A lithium battery remote monitoring and early warning method and system suitable for battery replacement mode

Abstract

The application relates to the technical field of lithium battery monitoring, and particularly discloses a lithium battery remote monitoring and early warning method and system suitable for a battery replacement mode, the method comprising the following steps: in response to a potential abnormal signal of a lithium battery collected, triggering a first-level early warning of the lithium battery; in response to the first-level early warning, fitting internal defect parameters of the lithium battery through a battery simulation model to calculate an internal risk value; and combining an external abnormal estimation value and the internal risk value to calculate a comprehensive risk value. By introducing a multi-level early warning mechanism and multi-dimensional consistency checking logic, the application can accurately eliminate scene interference and noise false alarms caused by complex working conditions, and ensure the authenticity and reliability of early warning triggering; meanwhile, the application breaks through the traditional mode and constructs a weight dynamic reconstruction mechanism under a thermal inertia lag state, and when external sensor data distortion is sensed, the application actively shifts the risk decision weight to an internal battery simulation model.

Description

Technical Field

[0001] This invention relates to the field of lithium battery monitoring technology, specifically to a method and system for remote monitoring and early warning of lithium batteries suitable for battery swapping modes. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the battery swapping model, which separates the vehicle and the battery, has been widely used due to its high energy replenishment efficiency. In this model, lithium batteries, in the form of battery pack asset pools, are frequently transferred between different vehicles and swapping stations, resulting in complex and variable operating conditions. To ensure the safe operation of the battery swapping network, the industry generally uses remote monitoring systems to collect external characterization data of the batteries in real time, and combines this with preset static thresholds for abnormal alarms and fault handling.

[0003] However, existing monitoring and early warning mechanisms are insufficient to cope with the multi-dimensional and complex challenges brought about by the high-frequency circulation under the battery swapping model. In particular, traditional monitoring and early warning methods are prone to failure under extreme working conditions and limited physical resources.

[0004] Specifically, existing technologies face two major technical challenges in practical applications:

[0005] First, there is a conflict between static warning weights and data source confidence distortion under extreme operating conditions. After a battery undergoes intense high-rate charging and discharging, due to the thermal inertia hysteresis effect of the battery pack's physical structure, the characterization data fed back by external sensors often suffers from severe time lag. At this point, the battery has actually accumulated a very high risk of thermal runaway. If statically configured internal and external risk weights are still used, the distorted external low-risk data will dilute the true internal high-risk value, thereby causing fatal safety misses.

[0006] Second, there is a conflict between the system's absolute safety isolation command and the rigid saturation of the limited physical isolation resources at the battery swapping station. When the monitoring system detects a real fault of extremely high risk level and issues a mandatory isolation command according to regulations, if the explosion-proof compartments and other special physical isolation resources in the battery swapping station are already full of historically faulty batteries, the robotic arm will face a physical scheduling deadlock due to the lack of safe compartments to place the batteries. Forcibly placing high-risk batteries back into regular compartments could easily trigger catastrophic consequences such as a chain fire. Summary of the Invention

[0007] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a method and system for remote monitoring and early warning of lithium batteries suitable for battery swapping, to ensure the authenticity and reliability of early warning triggering.

[0008] To achieve the above objectives, a first aspect of the present invention proposes a method for remote monitoring and early warning of lithium batteries suitable for battery swapping, comprising the following steps:

[0009] In response to the collected potential abnormal signals of the lithium battery, a first-level warning for the lithium battery is triggered;

[0010] In response to the first-level warning, the internal defect parameters of the lithium battery are fitted by a battery simulation model to calculate the internal risk value;

[0011] The comprehensive risk value is calculated by combining the external anomaly estimate and the internal risk value.

[0012] If the overall risk value reaches the second-level warning condition, then a consistency check is performed on the potential abnormal signal;

[0013] If the consistency check passes, it is determined to be a genuine anomaly. Based on the comprehensive risk value, the risk level of the lithium battery is determined, and the corresponding graded handling is performed.

[0014] To achieve the above objectives, a second aspect of the present invention provides a remote monitoring and early warning system for lithium batteries suitable for battery swapping, the system comprising:

[0015] The first-level early warning module is used to trigger a first-level early warning for the lithium battery in response to the collected potential abnormal signals of the lithium battery.

[0016] An internal risk calculation module is used to calculate the internal risk value by fitting the internal defect parameters of the lithium battery through a battery simulation model in response to the first-level warning.

[0017] The comprehensive risk calculation module is used to calculate the comprehensive risk value by combining the external anomaly estimate and the internal risk value;

[0018] The consistency verification module is used to perform consistency verification on the potential abnormal signal when the comprehensive risk value reaches the second-level warning condition;

[0019] The anomaly detection and handling module is used to determine a genuine anomaly when the consistency verification passes, determine the risk level of the lithium battery based on the comprehensive risk value, and perform corresponding graded handling.

[0020] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described method for remote monitoring and early warning of lithium batteries suitable for battery swapping modes.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] The lithium battery remote monitoring and early warning method and system applicable to battery swapping mode of this invention introduces a multi-level early warning mechanism and multi-dimensional consistency verification logic, which can accurately eliminate scenario-based interference and noise false alarms caused by complex operating conditions, and ensure the authenticity and reliability of early warning triggering. At the same time, this solution breaks through by constructing a weight dynamic reconstruction mechanism under thermal inertial lag. When external sensor data distortion is detected, the risk decision weight is actively transferred to the internal battery simulation model, which effectively overcomes the serious hidden danger of missed reports caused by external data lag under extreme operating conditions.

[0023] Furthermore, in the face of extreme scenarios where the physical isolation capacity of battery swapping stations reaches its limit, this solution designs a fallback mechanism based on dynamic risk gradient-based physical resource preemption scheduling and on-site proactive energy depletion and risk reduction in conventional storage locations. This effectively resolves the scheduling deadlock conflict between absolute network security commands and physical space capacity saturation, ensuring efficient circulation of battery assets while greatly improving the system robustness and safety baseline of the battery swapping network under extreme boundary conditions. Attached Figure Description

[0024] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0025] Figure 1 This is a flowchart illustrating the remote monitoring and early warning method for lithium batteries applicable to battery swapping modes provided by the present invention.

[0026] Figure 2 This is a schematic diagram showing the comparison of electrical performance data before and after pulse median denoising and Kalman filtering in the remote monitoring and early warning method for lithium batteries suitable for battery swapping provided by the present invention.

[0027] Figure 3 This is a schematic diagram of the convergence trajectory of internal defect parameters and internal risk values ​​of the battery based on the particle filter algorithm in the remote monitoring and early warning method for lithium batteries applicable to battery swapping mode provided by the present invention.

[0028] Figure 4 This is a schematic diagram of the matching path between the temporal evolution trajectory of potential abnormal signals and the fault precursor feature template based on the Dynamic Time Warping (DTW) algorithm in the lithium battery remote monitoring and early warning method applicable to battery swapping mode provided by the present invention.

[0029] Figure 5 This is a schematic diagram showing the dual Y-axis comparison of the surge in cumulative stress factor and the hysteresis divergence of external temperature change rate under extreme working conditions in the remote monitoring and early warning method for lithium batteries applicable to battery swapping mode provided by the present invention.

[0030] Figure 6 This is a cross-comparison verification curve of the exponential decay of the residual risk assessment value of historical faulty batteries and the current acute battery risk value in the lithium battery remote monitoring and early warning method applicable to battery swapping mode provided by the present invention.

[0031] Figure 7 This is a schematic diagram illustrating the implementation of the lithium battery remote monitoring and early warning system for battery swapping mode provided by the present invention.

[0032] Figure 8 This is a schematic diagram of the electronic device provided by the present invention. Detailed Implementation

[0033] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0034] The following description, with reference to the accompanying drawings, illustrates a method, system, and electronic device for remote monitoring and early warning of lithium batteries suitable for battery swapping modes, according to embodiments of the present invention.

[0035] Example 1:

[0036] In the battery swapping model, massive amounts of lithium batteries are transferred frequently between different new energy vehicles and distributed battery swapping stations in the form of battery pack asset pools. This operating model results in extremely complex operating conditions for lithium batteries, and there are significant differences in data structure between batteries from different batches and suppliers.

[0037] To achieve absolute security control over such dynamic assets, the method provided in this embodiment relies on a cloud-edge collaborative architecture, which includes edge computing nodes deployed at the edge of the battery swapping station and a central monitoring platform deployed in the cloud.

[0038] like Figure 1 As shown in the figure, the remote monitoring and early warning method for lithium batteries suitable for battery swapping mode provided in this embodiment specifically includes the following steps:

[0039] Step 1: Collection and standardization preprocessing of multi-source heterogeneous data.

