to step S50 to initiate verification and investigation of external anomalies.

[0092] In step S43, when the comparison result shows that the deviation probability is less than or equal to a preset probability threshold, it indicates that the current degradation rate of the test channel is low, or the current data fluctuation has not yet constituted a deterministic risk of exceeding tolerance. At this time, the downstream calibration and verification actions are not triggered, but the online continuous simulation state of the Long Short-Term Memory Network Model is maintained. As the test channel continues to advance the subsequent battery testing process, the sliding time window slides forward in the time domain, ingesting the latest statistical features extracted from the new test data in real time, and removing the earliest recorded historical features from the window, thereby updating the time series data sequence input to the Long Short-Term Memory Network Model; subsequently, the control flow returns to execute the step of recalculating the deviation probability based on the updated time series data using the Long Short-Term Memory Network Model, forming a rolling simulation loop.

[0093] In this embodiment, when the deviation probability has not reached the trigger limit, the latest test features are continuously absorbed and the prediction results are continuously corrected by sliding time window. This dynamically tracks the actual evolution path of channel accuracy and avoids premature interruption of the test process due to early weak disturbances. When the deviation probability exceeds the limit, the deterministic inaccuracy risk can be quickly captured and the process can be transferred to verification, ensuring the accuracy of calibration triggering timing and the stability of timing prediction mechanism.

[0094] In one embodiment, the external anomaly includes at least one of batch anomalies corresponding to the plurality of tested batteries and carrier anomalies corresponding to the test carrier carrying the plurality of tested batteries. Step S50 performs an investigation of external anomalies for the test channel, including: Step S51: Determine whether there is a batch anomaly based on the batch test data associated with the battery under test.

[0095] Step S52: If there is no batch anomaly, determine whether there is a vehicle anomaly based on the historical test data of the test vehicle carrying the battery under test.

[0096] In the production line testing environment, due to slight fluctuations in the activity of positive and negative electrode materials, electrolyte ratio, or previous manufacturing processes, the electrical reference center of different batches of tested batteries may have objectively slight differences. For example, under the same specification, the open circuit voltage reference of a normal batch is 1.50V, while the open circuit voltage reference of another normal batch is 1.58V. Both are within the qualified specification range of 1.5V to 1.6V, but the statistical characteristics of different batches are different.

[0097] When the production line switches incoming batches or the test channel receives a new batch of batteries, the test data will show a step change. If the model is only based on the time series of a single test channel, it is easy to misjudge this step change as a continuous accuracy drift in the channel hardware.

[0098] Therefore, this embodiment uses the control variable method, combined with the associated batch test data, to investigate whether the current accuracy deviation is caused by common fluctuations in the batch to which the tested battery belongs.

[0099] Once it is determined that there are no batch anomalies, it indicates that the data fluctuations are not due to batch-specific characteristics of the battery materials themselves. At this point, the investigation focus shifts to the carrier level. By retrieving the test data of the target test carrier carrying the battery under test during historical periods, it is possible to check for carrier anomalies such as probe wear, oxidation, or increased contact resistance, ensuring that carrier interference is effectively eliminated before entering the machine for self-calibration.

[0100] Specifically, step S51, based on the batch test data associated with the battery under test, determines whether there is a batch anomaly, including: Step S511: Obtain the deviation prediction results corresponding to the tests conducted by multiple test channels on the batches to which the tested batteries belong.

[0101] Step S512: Compare the deviation prediction results of each test channel.

[0102] Step S513: If the deviation prediction results of multiple test channels all show consistent deviation characteristics, it is determined that there is an abnormality in the batch, the calibration verification is terminated and the test channel continues to operate.

[0103] Step S514: If only the deviation prediction result of the current test channel shows deviation characteristics, then it is determined that there is no batch anomaly.

[0104] In step S511, the batch number of the battery under test is queried from the Manufacturing Execution System (MES) or the production line database based on the unique identification code (such as SN code) of the battery under test. Then, the future deviation prediction results output by the corresponding long short-term memory network models of multiple other test channels in operation on the production line when testing batteries of the same batch number are retrieved.

