A mobile power supply multi-dimensional battery cell safety state monitoring system

Abstract

The application discloses a kind of mobile power supply multidimensional battery safety state monitoring system, it is related to mobile power supply safety monitoring technical field, comprising: multivariate data acquisition module, time sequence feature extraction module, multi-feature fusion decision module, hierarchical response control module and storage unit, each module realizes data interaction and instruction transmission by internal bus;The application constructs the system level framework of multivariate time sequence feature fusion and fault hierarchical diagnosis, uses sliding window analysis to extract the change rate of each monitoring parameter, fluctuation variance, high-frequency noise energy etc.time sequence features, combined with the typical evolution trajectory in multi-dimensional feature space carries out pattern matching classification, realizes with multivariate time sequence feature fusion mode replaces single parameter threshold overrun fault determination logic, while supporting four-level hierarchical progressive response strategy, fundamentally solve the problem that traditional scheme cannot capture early implicit fault signal, safety hidden danger pre-judgment lag.

Description

Technical Field

[0001] This invention relates to the field of mobile power bank safety monitoring technology, specifically a multi-dimensional cell safety status monitoring system for mobile power banks. Background Technology

[0002] A portable power bank is a portable power supply device that integrates energy storage cells and charging / discharging management circuits. It provides emergency power to smartphones, portable digital devices, and other mobile electronic terminals when out of mains power, making it an indispensable accessory for daily travel, outdoor work, and other scenarios. With the increasing popularity of consumer electronic devices and the growing demand for longer battery life, the frequency of use and carrying scenarios for power banks are constantly expanding. Their operational safety directly relates to the personal and property safety of users and is a core performance indicator for power bank products. The battery cell is the core energy storage unit of a power bank. Currently, most mainstream power banks use lithium-ion cells as the energy storage carrier. The operating parameters of the battery cell, such as voltage, temperature, internal resistance, and structural deformation, directly reflect its safety status. Abnormal fluctuations in even a single parameter may indicate safety risks. Multi-dimensional safety status monitoring of the battery cell can comprehensively perceive changes in its state during operation, overcoming the limitations of single-parameter monitoring. It can identify potential safety hazards such as overcharging, overheating, internal micro-short circuits, and cell aging in advance, providing a reliable basis for triggering safety protection mechanisms. This is of great significance for avoiding serious safety accidents such as thermal runaway and extending the service life of power banks.

[0003] Patent CN120559487A discloses a portable real-time monitoring method and system for multi-parameter coupling of lithium batteries. The scheme first initializes the lithium battery operating state set, defines standard parameter ranges under different charge / discharge rates, and generates multi-parameter collaborative feature classes. It then collects real-time temperature and load parameters through integrated sensors, generates real-time parameter feature labels using a segmented time sliding window, and constructs feature points within the window. Next, it constructs a parameter coupling association set based on the feature points within the window, evaluates the coupling correlation between different operating states to complete risk screening, and finally evaluates the membership degree of the feature points within the window based on the risk screening results to form a corresponding early warning monitoring strategy. This reduces the false negative rate of the single-parameter threshold method and improves the accuracy of lithium battery monitoring.

[0004] While existing technologies have achieved multi-parameter coupled monitoring and risk warning, they still rely on whether the parameter values ​​fall within a preset standard range as the core basis for judgment. They can only conduct risk screening based on the coupling correlation of parameter amplitudes, without deeply exploring the temporal change characteristics of each monitored parameter. They cannot capture latent signals such as abnormal parameter fluctuations and high-frequency noise disturbances in the early stages of fault evolution. At the same time, fault classification is based solely on membership values, lacking a hierarchical diagnostic logic that matches the fault evolution trajectory. This makes it difficult to accurately identify early latent faults such as micro-short circuits and aging attenuation inside the battery cell. There are still problems such as delayed prediction of safety hazards and insufficient coverage of monitoring blind spots. They cannot fully adapt to the complex and ever-changing portable usage scenarios of power banks. Therefore, developing a multi-dimensional battery cell safety status monitoring system for power banks is of great significance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multi-dimensional battery cell safety status monitoring system for mobile power banks. It can solve the technical problems of existing mobile power bank battery cell monitoring schemes that only rely on parameter amplitude coupling to make risk judgments, fail to explore the characteristics of parameter time sequence changes, lack hierarchical diagnostic logic that matches the fault evolution trajectory, resulting in low accuracy of early hidden fault identification, delayed prediction of safety hazards, and insufficient coverage of monitoring blind spots.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a multi-dimensional battery cell safety status monitoring system for mobile power banks, the system comprising: a multi-parameter data acquisition module, a time-series feature extraction module, a multi-feature fusion decision module, a hierarchical response control module, and a storage unit, wherein each module realizes data interaction and command transmission through an internal bus;