[0040] The foundation of the above method lies in constructing a high-quality data base. In battery swapping networks, the capture of potential anomaly signals highly depends on the accuracy of the underlying data, which is extracted based on standardized time-series data. To obtain this data, the system first performs the following steps: collecting the electrical performance data of the lithium battery. The aforementioned electrical performance data covers the physical and electrochemical characterization quantities of the lithium battery at various levels, including total pack voltage, total pack current, highest single-cell voltage, lowest single-cell voltage, highest single-cell temperature, lowest single-cell temperature, and insulation resistance values ​​and contactor status flags reported by the battery management system.

[0041] It is important to note that because battery swapping networks are compatible with various vehicle bodies and battery packs, the data sources received by the system are highly heterogeneous. Therefore, after data collection, the system needs to match the corresponding battery management system feature fingerprint based on the battery pack identifier. The aforementioned battery pack identifier refers to the unique vehicle identification number or battery asset code written into the battery pack's physical read-only memory. The aforementioned battery management system feature fingerprint is a structured configuration file pre-stored in a cloud-based feature library. This fingerprint defines in detail the static parameter table and dynamic drift benchmark for a specific supplier and batch of batteries.

[0042] After a unique feature fingerprint is matched, the system performs data fusion and noise suppression on the electrical performance data based on the parameter mapping rules and noise type in the feature fingerprint of the battery management system to obtain the standardized time-series data.

[0043] The aforementioned parameter mapping rules specify the method for unifying the dimensions of heterogeneous data sources, the method for aligning sampling frequencies, and the logic for recombining byte order, thereby parsing the messy data from different manufacturers into a data matrix with a unified timestamp and a unified data type.

[0044] The aforementioned noise types pre-summarize the inherent error patterns of the corresponding battery management system hardware sampling circuits, covering Gaussian white noise, pulse spike noise caused by plugging and unplugging relays, and low-frequency baseline noise caused by temperature drift.

[0045] To address different noise types, the system employs different pure textual mathematical filtering algorithms for noise suppression. For impulse spike noise, the system uses a median filtering algorithm to extract the electrical performance data sequence within a preset sliding time window, sort it by amplitude, and take the value at the middle position of the sequence as the effective output at the center of the window, thereby filtering out extremely large or small abnormal impulses. For Gaussian white noise, the system uses a Kalman filtering algorithm to predict the prior state estimate at the current time using the optimal state estimate from the previous time step, calculate the prior estimate covariance, and then combine the actual observation value at the current time with the Kalman gain to calculate the optimal posterior state estimate at the current time step. This smooths the electrical performance data sequence, ultimately outputting high-quality standardized time-series data.

[0046] like Figure 2 This diagram illustrates the comparison between electrical performance data before and after pulse median denoising and Kalman filtering. In this diagram, the horizontal axis represents the continuous sampling time during battery operation, and the vertical axis represents the individual cell voltage values ​​collected by the underlying hardware network.

[0047] In the figure, the light-colored, volatile curves represent the raw, noisy voltage data without any processing, while the dark-colored, smooth curves represent the voltage data processed by the personalized noise self-suppression method provided by this invention.

[0048] As can be seen from the specific scheme of this embodiment, in the waveform transformation of the original noisy voltage data, in addition to the baseline noise with high frequency and dense distribution, there are also multiple severe pulse spikes that deviate from the true value by a few tenths of a volt at an instant. These spikes are usually caused by electrical transient impacts caused by the mechanical arm plugging and unplugging relays in the battery swapping network.

[0049] If such data containing spikes is directly input into the monitoring system, it is easy to cause model oscillations and false alarms. Through the system's built-in preprocessing engine, the median filtering algorithm is first used to extract and sort the values ​​within the set sliding time window, and to remove abrupt jump pulses that are extremely large or small. Then, the Kalman filtering algorithm is used to deeply smooth the residual baseline noise by calculating the prior estimate of the current state and the optimal posterior estimate of the state.

[0050] As can be observed from the trend of the dark curve in the figure, after waveform transformation processing by the above-mentioned hybrid filtering algorithm, the output filtered voltage data better restores the physical characteristics of the gradual decrease during the electrochemical discharge of lithium batteries, effectively filtering out discrete noise interference from the environment and hardware devices. This noise reduction and smoothing effect indicates that the standardized time-series data generated by this invention has high purity, providing a favorable data foundation for subsequent fitting of internal defect parameters in digital twin models.

[0051] Step 2: Triggering potential abnormal signals and fitting internal defect parameters.

[0052] After acquiring standardized time-series data, the system's anomaly detection engine, deployed in the cloud or at the edge, monitors this data in real time. In response to potential anomaly signals from the collected lithium batteries, a first-level warning is triggered. These potential anomaly signals refer to minor deviations in the standardized time-series data that exceed a preset health baseline but are insufficient to trigger hard power-off protection under traditional fixed-threshold alarm mechanisms. These deviations include a slow increase in individual cell differential pressure, slight abrupt changes in temperature rise rate, and a non-linear increase in charge / discharge internal resistance. This first-level warning is the system's initial screening alert state. Lithium batteries in this state will be marked as key monitoring targets by the backend, and the sampling and reporting frequency of their corresponding data will be increased.

[0053] The preset health baseline is different from the fixed hardware alarm dead zone. It is the expected normal characteristic range dynamically mapped by the system based on the current ambient temperature and state of charge of the battery, so that slight deviation characteristics can be captured before the hard power-off protection conditions are reached.

[0054] It is important to note that the first-level warning only indicates anomalies in externally observed data. To investigate the underlying cause, the system responds to the first-level warning by fitting the internal defect parameters of the lithium battery using a battery simulation model to calculate the internal risk value. The aforementioned battery simulation model is a mechanism- and data-driven digital twin model pre-trained based on extensive full-lifecycle aging experimental data of the same battery model. This model integrates equivalent circuit equations and simplified pseudo-two-dimensional electrochemical equations. The aforementioned internal defect parameters refer to electrochemical degradation indices that cannot be directly measured by external sensors, encompassing the degree of solid electrolyte interface film thickness, negative electrode lithium plating, active material loss ratio, and micro-internal short-circuit equivalent resistance.

[0055] Optionally, the process of fitting internal defect parameters relies on particle filtering or recursive least squares algorithms. The system uses the current and temperature from standardized time-series data as input excitations to the battery simulation model, and compares the residual between the model's output terminal voltage and the actual terminal voltage collected in the standardized time-series data. This residual is used to correct the hidden state variables within the model, iterating until the residual converges. At this point, the state variables within the model represent the actual internal defect parameters of the lithium battery. The system further maps these internal defect parameters to a preset hazard function and outputs the aforementioned internal risk value through a weighted summation. This internal risk value is a normalized continuous value; a higher value indicates a greater probability of thermal runaway caused by internal battery structural instability.

[0056] For example, when using the particle filter algorithm, the objective function is explicitly defined:

[0057] Let the actual collected terminal voltage be... The model predicts the terminal voltage as follows: The measurement noise variance is The system calculates the weights of each particle based on the Gaussian likelihood function. The formula is:

[0058] ;

[0059] For example, the residual convergence criterion is set as follows: a dual constraint of root mean square error (RMSE) and the number of iterations is adopted. That is, within the sliding observation window, the continuous... The RMSE of the terminal voltage residual in the next iteration. If this RMSE is less than a preset residual accuracy threshold (e.g., ... If the number of iterations reaches the set upper limit (e.g., 200 times) and is still not lower than the accuracy threshold, then the current best state posterior estimate is forcibly output and a low confidence label is attached.

[0060] like Figure 3 This diagram illustrates the convergence trajectory of battery internal defect parameters and internal risk values ​​based on the particle filter algorithm for backfitting. In this diagram, the horizontal axis represents the number of model iterations performed by the digital twin simulation model in the cloud computing container using particle filter backfitting; the left vertical axis represents the simulated voltage residual between the model output voltage and the actual acquired voltage; and the right vertical axis represents the normalized internal risk value obtained through mapping calculation.

[0061] The figure shows two different colored curves with opposite trends. The light-colored curve represents the terminal voltage residual convergence curve, while the dark-colored curve represents the internal risk value assessment trajectory.

[0062] As can be seen from the specific scheme of this embodiment, in the early stage of anomaly diagnosis, since the internal hidden state variables have not been effectively corrected, the light-colored terminal voltage residual convergence curve is initially at a high level of about 0.5, while the dark-colored internal risk value assessment trajectory, which represents the global danger level, is in an initial state of about 0.3.