[0105] In step S512, extreme outlier data with excessive dispersion can be removed first, and then the time series numerical sequences of the predicted output of each channel can be normalized or the pairwise ratios can be calculated to compare whether the numerical change magnitude, deviation direction and evolution curve of each channel prediction are highly consistent.

[0106] In step S513, if the comparison results show that the deviation prediction results output by multiple test channels of the battery of the same batch all exhibit the same deviation characteristics, such as each independent test channel predicting that the deviation will occur in the same direction and with the same magnitude in the future, statistically excluding the possibility that multiple independent test channels will experience the same hardware failure at the same time, it is determined that the data deviation is due to the electrical characteristic deviation of the battery of the same batch, that is, it is determined that there is an abnormality in the batch.

[0107] At this point, the measurement hardware of the test channel is determined to be functioning normally. The calibration and verification process is then terminated, no calibration command is issued to the test channel, and the test channel continues to perform normal testing. Once the batch of batteries has been tested and the test data has reached a steady state, the model's prediction results will automatically return to normal.

[0108] In step S514, it is indicated that only the test channel that initiated the calibration verification currently shows deviation characteristics, indicating that the deviation phenomenon is unrelated to the product batch, there is no batch abnormality, and interference from incoming material factors is excluded.

[0109] In another alternative embodiment, when performing batch screening, global operating condition analysis across different batches can also be taken into account to cope with large-scale environmental disturbances at the workshop level.

[0110] When acquiring test data and prediction results from multiple test channels, if multiple different test channels in the production line show the same test data fluctuation pattern and deviation trend within the same time period when testing products from multiple different incoming batches, it indicates that the anomaly is unrelated to hardware aging of a single test channel or the specificity of a single product batch. It is determined that the phenomenon is caused by common environmental interference or common equipment problems such as unstable workshop power supply voltage, sudden changes in ambient temperature and humidity, or changes in upstream global processes.

[0111] In this case, instead of implementing single-channel shielding and isolation, and directly issuing parameter calibration commands to individual test equipment, a global warning message is output to the Manufacturing Execution System (MES). After manual review of the production line utilities and environmental status, a decision is made on whether to perform a global parameter reset or batch scheduling calibration. This avoids blindly and isolatedly modifying individual parameters of normal test channels when a step change occurs in the global environment.

[0112] In one embodiment, step S52 determines whether there is a vehicle malfunction based on historical test data of the test vehicle carrying the battery under test, including: Step S521: Retrieve the historical test data of the test vehicle within a preset historical time window.

[0113] Step S522: Calculate the average vehicle deviation corresponding to the historical test data.

[0114] Step S523: Obtain the overall average deviation and standard deviation of all test vehicles within the preset historical time window.

[0115] Step S524: If the difference between the average deviation of the vehicle and the overall average deviation is greater than a preset outlier threshold, it is determined that there is an abnormality in the vehicle, the current calibration verification is terminated and the test channel continues to operate, wherein the preset outlier threshold is 3 times the standard deviation of the deviation.

[0116] Step S525: If the difference between the vehicle's average deviation and the overall average deviation is less than or equal to the preset outlier threshold, then it is determined that there is no vehicle anomaly.

[0117] In battery testing production lines, batteries under test are typically transported to the testing station in dedicated test carriers. These carriers are equipped with elastic probes or test grippers that contact the battery electrode tabs. Under prolonged mechanical insertion and removal and high-frequency current flow, the probe contact surfaces of the test carrier are prone to oxide deposition, mechanical wear, or probe elasticity attenuation, introducing additional contact resistance and causing measurement deviations in voltage or internal resistance. After determining the carrier identifier (such as a carrier barcode or carrier RFID tag) for the battery under test, step S521 retrieves all historical test data recorded from the production line database for the target test carrier during the testing of various batteries within a preset historical time window (e.g., a continuous operating period of the past 3 days).