[0007] The entire system adopts a low-power hierarchical scheduling architecture. Under normal circumstances, it only maintains the low-power operation of the multi-parameter data acquisition module and the basic judgment logic. When an abnormality is triggered, the higher-order computing modules are woken up step by step. While ensuring the real-time monitoring, it adapts to the portable and low-power requirements of mobile power banks.

[0008] The multi-parameter data acquisition module is used to acquire the raw data streams of voltage, temperature, expansion stress and internal resistance of the battery cell and transmit them to the time-series feature extraction module;

[0009] Each acquisition unit adopts a synchronous trigger acquisition mechanism to ensure that the acquisition timestamp deviation of the four types of parameters, namely voltage, temperature, stress and internal resistance, is controlled within milliseconds at the same time. This provides a unified time reference for the accurate coupling calculation of subsequent multi-parameter time-series characteristics and eliminates feature matching errors caused by time offset.

[0010] Furthermore, the multi-parameter data acquisition module integrates a voltage acquisition unit, a temperature acquisition unit, an expansion stress acquisition unit, and an internal resistance detection unit. The voltage acquisition unit is connected in parallel to the positive and negative terminals of the battery cell to acquire the open-circuit voltage and on-load operating voltage data of the battery cell. The temperature acquisition unit is attached to the surface of the battery cell and the tab position to acquire the temperature data of the battery cell body and the tab. The expansion stress acquisition unit is set at the gap between the battery cell encapsulation surface and the inner wall of the outer shell to acquire the expansion stress and deformation data of the battery cell. The internal resistance detection unit is connected to the battery cell circuit to acquire the AC internal resistance data of the battery cell.

[0011] Furthermore, the internal resistance detection unit adopts an AC injection detection circuit, which periodically injects a fixed amplitude AC detection signal into the cell circuit, collects the voltage response signal and current response signal of the circuit, and calculates the AC internal resistance value of the cell.

[0012] The frequency of the AC detection signal is set to the characteristic frequency point of the electrochemical impedance spectrum of the corresponding cell model, which can effectively avoid the interference of the inter-electrode capacitive reactance characteristics of the cell on the internal resistance detection results and improve the accuracy and stability of AC internal resistance detection. The internal resistance detection cycle is decoupled from the voltage and temperature acquisition cycle, and the detection frequency can be adjusted independently according to the operating conditions, taking into account both detection accuracy and power consumption control.

[0013] After the multi-parameter data acquisition module completes the acquisition of raw data for each parameter, it first performs filtering and noise reduction on the raw data, and then synchronizes the processed multi-channel raw data stream in time before transmitting it to the time-series feature extraction module.

[0014] Furthermore, the expansion stress acquisition unit employs a flexible thin-film pressure sensing sheet, which is fully attached to the central area of ​​the battery cell packaging surface. The detection surface of the sensing sheet faces the battery cell packaging side, and the back side of the sensing sheet abuts against the inner wall of the power bank casing. The temperature acquisition unit includes two independent temperature sensing elements: one set is attached to the geometric center of the battery cell body surface, and the other set is attached to the welding position of the positive electrode tab of the battery cell. The two sets of temperature sensing elements independently acquire and transmit data. After acquiring the stress data, the expansion stress acquisition unit combines it with the surface temperature data of the battery cell body to calculate the actual expansion deformation of the battery cell. The calculation formula is as follows: Where δ is the actual expansion deformation of the battery cell, and σ is the stress value collected by the expansion stress acquisition unit. γ is the surface temperature of the battery cell body, k is the stress-deformation conversion coefficient, and γ is the temperature compensation coefficient. k and γ are determined through constant temperature and pressure calibration tests on battery cell samples. Fixed calibration values ​​are used for battery cells of the same model.