[0063] As the particle filter algorithm continuously uses actual observation data to correct the internal defect parameters, the waveform of the terminal voltage residual convergence curve gradually exhibits an exponential smooth decay. After about 100 iterations, the residual waveform tends to stabilize and converge to a small fluctuation range close to 0.

[0064] Meanwhile, after experiencing initial random oscillations and particle resampling, the internal risk value assessment trajectory steadily increased as the residuals converged, eventually stabilizing in the high-risk quantification range of around 0.88.

[0065] The waveform transformation process of this dynamic curve demonstrates that the battery simulation model used in this invention can resolve the degree of change in electrochemical structures such as micro-short circuits within the iteration cycle, thereby quantifying the internal risk probability and providing an objective decision-making basis for triggering accurate early warnings.

[0066] Step 3: Quantification and assessment of the overall risk value.

[0067] After obtaining the internal risk value, the system combines the external anomaly estimate with the internal risk value to calculate the comprehensive risk value. The aforementioned external anomaly estimate is a statistical measure determined by the magnitude of the difference, duration of deviation, and rate of change of the deviation in the standardized time-series data representing parameters exceeding the safety baseline. Under extreme operating conditions in the battery swapping mode, relying solely on internal calculations or external observations has limitations; therefore, a fusion of risk perspectives is necessary.

[0068] For example, the external anomaly estimate The calculation formula is as follows:

[0069] ;

[0070] In the formula, This is an estimate of external anomalies; To standardize the magnitude of the difference between the characterization parameters in time series data and the preset health baseline. Duration of this deviation Deviation rate of change Set weight coefficients for the corresponding dimensions, and satisfy the following conditions: This is a linear normalization function used to eliminate dimensional differences, making the final output... It lies within the continuous interval [0, 1].

[0071] Optionally, the process of calculating the comprehensive risk value by combining the external anomaly estimate and the internal risk value specifically includes the following steps:

[0072] First, the external risk coefficient and internal risk coefficient corresponding to the battery type of the lithium battery are obtained. The aforementioned battery types cover different material systems such as lithium iron phosphate batteries, ternary nickel-cobalt-manganese batteries, and lithium titanate batteries. Different battery types have drastically different sensitivities to external stress and internal aging. For lithium iron phosphate batteries, which have good thermal stability but a flat voltage plateau, the system assigns a smaller external risk coefficient and a larger internal risk coefficient; for ternary lithium batteries, which have poor thermal stability, the system assigns a larger external risk coefficient.

[0073] Subsequently, the system sums the product of the external anomaly estimate and the external risk coefficient, and the product of the internal risk value and the internal risk coefficient, to obtain the total risk assessment value.

[0074] Finally, the system divides the total risk assessment value by the sum of the external risk coefficient and the internal risk coefficient to obtain the comprehensive risk value.

[0075] For example, the calculation process of the comprehensive risk value can be expressed as follows:

[0076] ;

[0077] In the formula, The comprehensive risk value represents the global quantitative probability of a thermal runaway event occurring in a lithium battery at the current moment. This is an external anomaly estimate, which characterizes the degree to which external physical parameters directly obtained from sensors deviate from the normal state; The internal risk value represents the degree of collapse of the battery's internal electrochemical structure as determined by the digital twin model. The external risk coefficient determines the weight of external representation data in the overall risk assessment. The internal risk coefficient determines the weight of internal mechanism data in the overall risk assessment.

[0078] Step 4: Second-level early warning triggering and multi-dimensional consistency verification.

[0079] After calculating the comprehensive risk value in real time, the system compares it with a preset tiered threshold. If the comprehensive risk value reaches the second-level warning condition, a consistency check is performed on the potential abnormal signal. The aforementioned second-level warning condition means that the comprehensive risk value exceeds the system's calibrated safety warning line, indicating that the battery has already posed a substantial safety threat. At this time, the system does not directly trigger a power outage or shutdown, but instead enters a rigorous logical self-verification process.

[0080] The safety warning line includes an initially set static default red line, and the system can subsequently dynamically adjust the warning line within an elastic threshold range consisting of the upper and lower tolerance limits based on the real-time physical inventory and other operational pressure status of the battery swapping station.

[0081] The aforementioned consistency verification includes parameter consistency verification, scenario consistency verification, and evolutionary consistency verification. This consistency verification aims to completely eliminate false positives caused by single-point sensor failures, communication packet loss, or normal extreme driving conditions from three independent dimensions: horizontal multi-parameter coupling relationships, vertical historical time series patterns, and external environmental interference.

[0082] For example, a consistency check is performed on the potential anomalous signal, specifically encompassing the following three parallel check steps:

[0083] In the first step, if the changing trends of multiple parameters of the lithium battery are consistent, the parameter consistency check is passed. Since a battery is a highly nonlinear electrochemical coupling system, a real fault will inevitably trigger a multi-dimensional chain reaction. The system extracts the voltage, temperature, and internal resistance sequences of the abnormal individual cells in real time, constructs a multi-dimensional parameter vector matrix, and calculates the Pearson correlation coefficient between each column vector. When an abnormal voltage drop occurs, if it is accompanied by an abnormal rise in the temperature of the individual cell and a sharp increase in the internal resistance at the corresponding location, and the eigenvalues ​​of the correlation coefficient matrix satisfy the preset consistency matrix criterion, the system determines that the anomaly conforms to the physical correlation law, thus passing the parameter consistency check.

[0084] In the second step, if the potential abnormal signals exclude the preset scenario-based action interference, the scenario consistency check is deemed passed. In battery swapping mode, there are numerous compliant scenario actions that can cause drastic data fluctuations. The aforementioned preset scenario-based action interference includes electrical transient shocks when the robotic arm inserts or removes the battery pack within the battery swapping station, continuous high-power kinetic energy recovery charging when the vehicle descends a long slope, and instantaneous high-rate discharge when the driver accelerates rapidly. The system determines whether any of the above actions have occurred at the current moment by analyzing the message commands reported by the vehicle controller or the process logs of the battery swapping station's main control computer.

[0085] Only when the occurrence time of a potential abnormal signal has no logical overlap with the occurrence time of all scenario-based actions, that is, when the abnormal signal is spontaneously generated in an interference-free state such as when the battery is at rest or undergoing slow charging and discharging, will the system determine that it has passed the scenario consistency check.

[0086] In the third step, if the temporal evolution trajectory of the potential abnormal signal matches a preset fault precursor evolution feature library with a degree greater than or equal to a matching threshold, the evolution consistency check is passed. Real battery thermal runaway or micro-short circuits are not instantaneous; there is always a degradation path from quantitative to qualitative change. The aforementioned fault precursor evolution feature library is a standard pathological curve library extracted from the entire lifecycle data of hundreds of thousands of real failed batteries in the cloud. The system extracts historical standardized time-series data within 72 hours prior to the occurrence of the potential abnormal signal to construct the aforementioned temporal evolution trajectory. The system uses a dynamic time warping algorithm to calculate the morphological distance between this temporal evolution trajectory and various standard pathological curves in the feature library, converting the distance value into a percentage matching degree. When the matching degree is greater than or equal to the set matching threshold of 85%, the abnormal signal is confirmed to have typical pathological developmental genes.

[0087] Optionally, the 85% matching threshold is not an empirical preset, but is determined by the system through mathematical statistical analysis of the entire lifecycle dataset of hundreds of thousands of historically known real-world faulty batteries in the cloud. Specifically, the system performs massive dynamic time warping (DTW) matching calculations between the temporal evolution trajectory of known faulty batteries before the outbreak and the fault precursor feature database, and plots the matching probability density distribution curve. The system selects the matching degree value corresponding to the optimal intersection point (e.g., the 95% confidence interval quantile) that balances the false alarm rate and the false alarm rate as the red line threshold, thereby ensuring that typical pathological development genes are accurately identified mathematically.

[0088] like Figure 4 This figure illustrates the matching path between the temporal evolution trajectory of a potential abnormal signal and the feature template of a fault precursor, based on a dynamic time warping algorithm. In this figure, the horizontal axis represents the historical sampling time within 72 hours prior to the occurrence of the potential abnormal signal, and the vertical axis represents the quantized amplitude of the battery's internal feature parameters.

[0089] The dark, smooth curve at the top of the figure represents the fault precursor feature template preset in the cloud-based historical failure medical record database, while the light-colored curve with local fluctuations at the bottom represents the time-series evolution trajectory of potential abnormal signals captured in real time and preprocessed. The light-colored dashed line connecting these two main curves represents the optimal time warping matching path sought by the dynamic time warping algorithm when calculating morphological distance.