[0118] Step S522 performs statistical analysis on all the measured values ​​of the target test vehicle retrieved within a preset historical time window, calculates the average deviation of each electrical measurement data relative to the nominal reference value or test median, and obtains the vehicle average deviation characterizing the inherent contact measurement characteristics of the target test vehicle itself.

[0119] To establish a unified tooling benchmark reference system, step S523 synchronously accesses the production line database and retrieves the test dataset of all test vehicles in circulation within the same preset historical time window. Normal statistical distribution calculations are performed on the test deviations of all test vehicles, and the overall average deviation (i.e., the expected value of the group benchmark) and the standard deviation reflecting the degree of deviation dispersion among the vehicles are calculated for all test vehicles on the production line.

[0120] Step S524 compares the average deviation of the target test vehicle with the overall average deviation of all test vehicles, and compares the difference with a preset outlier threshold. The preset outlier threshold is set to 3 times the standard deviation of the deviation, i.e., the 3-times-standard-deviation (3σ) outlier test based on the normal distribution.

[0121] When the difference between the two values ​​exceeds three times the standard deviation, it indicates that the contact deviation of the target test vehicle has statistically deviated significantly from the normal distribution of all vehicles, belonging to a typical outlier. This suggests that the probe contact impedance may have deteriorated, and the deviation in the measurement data is due to the vehicle's abnormality, rather than a measurement error in the test channel itself. Therefore, a vehicle abnormality is determined, the calibration and verification process for the test channel is terminated, the internal measurement parameters of the test channel are not adjusted, and the test channel continues to operate normally.

[0122] At the execution level, after determining that there is a carrier abnormality, a carrier removal control command can be generated to control the unloading mechanism connected to the production line to remove the target test carrier from the production flow line and transfer it to the maintenance station for probe cleaning or replacement, so as to prevent the abnormal carrier from continuing to flow into subsequent stations.

[0123] Step S525 indicates that the deviation performance of the target test vehicle is within the normal and controlled statistical fluctuation range of the entire tooling line, eliminating the interference of the vehicle probe contact impedance degradation, and determining that there is no abnormality in the vehicle. The control process continues to step S60 to perform the acquisition of historical calibration frequency and assessment of channel hardware health for the test channel itself.

[0124] This embodiment introduces a longitudinal history tracking mechanism for the vehicle based on statistical outlier tests, which separates vehicle anomalies from the drift of internal circuit parameters in the test channel, preventing the misjudgment of poor vehicle contact as machine misalignment and the execution of incorrect calibrations.

[0125] In one embodiment, after ruling out the batch anomalies and the vehicle anomalies, step S60 is performed to obtain the historical calibration frequency of the test channel.

[0126] Specifically, once it is confirmed that no batch anomalies have occurred in the batch test data associated with the battery under test, and the target test vehicle carrying the battery under test has not exceeded the preset outlier threshold, the deviation in the test data has been determined to be caused by the drift of the electrical parameters of the test channel itself. At this time, the identification information of the current test channel is obtained, and the maintenance and calibration records of the test channel within the preset backtracking time period are retrieved.

[0127] The preset backtracking time period can be configured according to the production line testing cycle, for example, configured as the past 1 month, the past 3 months, or the past 6 months. The total number of parameter calibrations performed on the test channel within the preset backtracking time period is counted, and the resulting value is used as the historical calibration frequency. By introducing the backtracking time period to perform sliding statistics on historical calibration actions, the evolution trend of the test channel in long-term operation can be determined.

[0128] After obtaining the historical calibration frequency, if the historical calibration frequency does not exceed the preset frequency threshold, step S70 is executed to send a calibration command to the test channel to calibrate the test channel.

[0129] Specifically, a pre-stored preset frequency threshold is retrieved, and the obtained historical calibration frequencies are compared with the preset frequency threshold. The preset frequency threshold represents the maximum number of recoverable drifts that the test channel can withstand within a set time period, for example, 2 times within 3 months.