[0015] This temperature compensation mechanism can eliminate the interference of thermal expansion deformation of the packaging material caused by Joule heating during normal charging and discharging of the battery cell, ensuring that the final deformation only reflects the structural changes of the active material inside the battery cell and fault deformations such as lithium plating and bulging, greatly improving the ability of expansion deformation data to characterize hidden faults inside the battery cell.

[0016] The time-series feature extraction module is used to perform sliding window analysis on the raw data stream of each channel, extract the time-series features of each parameter, and transmit them to the multi-feature fusion decision module.

[0017] Furthermore, the temporal feature extraction module performs the following operations when extracting temporal features:

[0018] It receives the raw data streams from each channel transmitted by the multi-parameter data acquisition module, and performs real-time segmentation of the continuous data stream according to a fixed window length and sliding step size to generate single-parameter data segments corresponding to the time window.

[0019] The window length and sliding step size of the sliding window can be adaptively adjusted according to the operating conditions. Under high-rate charging and discharging conditions, the window length and sliding step size are automatically shortened to improve the sensitivity of capturing transient abnormal features; under low-current discharging or standby conditions, the window length is automatically increased to reduce system power consumption and adapt to the needs of different usage scenarios of the power bank.

[0020] Numerical calculations were performed on the single-parameter data segments within each time window to obtain three types of time-series characteristics: the rate of change of the corresponding parameter, the fluctuation variance, and the high-frequency noise energy.

[0021] The high-frequency noise energy is obtained by integrating the noise components of the specified frequency band after separating them by bandpass filtering. This frequency band corresponds to the characteristic noise frequency band caused by faults such as micro short circuits and poor contact inside the battery cell, which can accurately capture early weak fault signals that cannot be identified by conventional amplitude monitoring.

[0022] The time-series features of all parameters within the same time window are timestamped and packaged, and then transmitted to the multi-feature fusion decision module.

[0023] The multi-feature fusion decision module is used to map multi-dimensional temporal features to the feature space, perform pattern matching with pre-stored typical fault trajectories, and output the hierarchical diagnosis results to the hierarchical response control module.

[0024] Furthermore, the multi-feature fusion decision module performs the following operations when performing fault mode matching:

[0025] Receive multi-dimensional time-series feature data transmitted by the time-series feature extraction module, map all time-series features to a unified high-dimensional feature space, and generate cell state feature points at the corresponding time.

[0026] The state feature points of the continuous time series are connected in series to form a real-time state evolution trajectory, and the internally stored fault feature trajectory library is called.

[0027] After normalizing the features of each dimension, the matching degree between the real-time state evolution trajectory and each typical fault trajectory in the trajectory database is calculated. Based on the matching degree results, fault classification is completed, and the corresponding diagnostic results are output. The matching degree calculation formula is as follows: Where S is the matching degree between the real-time trajectory and the typical fault trajectory, and n is the total number of feature dimensions. This represents the difference between the normalized real-time value of the k-th dimension temporal feature and the corresponding value of the typical trajectory. The weight coefficients for the k-th dimension feature are determined by analyzing the feature contribution of fault samples, with core fault-related features corresponding to higher weight values.

[0028] The normalization process adopts the extreme value normalization method. The extreme value boundaries of each dimension feature are obtained by calibrating the full working condition safety range of the corresponding parameters. This can eliminate the influence of weight offset on the matching degree calculation results of features with different dimensions and orders of magnitude, and ensure the fairness and accuracy of multi-dimensional feature fusion judgment.