[0090] As can be seen from the specific scheme of this embodiment, due to the randomness of the charging and discharging conditions of batteries in actual battery swapping networks, the abnormal evolution trajectory collected in real time often exhibits local stretching, compression, or temporal misalignment compared to the standard pathological curve. If traditional direct comparison based on spatial distance is used, it is easy to draw biased conclusions due to the aforementioned temporal misalignment.

[0091] By introducing a dynamic time warping algorithm, the system allows for non-linear, elastic comparisons on the time axis. As can be observed from the distribution of the light-colored dashed lines in the attached figure, even though the peaks and troughs of the two waveforms are not strictly aligned on the horizontal time axis, the algorithm can still capture the correlation of the waveform evolution trend and perform a one-to-one cross-temporal oblique connection mapping of similar morphological feature points.

[0092] This evolutionary trajectory matching mechanism based on morphological distance calculation helps overcome temporal misalignment interference caused by complex operating conditions. Ultimately, the calculated morphological distance is converted into a percentage matching degree. When this matching degree is greater than or equal to a set threshold of 85%, the system can confirm that the abnormal signal has pathological development characteristics, thus completing the evolutionary consistency verification. It should be noted that the 85% threshold is a purely logical similarity score threshold evaluation indicator and is not included in the... Figure 4 It is represented by lines.

[0093] It is also important to note that the above three checks are strictly logically ANDed. The consistency check is considered passed when all three checks—parameter consistency, scene consistency, and evolution consistency—pass. If any check fails, the system will label the potential abnormal signal as a false anomaly or environmental noise, terminate the current warning process, and archive it in the feature library for subsequent training of the filtering algorithm, ensuring that no healthy battery asset is mistakenly affected.

[0094] Step 5: Risk classification, adaptive threshold adjustment and graded treatment.

[0095] If the consistency verification passes, it is determined to be a genuine anomaly. Based on the comprehensive risk value, the risk level of the lithium battery is determined, and corresponding graded handling is performed. The aforementioned risk level refers to the system classifying the battery's dangerous state into multiple tiers based on the magnitude of the comprehensive risk value, covering mild consistency degradation, moderate self-discharge aggravation, severe internal short circuit initiation, and critical thermal runaway.

[0096] In the battery swapping business model, which pursues extreme efficiency and asset turnover, the disposal method cannot be a rigid one-size-fits-all approach. The system's execution of corresponding tiered disposal is deeply tied to real-world operational pressures. This tiered disposal process first includes obtaining the current operational status information of the battery swapping station, which includes the battery swapping time period and the battery inventory quantity. The aforementioned battery swapping time periods cover morning and evening rush hours, peak refueling periods for logistics vehicles, and off-peak hours at midnight. The aforementioned battery inventory quantity refers to the total number of fully charged and fault-free dispatchable battery packs currently in the battery swapping station's storage area.

[0097] Subsequently, the system determines whether the battery swapping station is in a high-pressure or low-pressure state based on the battery swapping period and the battery inventory quantity. A high-pressure state is defined as the current period being a peak battery swapping period when the inventory of fully charged batteries is below the minimum turnover threshold to ensure uninterrupted battery swapping services. A low-pressure state is defined as the current period being a low-pressure battery swapping period when the inventory of fully charged batteries is sufficient.

[0098] It is also important to note that under extreme operating conditions, the system must dynamically and flexibly balance absolute safety with guaranteed supply. When the risk level is at a preset low-to-medium risk level, the system acquires a corresponding elastic threshold range. The aforementioned low-to-medium risk level typically corresponds to mild uniform degradation or minor capacity decay, posing no immediate safety threat of ignition or explosion. The aforementioned elastic threshold range refers to the upper and lower limits of the overall risk tolerance value that the system allows the battery to continue operating with defects and current limiting.

[0099] If the battery swapping station is under high pressure, to avoid large-scale vehicle breakdowns and queues that could lead to mass incidents, the system adjusts the warning threshold to the upper limit of the elastic threshold range. By increasing the tolerance, batteries with minor defects are allowed to continue swapping, maximizing the guarantee of transportation capacity.

[0100] Conversely, if the battery swapping station is in a low-pressure state, where resources are abundant and the system prioritizes the lifespan and absolute safety of battery assets, the warning threshold will be adjusted to the lower limit of the elastic threshold range. By tightening the tolerance, potentially hazardous batteries can be intercepted within the battery swapping station in advance.

[0101] Finally, the system executes the tiered response based on the adjusted warning threshold. The aforementioned tiered response covers issuing a power reduction operation command and requiring the vehicle to travel at a limited speed for low-risk vehicles, locking the battery to prevent it from being transferred across stations and allowing it to be slowly charged within the station for medium-risk vehicles, and triggering the station's robotic arm to forcibly grab the battery and place it in a special explosion-proof isolation tank within the station for immersion physical isolation for high-risk vehicles.

[0102] Step Six: Step-by-step verification and upstream tracing after handling.

[0103] After detaining suspected faulty batteries at the battery swapping station and implementing initial control measures, to prevent permanent misjudgment of high-quality assets, following the corresponding tiered disposal, a verification process for the lithium batteries is triggered. This verification process involves the battery swapping station utilizing its high-precision bidirectional charge / discharge inverter equipment and on-site simulated load cabinet to perform in-depth offline and offline testing on the isolated batteries. The tests include constant current / constant voltage full charge testing, high-current deep discharge testing, and step pulse internal resistance AC impedance spectroscopy scanning testing.

[0104] If the verification process shows no real fault, meaning the battery's electrochemical indicators are perfectly restored during the offline in-depth inspection, it proves that the earlier warning was a phantom alarm triggered by an extremely rare, sporadic communication disorder or software deadlock. In this case, the system will deactivate the warning for the potential abnormal signal, resume the lithium battery's operation, re-mark its status code as healthy, and push it into the available queue of the battery swapping asset pool.

[0105] Conversely, if the verification result indicates the existence of a real fault—that is, severe voltage drops, insulation collapse, or uncontrolled temperature rise are reproduced in offline testing—it confirms that the battery's internal defects have been thoroughly identified. In this case, the system escalates the handling level of the tiered treatment. Specific escalation measures include disconnecting any physical charging or discharging circuits of the battery, activating the battery pack's built-in aerosol fire extinguisher for preventative spraying, automatically generating a diagnostic work order containing a complete data chain, and dispatching a special hazardous waste transport vehicle for emergency cross-city collection.

[0106] Step 7: Monitoring core quantitative indicators.

[0107] Since the method provided in this embodiment is a network model with lifelong self-evolution capability, the system continuously calculates quantitative indicators for the first-level early warning and the consistency verification process throughout the entire operation cycle. The aforementioned quantitative indicators cover the false negative rate and false positive rate, which characterize the accuracy of the early warning; the single calculation time, which characterizes the efficiency of the verification process; and the success rate of flexible disposal, which characterizes the health of asset operation.

[0108] The system backend pre-constructs a multi-dimensional hierarchical benchmark library, which sets optimal expected values ​​for various quantitative indicators for batteries at different life stages and from different manufacturers. The system compares current performance in real time in the backend. If the quantitative indicators deviate from the target thresholds in the pre-defined hierarchical benchmark library—for example, if the false alarm rate significantly exceeds the set tolerance threshold of 5% in the past month—the system iteratively optimizes the parameters used to determine anomalies. This iterative optimization includes using negative false alarm sample data collected in the past month to fine-tune the internal neural network connection weights of the battery simulation model using the gradient descent backpropagation algorithm, expanding the tolerance window size in scenario consistency verification, or recalibrating the internal and external risk coefficient allocation ratio in the comprehensive risk value formula.

[0109] After fine-tuning and parameter correction of the large-scale model in the background, the system distributes the optimized complete set of configuration files to each edge computing node and updates the hierarchical benchmark library simultaneously. This achieves the adaptive evolution of the remote monitoring and early warning algorithm for the battery swapping network and its ultimate goal of approaching zero false alarms in a closed-loop state without human intervention. Thus, the objective set by the method in this embodiment is achieved.

[0110] Example 2:

[0111] Based on Example 1, this example further provides a specific implementation method for handling data source confidence conflicts under extreme operating conditions.

[0112] In the battery swapping mode, new energy heavy trucks and other commercial vehicles often face complex and harsh operating environments. When the battery pack undergoes a long period of extreme discharge and is immediately connected to the battery swapping station for high-power rapid charging, the thermal stress inside the battery is extremely unbalanced.