[0130] When the comparison results show that the historical calibration frequency does not exceed the preset frequency threshold, for example, if calibration has been performed once within 3 months, it indicates that the current accuracy deviation of the test channel is a normal linear drift caused by fluctuations in environmental temperature and humidity or slight aging of components, which is within the compensable range of electrical parameters allowed by the hardware circuit design. A calibration command carrying calibration parameters is then generated and sent to the test channel or its corresponding channel controller to initiate the online calibration process for the test channel.

[0131] Upon receiving the calibration command, the test channel or channel controller performs specific parameter compensation and calibration actions. Specifically, the calibration command carries a correction compensation amount calculated based on the previously predicted deviation, and compensation is performed according to this correction compensation amount. Alternatively, the calibration command controls the test channel to disconnect from the current battery test circuit, switch to an internal standard reference source to acquire standard electrical signals, and calculate the actual error curve of the current channel by comparing it with the standard signal. Then, the calibration lookup table or polynomial fitting parameters are updated in the underlying driver program.

[0132] After completing the bias parameter update or reference calibration, the channel controller uses a standard reference source to perform a backtest self-check on the calibrated test channel. After confirming that the backtest deviation falls within the preset accuracy specification tolerance range, it returns a calibration success signal, enabling the test channel to return to the ready state to continue performing the test task of the battery under test. At the same time, it records the calibration timestamp and updated parameters in the database to complete the closed-loop management of channel status and calibration control.

[0133] In one embodiment, after obtaining the historical calibration frequency of the test channel in step S60, the method further includes: Step S80: If the historical calibration frequency exceeds the preset frequency threshold, then the calibration of the test channel is terminated, and the test channel is shielded and isolated.

[0134] If the comparison results show that the historical calibration frequency exceeds the preset frequency threshold, it indicates that the test channel has repeatedly experienced measurement inaccuracies within a short period of time. This phenomenon cannot be recovered by conventional software parameter bias compensation, indicating that the reference sampling chip, amplification circuit, or main circuit components inside the channel have suffered irreversible damage, breakdown, or poor soldering, or other hardware-level failures. In this state, continuing to issue calibration commands to forcibly modify the bias parameters will not only fail to maintain the long-term stability of subsequent tests but will also mask the true hardware damage and increase the risk of undetected faulty batteries.

[0135] Therefore, step S80 of this embodiment terminates the calibration process for the test channel and performs shielding isolation on the test channel. The shielding isolation is performed by setting the status of the test channel to a fault-disabled state in the production line test scheduling software, skipping the channel in subsequent task allocation to prevent new batteries under test from being assigned to the channel for testing, so that the test channel is completely removed from the test operation, pending later hardware maintenance.

[0136] In addition, the training process of the random forest model and long short-term memory network model mentioned above is explained below.

[0137] For the random forest classification model, historical electrical test datasets generated by historical battery tests are collected from the test channel. Using a preset number (e.g., 40 consecutive test data sets) as statistical units, the average, variance, range, and other multi-dimensional statistical features of the electrical test data within each sliding window are calculated as model inputs.

[0138] The sample data are labeled based on the actual operating performance of the test channel and the quality inspection records: if the electrical test data is stably distributed near the specification center line and there is no abnormal fluctuation, it is labeled as a normal state sample indicating that the channel is in normal condition; if the electrical test data deviates from the specification center line and fluctuates but does not produce test failures, indicating that the channel has potential accuracy drift, it is labeled as a warning state sample; if the electrical test data fluctuates greatly and is close to or exceeds the accuracy specification tolerance range, it is labeled as an out-of-tolerance state sample.

[0139] The constructed sample dataset is input into the random forest classification algorithm to complete the training, resulting in a random forest classification model that can output the corresponding state classification results in real time based on the input statistical features.

[0140] For the long short-term memory network model, the model input includes multi-dimensional parameters such as input current, output current, input voltage, output voltage, and internal resistance of the test channel in continuous time series. Similarly, the mean, variance, and range are calculated according to a fixed number of consecutive groups (e.g., 40 consecutive groups of data), and normalization is performed to construct the time series feature sequence.