[0113] To address this specific scenario, this embodiment includes a weight reconstruction step before calculating the comprehensive risk value by combining the external anomaly estimate and the internal risk value. The system first extracts the charge / discharge rate sequence of the lithium battery within a preset historical time window and calculates the cumulative stress factor characterizing the severity of the operating conditions. The aforementioned preset historical time window refers to a fixed observation period shifted backward from the moment the system currently triggers a warning. The aforementioned charge / discharge rate sequence refers to the set of ratios of the absolute value of the battery's charge / discharge current to the battery's rated capacity, recorded at a fixed sampling frequency during this observation period.

[0114] For example, to quantify the discrete charge / discharge rate sequence into characteristic parameters that can measure the risk of battery damage, the system introduces a square integral algorithm to calculate the aforementioned cumulative stress factor. The specific mathematical formula is as follows:

[0115] ;

[0116] In the formula, Defined as the aforementioned cumulative stress factor, it characterizes the overall load and intensity of charging and discharging that the battery has continuously endured over a period of time. Defined as the total number of sampling points contained within the aforementioned preset historical time window; Defined as in the first The instantaneous charge / discharge rate values ​​collected at each sampling moment; Defined as the time interval step between two adjacent sampling points. From a physical mechanism perspective, the square integral of the charge / discharge rate nonlinearly amplifies the thermal shock effect caused by extreme high currents. This cumulative stress factor Physically equivalent to mapping the Joule heat inside the battery ( Among them, physical current With the charge / discharge rate A dimensionless characterization of the cumulative effect (showing a linear positive correlation). The larger the value, the stronger the internal heat source excitation the battery experiences within the observation window, and the higher the risk of fatigue instability.

[0117] By calculating the square integral of the charge / discharge rate, the system can nonlinearly amplify the thermal stress impact caused by extremely high current rates on the battery's electrochemical structure, thereby accurately characterizing the battery's fatigue state.

[0118] After calculating the cumulative stress factor, the system obtains the external temperature change rate of the lithium battery through the underlying sensor interface. The aforementioned external temperature change rate refers to the difference in temperature rise per unit time of the physical temperature sensor attached to the battery module surface, representing the external heating phenomenon of the battery. Subsequently, the system executes data confidence conflict determination logic. When the cumulative stress factor is greater than or equal to a preset stress threshold, and the external temperature change rate is lower than a preset temperature change rate threshold, the lithium battery is determined to be in a thermal inertia hysteresis state. The aforementioned preset stress threshold refers to the maximum historical operating condition intensity threshold that the system allows the battery to withstand within the safety boundary. The aforementioned preset temperature change rate threshold refers to the minimum normal charge / discharge heating response rate calibrated by the system.

[0119] It is important to note that the aforementioned thermal inertia hysteresis state reveals the inherent limitations of physical sensors under extreme conditions. In this state, the battery interior, subjected to extreme charging and discharging stresses, has actually accumulated a large amount of heat and may even have experienced localized micro-internal short circuits. However, due to the large specific heat capacity of the electrolyte inside the battery, the thermal impedance between physical layers, and the strong intervention of the external liquid cooling system, the external physical temperature sensor fails to detect the sudden temperature rise in the core area in time, thus presenting a false sense of safety with a gradual change in external temperature. If the system continues to use a fixed static risk weight at this time, the distorted and low-risk external data will inevitably excessively dilute the extremely high-risk internal data, leading to serious catastrophic underreporting.

[0120] like Figure 5 The diagram illustrates the contrast between the surge in cumulative stress factor and the hysteresis divergence of the external temperature change rate under extreme operating conditions. In this diagram, the horizontal axis represents the continuous sampling time points of the battery in the battery swapping network experiencing changes in the operating environment, the left vertical axis represents the cumulative stress factor value calculated by the square integral algorithm, and the right vertical axis represents the external temperature change rate collected by the underlying physical sensors.

[0121] The figure shows two curves of physical quantity evolution with significantly different colors and trends. The light-colored curve with a circular mark represents the cumulative stress factor surge curve, while the dark-colored curve with a square mark represents the external temperature change rate hysteresis curve.

[0122] As can be seen from the specific scheme of this embodiment, when a new energy vehicle undergoes a long downhill high-load kinetic energy recovery and immediately enters a battery swapping station for rapid charging, the cumulative stress factor surge curve, which represents the fatigue state inside the battery, shows a relatively steep nonlinear climbing waveform. In a short period of time, it breaks through the stress threshold of 15 set by the system and rises to a higher region of about 28.

[0123] However, the external temperature change rate hysteresis curve, which represents the external heat generation phenomenon, shows a flat waveform with small fluctuations, fluctuating within a value range of about 0.5, which is lower than the normal heat generation response threshold of 2 set by the system.

[0124] The divergence in the trends and numerical differences of the two curves on the same time axis intuitively reflects the thermal inertia hysteresis effect of the battery pack's physical structure. The attached figure illustrates that under extreme boundary conditions, the response speed of external physical temperature sensors may lag. If the dynamic transfer of decision-making power and weight reconstruction are not triggered in such cases, the system may be misled by smooth external data, leading to missed risk detections. This demonstrates the rationality and practical application value of the weight reconstruction mechanism introduced in this invention.

[0125] To completely eliminate the safety blind spot caused by the aforementioned physical lag, the system must dynamically transfer decision-making control. Responding to the thermal inertia lag state, the system reduces the external risk coefficient based on the accumulated stress factor to obtain a reconstructed external risk coefficient; and increases the internal risk coefficient based on the accumulated stress factor to obtain a reconstructed internal risk coefficient. The aforementioned reconstructed external risk coefficient refers to the dynamic coefficient after processing with the attenuation penalty function, and the aforementioned reconstructed internal risk coefficient refers to the dynamic coefficient after processing with the trust compensation function. The system uses the aforementioned accumulated stress factor to construct a mathematical relationship between exponential attenuation and linear compensation, forcibly stripping away the weight proportion of externally distorted data.

[0126] For example, the specific mathematical expression of the weight reconstruction process described above, which obtains reconstruction coefficients through attenuation penalty and trust compensation, is as follows:

[0127] ;

[0128] ;

[0129] In the formula, Defined as the aforementioned external risk coefficient for restructuring; Defined as the aforementioned external risk coefficient, it is the basic weight value when the hysteresis state is not triggered; Defined as a natural constant; Defined as a preset attenuation penalty adjustment base, it determines the severity of the external weight stripping as stress increases; Defined as the aforementioned cumulative stress factor; Defined as the aforementioned internal risk coefficient of the restructuring; Defined as the aforementioned internal risk coefficient, it is the basic weight value when the lag state is not triggered; Defined as a preset trust compensation adjustment base, it determines the slope at which the internal weight increases with increasing stress. It should be noted that, to prevent excessive imbalance in weight allocation under extreme operating conditions, the system sets upper and lower thresholds for the reconstructed internal risk coefficient and the reconstructed external risk coefficient, respectively; when the reconstructed internal risk coefficient or the reconstructed external risk coefficient calculated according to the above formula exceeds the corresponding threshold boundary, the boundary value is forcibly taken.

[0130] Based on the above formula, it can be seen that the larger the cumulative stress factor, the lower the system's trust in external sensors. At this time, the system will fully and smoothly transfer the decision-making power to the internal simulation model that can see through the internal mechanism of the battery.

[0131] After completing the dynamic reconstruction of the decision weights, the system enters the final numerical replacement and calculation execution stage. The system replaces the external risk coefficient and the internal risk coefficient used in calculating the total risk assessment value with the reconstructed external risk coefficient and the reconstructed internal risk coefficient, respectively.

[0132] In detail, the system replaces the aforementioned external risk coefficient with the reconstruction external risk coefficient calculated by the aforementioned formula, and replaces the aforementioned internal risk coefficient with the reconstruction internal risk coefficient. Subsequently, the system sums the product of the external anomaly estimate and the reconstruction external risk coefficient, and the product of the internal risk value and the reconstruction internal risk coefficient to obtain a new total risk assessment value, and divides this new total risk assessment value by the sum of the reconstruction external risk coefficient and the reconstruction internal risk coefficient.

[0133] Through the aforementioned variable substitution logic, the system ultimately calculates the comprehensive risk value after hysteresis correction. With this weight reconstruction mechanism, even in extreme boundary scenarios where physical sensors collectively become thermally insensitive, the system can still penetrate physical impedance and keenly detect the critical hidden danger of near-thermal runaway within the battery, significantly improving the reliability and robustness of the battery swapping network security defense system.

[0134] For example, suppose a heavy-duty truck undergoing battery swapping has just experienced a long downhill stretch with high-load kinetic energy recovery and then immediately enters the battery swapping station. The system's preset historical time window contains four sampling points, namely... The time interval step between two adjacent sampling points. The instantaneous charge / discharge rate sequences collected within this window are as follows: , , , .