[0141] In terms of time-series sample annotation, the calibration nodes in the history of the test channel are used as time anchors to backtrack and extract the statistical feature sequence of electrical test data in the preset time period (e.g., several weeks before calibration) before each calibration node.

[0142] If the calibration record indicates that the measurement accuracy of the test channel has indeed been adjusted or compensated, the corresponding time-series characteristic sequence before calibration will be marked as a positive sample indicating that the accuracy has deviated in the future time interval; if the calibration test shows that the accuracy of the test channel remains normal and no adjustment is required, it will be marked as a negative sample indicating that no accuracy has deviated.

[0143] The labeled time series sample set is input into the long short-term memory network model to complete iterative training, thereby constructing a long short-term memory network model for predicting the probability of deviation of the test channel from the accuracy specification tolerance range in the future preset period (e.g., the next 10 days, the next month).

[0144] After training, the random forest classification model and the long short-term memory network model are deployed in the prediction module of test control. During actual production line operation, the front-end random forest classification model calculates and monitors the discrete health status of the test channel in real time. Only when the status monitoring information repeatedly meets the preset early warning trigger conditions will the back-end call the long short-term memory network model to infer the probability of time series deviation in future time intervals. This achieves reasonable allocation of computing resources and proactive perception of the gradual accuracy drift of the test channel.

[0145] Compared with commonly used technologies, this embodiment has the following advantages: This application can trigger calibration verification in advance before accuracy deviation occurs, avoiding inaccurate test data caused by actual deviations in the test channel; at the same time, it checks and eliminates batch anomalies and carrier anomalies during calibration verification, and eliminates external interference such as poor contact between battery material characteristics and test carrier, avoiding miscalibration of normal test channels; moreover, it only issues calibration instructions when the historical calibration frequency does not exceed the preset frequency threshold, preventing continuous invalid calibration of frequently inaccurate test channels, and ensuring the accuracy and reliability of test channel calibration control.

[0146] In one embodiment, a calibration control device for a battery test channel is provided, such as... Figure 2 As shown. The calibration control device for this battery test channel includes the following modules, and the specific functions of each module are as follows: The feature extraction module is used to acquire electrical test data generated by the test channel for testing several batteries under test, and to extract the statistical features of the electrical test data.

[0147] The status monitoring module is used to classify the status based on the statistical characteristics in order to continuously monitor the channel status of the test channel and obtain status monitoring information.

[0148] The trend prediction module is used to predict the accuracy deviation trend of the test channel in the future time interval when the status monitoring information meets the preset early warning triggering conditions, and obtain the deviation prediction result.

[0149] The verification trigger module is used to trigger calibration verification when the deviation prediction result meets the preset calibration trigger conditions.

[0150] An anomaly detection module is used to perform anomaly detection for the test channel in response to the calibration verification.

[0151] The frequency acquisition module is used to acquire the historical calibration frequency of the test channel after excluding the external anomalies.

[0152] The calibration control module is used to send a calibration command to the test channel to calibrate the test channel when the historical calibration frequency does not exceed a preset frequency threshold.

[0153] It should be noted that for details not disclosed in the calibration control device for the battery test channel in this application embodiment, please refer to the details disclosed in the calibration control method for the battery test channel in this application embodiment.

[0154] Those skilled in the art will understand that the schematic diagram is merely an example of the calibration control device for the battery test channel and does not constitute a limitation on the terminal device of the calibration control device for the battery test channel. It may include more or fewer components than shown in the diagram, or combine certain components, or different components. For example, the calibration control device for the battery test channel may also include input / output devices, network access devices, buses, etc.

[0155] The calibration control device for the battery test channel may further include computing devices such as computers, laptops, handheld computers, and cloud servers, as well as, but not limited to, a processing module, a storage module, and a computer program stored in the storage module and executable on the processing module, such as the calibration control method program for the battery test channel described above. When the processing module executes the computer program, it implements the steps in the calibration control method embodiments of the various battery test channels described above, for example... Figure 1 The steps are shown.