[0135] Meanwhile, the system's preset stress threshold is 15, and the preset temperature change rate threshold is 2 degrees Celsius per minute. The system's initial external risk coefficient is also set. Initial internal risk coefficient Preset decay penalty adjustment base Preset trust compensation adjustment base .

[0136] First, the system executes the square integral algorithm to calculate the cumulative stress factor. Substituting the above values ​​into the formula, the calculation process is as follows:

[0137] ;

[0138] The cumulative stress factor was calculated. .

[0139] Subsequently, the system obtains the current external temperature change rate as 0.5 degrees Celsius per minute. The system performs a condition comparison. Since the cumulative stress factor of 28 is greater than the preset stress threshold of 15, and the external temperature change rate of 0.5 degrees Celsius per minute is lower than the preset temperature change rate threshold of 2 degrees Celsius per minute, the system clearly determines that the lithium battery is in a state of thermal inertia hysteresis, and then triggers the weight reconstruction step.

[0140] At this point, due to severe hysteresis in the external sensors, the estimated value of external anomalies fed back from the standardized time-series data is extremely low. It is assumed that... The battery simulation model, when viewed from within, is approaching the critical point of a micro-internal short circuit, and the output internal risk value is extremely high. (Assuming...) .

[0141] If the weight reconstruction mechanism of this embodiment is not used, and the comprehensive risk value is calculated directly, the result is:

[0142] ;

[0143] This value of 0.55 is far below the usual high-risk alarm threshold, and the system may suffer fatal false alarms due to being misled by false external security data.

[0144] To address this issue, the system substitutes the numerical values ​​into the attenuation penalty and trust compensation formulas, and performs weight reconstruction calculations: reconstructing the external risk coefficient. The derivation process is as follows:

[0145] ;

[0146] Reconstruct internal risk coefficients The derivation process is as follows:

[0147] ;

[0148] ;

[0149] It is worth noting that, as can be seen, after dynamic weight reconstruction, the weight ratio of external distorted data is drastically compressed, while the weight of internal high-fidelity simulation data is significantly amplified. The final calculated comprehensive risk value jumps from an underestimated 0.55 to 0.835, thus accurately crossing the high-risk alarm threshold and successfully triggering a high-level warning and mandatory physical isolation. This verifies the effectiveness of this technical solution in overcoming physical lag and eliminating hidden underreporting under extreme conditions.

[0150] Example 3:

[0151] Based on the above embodiments, this embodiment further provides a specific implementation method to resolve the conflict between the system's absolute security isolation command and the hard saturation of the limited physical isolation capacity of the swapping station in extreme scenarios where the physical isolation resources of the swapping station are limited.

[0152] In the daily heavy-asset operation of battery swapping stations, when the system verifies that a battery has a genuine fault and faces an extremely high risk of thermal runaway, the system must implement the highest level of safety control. This embodiment provides a complete resource conflict scheduling and in-situ energy depletion and risk mitigation mechanism for the process of escalating the tiered handling process.

[0153] The system first acquires the target isolation compartments required to escalate the response level and then checks the remaining capacity of these compartments within the battery swapping station. The aforementioned target isolation compartments refer to specialized safety fire-fighting compartments within the battery swapping station, equipped with independent explosion-proof structures, aerosol fire extinguishing immersion systems, and reinforced smoke exhaust ducts. Due to both construction cost and physical space constraints, the number of target isolation compartments deployed within a single battery swapping station is extremely limited. The aforementioned remaining capacity refers to the number of vacant physical workstations within the target isolation compartments that are not currently occupied by other faulty batteries.

[0154] It is important to note that under extreme weather conditions or when the same batch of battery cells experiences concentrated aging and deterioration, multiple batteries within the battery swapping station may trigger high-level warnings sequentially. The system uses a low-level programmable logic controller (PLC) to poll the occupancy sensor signals of the physical workstations in real time. When the remaining capacity falls below the capacity alarm threshold, a resource allocation conflict is triggered. The aforementioned capacity alarm threshold refers to the minimum idle workstation threshold set by the system to trigger a physical deadlock warning, typically set to zero. The aforementioned resource allocation conflict represents a severe deadlock situation where the absolute security isolation scheduling instructions in the digital space cannot be directly executed by the robotic arm due to the complete depletion of real-world three-dimensional physical space resources. Forcibly returning lithium batteries in an acutely high-risk state to conventional storage areas lacking explosion-proof capabilities under such circumstances would pose a significant risk of cascading fires or even destroying the entire battery swapping station.

[0155] To break the physical deadlock caused by the aforementioned resource allocation conflict, the system introduces a physical resource preemption and unequal replacement logic based on dynamic risk gradients. In response to the resource allocation conflict, the system obtains a first risk assessment value for the historically faulty battery currently occupying the target isolation compartment, and a second risk assessment value for the lithium battery. The aforementioned historically faulty battery refers to an old faulty battery that, before the current moment, has been determined by the system to be faulty and has been legally moved into the target isolation compartment for closed storage.

[0156] The aforementioned first risk assessment value refers to the quantitative value of the residual danger level of the historically faulty battery after it has been left to cool and release internal stress for a long time in the target isolation compartment; the aforementioned second risk assessment value refers to the quantitative value of the immediate high-risk level of the lithium battery that is currently triggering a high-level warning and is in the acute outbreak stage of the pathology.

[0157] For example, to accurately quantify the residual danger of the aforementioned historically faulty batteries as they gradually decrease over time, the system introduces an exponential decay law to calculate the aforementioned first risk assessment value. The specific mathematical formula is as follows:

[0158] ;

[0159] In the formula, Defined as the aforementioned first risk assessment value; Rinitial is defined as the initial detention risk value, that is, the original comprehensive risk value calculated by the system when the historical faulty battery was initially moved into the target isolation unit; Defined as a preset risk decay constant, it is jointly determined by the chemical system of the historically faulty battery and the environmental cooling efficiency within the target isolation compartment; Defined as the isolation settling time, it is the total physical time elapsed from the moment the historically faulty battery was moved into the target isolation compartment until the current moment. Using this formula, the system can accurately depict the physical reality of the historically faulty battery gradually dissipating internal thermal stress and gradually decreasing electrochemical activity within the isolation compartment.

[0160] like Figure 6 The figure shows a cross-validation curve comparing the decay of the residual risk assessment value of historical faulty batteries with the current acute battery risk value. In this figure, the horizontal axis represents the isolation and resting time experienced by historical faulty batteries in the target isolation compartment, and the vertical axis represents the quantitative risk assessment value calculated based on the risk model.

[0161] The figure shows a dark curve that smoothly decreases over time and a light-colored straight line that remains stable. The dark curve represents the evolution trajectory of the first risk assessment value of historical faulty batteries, while the light-colored straight line represents the second risk assessment value of new faulty lithium batteries that are currently triggering high-level warnings.

[0162] As can be seen from the specific scheduling scheme in this embodiment, when the historical faulty battery is first moved into the isolation compartment, its initial risk assessment value is 0.95, which is a high-risk state. However, as the isolation and resting time goes by, the internal thermal stress of the battery is gradually released and the electrochemical activity decreases, and the dark risk assessment curve shows an exponential decay waveform.

[0163] Meanwhile, the risk assessment value of the currently malfunctioning acute battery remains constant at the warning line of 0.75. From the waveform transformation trajectories of both, it can be observed that at a point where the isolation and resting time reaches approximately 3.6 hours, the dark exponential decay curve crosses downwards across the light-colored warning line, forming a risk reversal crossover point marked with a star. The appearance of this crossover point has significance for resource allocation decisions; it indicates that after the initial resting and cooling process, the residual danger level of the old battery is now lower than the immediate danger level of the newly malfunctioning battery.

[0164] Based on the risk replacement logic revealed by this dynamic intersection, the system generates a preemptive scheduling command to move the cooled old batteries out of the regular storage space, thereby allocating a special safety fire-fighting storage space to the new batteries that are on the verge of danger when physical space resources are scarce.

[0165] This physical resource preemption scheduling mechanism based on dynamic risk attenuation and cross-comparison provides a scheduling scheme to deal with limited fire compartment capacity, enhancing the invention's defense capabilities and feasibility in the face of complex fault scenarios.

[0166] After obtaining the aforementioned first risk assessment value and the aforementioned second risk assessment value, the system executes a strict asymmetric risk replacement comparison logic: if the first risk assessment value is lower than the second risk assessment value, a preemptive scheduling command is generated to control the battery swapping station to move the historically faulty battery out of the regular storage area and move the lithium battery into the target isolation storage area. The aforementioned preemptive scheduling command is the underlying motion control code issued to the battery swapping station's 3D stacker crane and charging / swapping robotic arm.