[0156] In addition, this application also proposes an electronic device, which includes a storage module and a processing module. When the processing module executes the computer program, it can implement the steps in the above-mentioned calibration control method for the battery test channel, that is, implement the steps in any of the technical solutions of the above-mentioned calibration control method for the battery test channel.

[0157] The electronic device may be part of a calibration control unit integrated into the battery test channel, a local terminal device, or part of a cloud server.

[0158] The processing module can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processing module is the control center of the calibration control device for the battery testing channel, connecting various parts of the calibration control device throughout the battery testing channel using various interfaces and lines.

[0159] The storage module can be used to store the computer programs and modules. The processing module implements various functions of the calibration control device for the battery test channel by running or executing the computer programs and modules stored in the storage module and calling the data stored in the storage module. The storage module may mainly include a program storage area and a data storage area, wherein the program storage area may store the operating system, at least one application program required for a function, etc. In addition, the storage module may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, plug-in hard disk, at least one disk storage device, flash memory device or other non-volatile solid-state storage device.

[0160] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0161] The detailed descriptions listed above are merely specific descriptions of feasible implementation methods of this application, and are not intended to limit the scope of protection of this application. All equivalent implementation methods or modifications made without departing from the specific spirit of this application should be included within the scope of protection of this application.

Claims

1. A calibration control method for a battery test channel, characterized in that, Includes the following steps: Acquire electrical test data generated by the test channel for testing several batteries under test, and extract the statistical characteristics of the electrical test data; Based on the statistical characteristics, state classification is performed to continuously monitor the channel state of the test channel and obtain state monitoring information; When the status monitoring information meets the preset early warning triggering conditions, the accuracy deviation trend of the test channel in the future time interval is predicted to obtain the deviation prediction result. When the deviation prediction result meets the preset calibration trigger condition, calibration verification is triggered; In response to the calibration verification, troubleshooting for external anomalies related to the test channel is performed; After ruling out the external anomalies, obtain the historical calibration frequency of the test channel; When the historical calibration frequency does not exceed the preset frequency threshold, a calibration command is sent to the test channel to calibrate the test channel.

2. The calibration control method for the battery test channel according to claim 1, characterized in that, The extraction of statistical features from the electrical test data includes: The electrical parameters generated by the test channel testing a preset number of the tested batteries within a sliding data window are obtained. The electrical parameters include at least one of input current, output current, input voltage, output voltage, and internal resistance. Calculate at least one of the mean, variance, and range of each electrical parameter within the sliding data window, and use the calculation result as the statistical feature.

3. The calibration control method for the battery test channel according to claim 1, characterized in that, The process of classifying states based on the statistical characteristics to continuously monitor the channel state of the test channel, and obtaining state monitoring information, includes: The statistical features are classified in real time using a classification model, and the classification results are output to indicate whether the test channel is currently in a normal state, a warning state, or an out-of-tolerance state. The normal state indicates that the electrical test data has no abnormal fluctuations, the warning state indicates that the test channel has a potential accuracy drift trend and the test has no defects, and the out-of-tolerance state indicates that the test channel is close to or exceeds the accuracy specification tolerance range. The classification results are continuously recorded to generate the status monitoring information.

4. The calibration control method for the battery test channel according to claim 3, characterized in that, The classification model is a random forest classification model; The preset warning triggering condition is: all the statistical features extracted from the preset number of groups in the status monitoring information are classified as the warning status, wherein the preset number of groups is at least two groups; The calibration control method for the battery test channel further includes: When the classification model outputs the out-of-tolerance state, the probability of the out-of-tolerance state occurring in the preset number of analysis groups is statistically analyzed. Based on the probability range in which the occurrence probability falls, a corresponding graded handling strategy is executed for the test channel. The graded handling strategy includes at least one of logging, outputting early warning signals, and executing channel blocking.