[0167] Under this logic, since the historically faulty batteries have been sufficiently cooled, the probability of spontaneous combustion has been significantly reduced. The system determines that moving them to a conventional storage area lacking explosion-proof measures is within an acceptable secondary risk range. By removing the old batteries, the system successfully freed up a life-saving special safety fire-fighting compartment for the lithium batteries currently on the verge of acute explosion, thus completing resource optimization and reallocation based on real-time risk gradients even when physical resources were depleted.

[0168] It is also important to note that in some extremely severe cascading failure scenarios, the old batteries inside the explosion-proof compartment are also in an extremely dangerous and uncontrollable state. The system executes a comparison logic; if the first risk assessment value is greater than or equal to the second risk assessment value, the system absolutely prohibits any physical movement of the historically faulty battery, as its danger level remains extremely high or it is already in the irreversible early stage of thermal runaway. Due to the failure to seize the battery, the newly triggered lithium battery will face the desperate situation of having no explosion-proof compartment available. Therefore, the system generates an in-situ active de-energization command, controlling the conventional compartment where the lithium battery is currently located to perform a forced discharge operation on the lithium battery until its state of charge drops below a safe threshold.

[0169] Optionally, the aforementioned in-situ active de-energization command is the highest-priority inverter scheduling message issued by the system to the bidirectional charge / discharge inverter equipment configured in the conventional bay. Although the aforementioned conventional bay lacks explosion-proof isolation capabilities, it connects to the internal DC bus of the battery swapping station and the bidirectional power conversion system. Ignoring the cycle life damage of the lithium battery, the system forcibly controls the charger to operate at full reverse power, rapidly converting the vast chemical energy contained within the lithium battery into electrical energy and feeding it back to the internal energy storage system of the battery swapping station or the external AC power grid. The aforementioned state of charge refers to a physical parameter characterizing the current remaining percentage of battery capacity. The aforementioned safety threshold refers to a low-energy red line standard pre-calibrated through numerous thermal runaway destructive experiments.

[0170] Numerous experiments have shown that the intensity and destructive power of thermal runaway in lithium batteries are absolutely positively correlated with the total amount of chemical energy contained within them. When the state of charge of a battery is forcibly depleted and reduced below the safety threshold, even if physical defects such as micro-short circuits or separator punctures still exist inside, the probability of deflagration and the potential fire will drop precipitously due to the lack of sufficient energy support.

[0171] This embodiment utilizes an in-situ active de-energization mechanism to successfully achieve in-situ safety mitigation in conventional storage areas lacking explosion-proof capabilities, even when the physical explosion-proof resources of the battery swapping station are completely exhausted and physical location relocation is impossible. This provides an unbreakable safety net for emergency response in extreme scenarios of the battery swapping network, ensuring battery swapping safety.

[0172] Example 4:

[0173] like Figure 7 As shown, this embodiment provides a remote monitoring and early warning system for lithium batteries suitable for battery swapping. The system in this embodiment relies on a cloud-edge collaborative distributed network architecture, deeply integrating and linking the underlying IoT sensing devices, edge-side automated actuators, and cloud-based high-performance computing clusters.

[0174] The lithium battery remote monitoring and early warning system for battery swapping mode provided in this embodiment includes a first-level early warning module, an internal risk calculation module, a comprehensive risk calculation module, a consistency verification module, and an anomaly judgment and handling module.

[0175] For example, the system includes a first-level early warning module for triggering a first-level early warning for the lithium battery in response to a potential abnormal signal collected from the battery. In practical physical applications, this first-level early warning module primarily relies on edge computing gateways deployed at various battery swapping stations and IoT communication terminals that travel with the vehicle or battery pack.

[0176] In the battery swapping network, the lithium battery is internally equipped with a low-level hardware sensing network consisting of a high-precision voltage sampling chip, a current Hall sensor, and multi-point thermistors. This hardware sensing network collects the battery's physical characterization data in real time, packages it through a microcontroller, and uploads it to the station control host of the swapping station or a cloud receiving server. The streaming data processing engine in the first-level early warning module processes this standardized time-series data in real time. Once it detects a potential abnormal signal, such as a nonlinear increase in differential pressure or a small change in the rate of temperature rise, the module immediately highlights the physical medium access control address of the lithium battery in the system's overall asset scheduling topology map, triggering the first-level early warning. It also coordinates with the edge computing gateway to increase the battery's data reporting frequency from the conventional low-frequency heartbeat mode to a high-frequency diagnostic mode at the millisecond level.

[0177] It should also be noted that, in order to overcome the fatal flaw of external sensor data exhibiting a false safety appearance under thermal inertial hysteresis as pointed out in the background art, the system includes an internal risk calculation module for calculating an internal risk value by fitting the internal defect parameters of the lithium battery through a battery simulation model in response to the first-level warning.

[0178] In actual deployment, the aforementioned internal risk calculation module is mounted on a GPU-accelerated computing cluster in a cloud data center. Because fitting the internal defect parameters of the battery requires massive high-dimensional matrix inversions and partial differential equation solutions, the computing power of edge devices cannot meet real-time requirements. Therefore, upon triggering the first-level warning, the cloud cluster immediately allocates a dedicated computing container and memory thread to the lithium battery, retrieving a pre-trained offline digital twin simulation model. This module uses the received high-frequency current and temperature sequences as the driving force input model, employing a particle filtering algorithm to reconstruct the battery's true electrochemical state in the cloud virtual space, such as the degree of local thermal shrinkage of the separator or the thickness of lithium plating on the negative electrode surface. Through this high-computing-power mapping mechanism in the cloud, the system successfully penetrates the thermal impedance caused by the heavy metal casing and coolant of the physical battery pack, calculating accurate internal risk values. Thus, even under extreme boundary conditions of physical sensor failure or hysteresis, it can still accurately assess the battery's true degradation status.

[0179] After acquiring data from external observations and internal calculations, the system includes a comprehensive risk calculation module, which combines external anomaly estimates and the internal risk values ​​to calculate a comprehensive risk value. This comprehensive risk calculation module serves as the central decision-making unit of the entire system, containing a dynamic weight reconstruction logic unit.

[0180] As described in Example 2, when the vehicle operates under severe conditions such as long downhill kinetic energy recovery, this module calculates the cumulative stress factor by extracting historical charge-discharge rate sequences. In practical applications, this module is connected to the weather micro-environment monitoring station of the battery swapping station and the vehicle's in-vehicle infotainment system bus to cross-observe the external ambient temperature change rate. Once it is determined that the battery has entered a state of thermal inertia hysteresis, the microprocessor in the comprehensive risk calculation module will use a preset exponential decay calculation circuit and a linear compensation calculation circuit to instantly weaken the influence of external physical sensor data in the decision-making system and proportionally amplify the data weight output by the cloud-based internal risk calculation module. Through this algorithm engine that integrates dynamic stress penalty, the comprehensive risk value output by this module can accurately and without distortion reflect the global thermal runaway probability of the battery, completely solving the hidden danger of missed detection caused by static warning weights.

[0181] Optionally, in order to cope with complex scenario-based interference such as plugging and unplugging shocks and sudden changes in operating conditions under the battery swapping mode, and to avoid false alarms causing high-quality assets to be incorrectly sealed, the system includes a consistency verification module, which is used to perform consistency verification on the potential abnormal signals when the comprehensive risk value reaches the second-level warning condition.

[0182] The aforementioned consistency verification module is physically implemented as a multi-source data cross-comparison server. When the comprehensive risk value exceeds the second-level warning threshold, the module does not immediately issue a command to cut off the physical circuit. Instead, it concurrently calls three independent data interfaces. The first interface connects to the battery management system to verify the physical coupling trends of voltage, current, and temperature to complete parameter consistency verification. The second interface connects to the programmable logic controller and vehicle control unit of the battery swapping station to retrieve robotic arm action logs, charger start / stop logs, and driver operation logs to complete scenario consistency verification. The third interface connects to the distributed big data storage architecture in the cloud to call the historical failure record database to perform dynamic time warping matching to complete evolutionary consistency verification. Only when the underlying hardware logs and data characteristics of these three dimensions completely point to real physical lesions will the module issue a warning signal.

[0183] After all the above verification links are established, the system includes an anomaly detection and handling module, which is used to determine that the anomaly is real when the consistency verification passes, determine the risk level of the lithium battery based on the comprehensive risk value, and perform corresponding graded handling.