5. The calibration control method for the battery test channel according to claim 1 or 4, characterized in that, When the status monitoring information meets the preset early warning triggering conditions, the accuracy deviation trend of the test channel in the future time interval is predicted to obtain the deviation prediction result, including: In response to the state monitoring information satisfying the preset early warning triggering condition, multiple statistical features within the sliding time window are input into the long short-term memory network model; The long short-term memory network model is used to calculate the probability of the test channel's accuracy deviating from the accuracy specification tolerance range within a future preset period, and the deviation probability is used as the deviation prediction result.

6. The calibration control method for the battery test channel according to claim 5, characterized in that, When the deviation prediction result meets the preset calibration trigger condition, the calibration verification is triggered, including: Determine whether the deviation probability in the deviation prediction result is greater than a preset probability threshold; If the deviation probability is greater than the preset probability threshold, then the preset calibration trigger condition is met, and calibration verification is triggered. If the deviation probability is less than or equal to the preset probability threshold, then new statistical features are obtained by scrolling through the sliding time window, the data input to the long short-term memory network model is updated, and the process returns to the step of calculating the deviation probability.

7. The calibration control method for the battery test channel according to claim 1, characterized in that, The external anomaly includes at least one of the batch anomaly corresponding to the plurality of batteries under test and the vehicle anomaly corresponding to the test vehicle carrying the plurality of batteries under test. The process of troubleshooting external anomalies for the test channel includes: Determine whether there are any batch anomalies based on the batch test data associated with the battery under test; If no batch anomaly is found, then a determination of whether a vehicle anomaly exists is made based on the historical test data of the test vehicle carrying the battery under test.

8. The calibration control method for the battery test channel according to claim 7, characterized in that, The determination of whether there is a batch anomaly based on the batch test data associated with the tested battery includes: Obtain the deviation prediction results corresponding to the testing of the batches of the several tested batteries by multiple test channels; Compare the deviation prediction results for each of the test channels: If the deviation prediction results of multiple test channels all show the same deviation characteristics, it is determined that there is an abnormality in the batch, the current calibration verification is terminated and the test channel continues to operate. If only the deviation prediction result of the current test channel shows deviation characteristics, then it is determined that there is no batch anomaly; The determination of whether there is a vehicle anomaly based on the historical test data of the test vehicle carrying the battery under test includes: Retrieve historical test data of the test vehicle within a preset historical time window; Calculate the average vehicle deviation corresponding to the historical test data; Obtain the overall average deviation and standard deviation of all test vehicles within the preset historical time window; If the difference between the average deviation of the vehicle and the overall average deviation is greater than a preset outlier threshold, it is determined that there is an abnormality in the vehicle, the current calibration verification is terminated and the test channel continues to operate, wherein the preset outlier threshold is 3 times the standard deviation of the deviation; If the difference between the vehicle's average deviation and the overall average deviation is less than or equal to the preset outlier threshold, then it is determined that there is no vehicle anomaly.

9. The calibration control method for the battery test channel according to claim 1, characterized in that, After obtaining the historical calibration frequency of the test channel, the method further includes: If the historical calibration frequency exceeds the preset frequency threshold, the calibration of the test channel is terminated, and the test channel is shielded and isolated.

10. A calibration control device for a battery test channel, characterized in that, include: The feature extraction module is used to acquire electrical test data generated by the test channel for testing several batteries under test, and to extract the statistical features of the electrical test data; The status monitoring module is used to classify the status based on the statistical characteristics to continuously monitor the channel status of the test channel and obtain status monitoring information. The trend prediction module is used to predict the accuracy deviation trend of the test channel in a future time interval when the status monitoring information meets the preset early warning triggering conditions, and obtain the deviation prediction result. The verification trigger module is used to trigger calibration verification when the deviation prediction result meets the preset calibration trigger conditions; An anomaly detection module is used to perform anomaly detection for the test channel in response to the calibration verification. The frequency acquisition module is used to acquire the historical calibration frequency of the test channel after excluding the external anomalies. The calibration control module is used to send a calibration command to the test channel to calibrate the test channel when the historical calibration frequency does not exceed a preset frequency threshold.

Images