[0184] The aforementioned anomaly detection and handling module is the actuator that directly intervenes in the real three-dimensional physical world of the system. This module is hardwired with the three-dimensional stacker crane, charging and swapping robotic arm, special explosion-proof fire-fighting compartment, and bidirectional power conversion system within the battery swapping station via industrial Ethernet. As described in Example 3, when encountering a conflict between an extremely high-risk fault and the saturation of the explosion-proof isolation capacity of the battery swapping station, the resource scheduling engine within this module will capture the cooling attenuation data of historical faulty batteries in each compartment in real time. If it is determined that preemption is possible, this module will directly issue instructions to the main control programmable logic controller of the battery swapping station, controlling the massive steel robotic arm to physically swap and replace batteries in the conventional idle compartment with those in the explosion-proof compartment.

[0185] It is important to note that if the explosion-proof compartment resources become completely locked, the anomaly detection and handling module will activate the ultimate flexible handling bottom line. This module directly issues an in-situ active power depletion command to the conventional charging and swapping station where the target lithium battery is currently located. Under this command, the bidirectional AC / DC inverter connected to this station instantly switches to full-load inverter discharge mode. The charger not only stops injecting energy into the battery, but also uses its built-in high-power insulated-gate bipolar transistor inverter circuit to forcibly invert the electrical energy inside the battery into industrial frequency AC power and connect it to the external power grid, draining the chemical energy of the high-risk battery in a very short time.

[0186] By combining the collaborative operation of all the aforementioned modules, the system in this embodiment integrates big data algorithms in the cloud, real-time decision-making at the edge, and complex electromechanical equipment within the battery swapping station, breaking the deadlock between absolute safety commands and physical space capacity limitations, and providing reliable software and hardware system-level protection for the large-scale safe commercial operation of the battery swapping mode.

[0187] Example 5:

[0188] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0189] like Figure 8 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 100 includes a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, via a bus 102. Optionally, the electronic device 100 may further include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one unit, and the structure of this electronic device 100 does not constitute a limitation on the embodiments of the present invention.

[0190] Processor 101 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0191] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI bus or an EISA bus, etc. Bus 102 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0192] The memory 103 stores a computer program corresponding to the lithium battery remote monitoring and early warning method applicable to the battery swapping mode according to the above embodiments of the present invention. The computer program is controlled and executed by the processor 101. The processor 101 executes the computer program stored in the memory 103 to implement the content shown in the foregoing method embodiments.

[0193] Among them, electronic devices 100 include, but are not limited to: mobile terminals such as laptops and PADs (tablet computers) and fixed terminals such as desktop computers. Figure 8 The electronic device 100 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0194] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for remote monitoring and early warning of lithium batteries suitable for battery swapping, characterized in that, Includes the following steps: In response to the collected potential abnormal signals of the lithium battery, a first-level warning for the lithium battery is triggered; In response to the first-level warning, the internal defect parameters of the lithium battery are fitted by a battery simulation model to calculate the internal risk value; The comprehensive risk value is calculated by combining the external anomaly estimate and the internal risk value. If the overall risk value reaches the second-level warning condition, then a consistency check is performed on the potential abnormal signal; If the consistency check passes, it is determined to be a genuine anomaly. Based on the comprehensive risk value, the risk level of the lithium battery is determined, and the corresponding graded handling is performed.

2. The method according to claim 1, characterized in that, The potential abnormal signals are extracted based on standardized time-series data, and the process of generating the standardized time-series data includes: Collect the electrical performance data of the lithium battery and match the corresponding battery management system feature fingerprint according to the battery pack identifier; Based on the parameter mapping rules and noise types in the characteristic fingerprint of the battery management system, the electrical performance data is fused and noise suppressed to obtain the standardized time-series data.

3. The method according to claim 1, characterized in that, The process of calculating the comprehensive risk value by combining the external anomaly estimate and the internal risk value includes: Obtain the external risk coefficient and internal risk coefficient corresponding to the battery type of the lithium battery; The total risk assessment value is obtained by summing the product of the external anomaly estimate and the external risk coefficient, and the product of the internal risk value and the internal risk coefficient. The total risk assessment value is obtained by dividing the total risk assessment value by the sum of the external risk coefficient and the internal risk coefficient.

4. The method according to claim 3, characterized in that, Before the process of calculating the comprehensive risk value by combining the external anomaly estimate and the internal risk value, a weight reconstruction step is also included: Extract the charge / discharge rate sequence of the lithium battery within a preset historical time window, and calculate the cumulative stress factor characterizing the severity of the operating conditions; The external temperature change rate of the lithium battery is obtained. When the cumulative stress factor is greater than or equal to a preset stress threshold and the external temperature change rate is lower than a preset temperature change rate threshold, the lithium battery is determined to be in a thermal inertia hysteresis state. In response to the thermal inertia hysteresis state, the external risk coefficient is reduced according to the cumulative stress factor to obtain the reconstructed external risk coefficient; The internal risk coefficient is then increased based on the accumulated stress factor to obtain the reconstructed internal risk coefficient. The external risk coefficient and the internal risk coefficient used in calculating the total risk assessment value are replaced with the reconstructed external risk coefficient and the reconstructed internal risk coefficient, respectively.

5. The method according to claim 1, characterized in that, The consistency verification includes parameter consistency verification, scenario consistency verification, and evolutionary consistency verification. The consistency check performed on the potential abnormal signal includes: If the changing trends of multiple parameters of the lithium battery are consistent, then the parameter consistency check is passed. If the potential abnormal signal excludes the preset scenario-based action interference, then the scenario consistency check is passed. If the temporal evolution trajectory of the potential abnormal signal matches the preset fault precursor evolution feature library with a degree greater than or equal to the degree of matching threshold, then the evolution consistency check is passed. When the parameter consistency check, the scenario consistency check, and the evolution consistency check all pass, the consistency check is determined to be successful.

6. The method according to claim 1, characterized in that, The execution of the corresponding tiered processing includes: Obtain the current operating status information of the battery swapping station, including the battery swapping period and the number of batteries in stock; Based on the battery swapping period and the battery inventory quantity, it is determined whether the battery swapping station is in a high-pressure state or a low-pressure state. When the risk level is at a preset low to medium risk level, obtain the elastic threshold range corresponding to the risk level; If the battery swapping station is in the high-pressure state, the warning threshold will be adjusted to the upper limit of the elastic threshold range; If the battery swapping station is in the low-pressure state, the early warning threshold will be adjusted to the lower limit of the elastic threshold range. The tiered response is executed based on the adjusted warning threshold.

7. The method according to claim 1, characterized in that, After performing the corresponding graded processing, it also includes: Trigger the verification process for the lithium battery; If the verification result of the verification process is that there is no real fault, then the warning for the potential abnormal signal is lifted and the lithium battery operation is restored. If the verification result indicates the existence of a real fault, the handling level of the graded handling will be upgraded.

8. The method according to claim 7, characterized in that, The elevation of the treatment level in the tiered treatment includes: Obtain the target isolation compartment required to upgrade the treatment level, and detect the remaining capacity of the target isolation compartment within the battery swapping station; When the remaining capacity is lower than the capacity alarm threshold, a resource allocation conflict is triggered. In response to the resource allocation conflict state, obtain the first risk assessment value of the historical faulty battery currently occupying the target isolation compartment, and the second risk assessment value of the lithium battery; If the first risk assessment value is lower than the second risk assessment value, a preemptive scheduling instruction is generated to control the battery swapping station to move the historically faulty battery out to the regular storage location and move the lithium battery into the target isolation storage location. If the first risk assessment value is greater than or equal to the second risk assessment value, an in-situ active de-energization command is generated to control the conventional storage location where the lithium battery is currently located to perform a forced discharge operation on the lithium battery until the state of charge of the lithium battery drops below the safety threshold.

9. The method according to claim 1, characterized in that, The method further includes: Calculate the quantitative indicators for the first-level early warning and the consistency verification process; If the quantitative indicator deviates from the target threshold in the preset hierarchical benchmark library, the parameters used to determine the anomaly are iteratively optimized, and the hierarchical benchmark library is updated synchronously.

10. A remote monitoring and early warning system for lithium batteries suitable for battery swapping, characterized in that, include: The first-level early warning module is used to trigger a first-level early warning for the lithium battery in response to the collected potential abnormal signals of the lithium battery. An internal risk calculation module is used to calculate the internal risk value by fitting the internal defect parameters of the lithium battery through a battery simulation model in response to the first-level warning. The comprehensive risk calculation module is used to calculate the comprehensive risk value by combining the external anomaly estimate and the internal risk value; The consistency verification module is used to perform consistency verification on the potential abnormal signal when the comprehensive risk value reaches the second-level warning condition; The anomaly detection and handling module is used to determine a genuine anomaly when the consistency verification passes, determine the risk level of the lithium battery based on the comprehensive risk value, and perform corresponding graded handling.

Images