[0029] Furthermore, the fault feature trajectory library pre-stores typical feature evolution trajectories corresponding to four types of faults: overcharging, overheating, internal micro-short circuit, and cell aging. Each typical trajectory is obtained through sample testing and calibration. After the multi-feature fusion decision module calculates the matching degree, it further calculates the fault severity index. Based on the range of the fault severity index, the diagnostic results are divided into four levels: normal fluctuation, suspected abnormality, clear fault, and urgent danger. Each level corresponds to a different degree of fault severity and evolution trend. The diagnostic results carry the corresponding fault type information and level identifier. The formula for calculating the fault severity index is: Where i is the fault severity index, and S is the matching degree between the real-time trajectory and the corresponding fault type. τ is the rate of change of matching degree per unit time, β is the time reference constant, and β is the evolution rate weighting coefficient. τ is determined by statistical analysis of the evolution time of fault samples, and β is determined by fault evolution characteristic experiments. Different fault types correspond to independently calibrated β values.

[0030] This calculation method takes into account both the current matching degree and the evolution speed of the fault. It can trigger higher-level protection actions in advance for rapidly deteriorating transient faults, avoiding the rapid spread of faults and serious safety accidents such as thermal runaway. At the same time, for aging-type slowly evolving faults, the fault level can be accurately determined by the steady-state value of the matching degree, avoiding excessive triggering of protection and affecting normal use.

[0031] The graded response control module is used to execute system response actions corresponding to the diagnostic level;

[0032] Furthermore, under the normal fluctuation level, the graded response control module does not interfere with the normal operation of the charging and discharging circuit, but only writes operating data to the storage unit. Under the suspected abnormal level, the graded response control module sends a current limiting command to the charging and discharging management circuit and a frequency increase command to the multi-parameter data acquisition module. Under the clear fault level, the graded response control module sends a cut-off command to the main circuit protection switch to disconnect the main charging and discharging circuit. Under the emergency danger level, the graded response control module triggers the physical isolation device of the circuit to disconnect the connection between the battery cell and the external circuit, and triggers the external early warning mechanism to issue a warning signal.

[0033] The fault trigger thresholds for each level can be dynamically adjusted based on the cumulative number of battery cell cycles and real-time health status. As the battery cell ages, the trigger thresholds for each level are narrowed to accommodate the lower safety tolerance boundaries of aging battery cells. The current limiting and power reduction process adopts a smooth and gradual adjustment method to avoid voltage surges and power outages caused by sudden power changes, ensuring the stability of the power bank's external power supply process.

[0034] Furthermore, the hierarchical response control module performs the following operations when executing a system response:

[0035] Receive the hierarchical diagnosis results output by the multi-feature fusion decision module, identify the level identifier and fault type information in the diagnosis results, and call the corresponding level response strategy;

[0036] According to the response strategy, control commands are sent to the corresponding actuators to adjust the charging and discharging status of the power bank and trigger the corresponding level of protection action.

[0037] The response action information and current status data are synchronously sent to the storage unit for logging, while continuously receiving subsequent diagnostic results to adjust the response level.

[0038] The storage unit is used to store operation logs and system configuration data.

[0039] Furthermore, the storage unit adopts a non-volatile storage medium, and the stored content includes system operation logs, fault feature trajectory database data, and system configuration parameters. The system operation log records the original acquisition data of each parameter, time-series feature data, fault diagnosis results, and system response actions. The storage unit supports data reading and writing, and supports updating the fault feature trajectory database and modifying the system configuration parameters.

[0040] The storage unit adopts a partitioned storage architecture. The fault feature trajectory library and system configuration parameters are stored in a read-only protected partition, which can only be modified during authorized firmware upgrades. The operation log is stored in a rewritable partition. When the storage space is full, the oldest log data is automatically overwritten, which not only ensures the security of core configuration data, but also achieves long-term cyclic recording of operation logs.

[0041] Compared with existing technologies, this multi-dimensional battery cell safety status monitoring system for mobile power banks has the following advantages:

[0042] This invention constructs a system-level framework for multi-parameter time-series feature fusion and fault classification diagnosis. It utilizes sliding window analysis to extract time-series features such as the rate of change, fluctuation variance, and high-frequency noise energy of each monitored parameter. Combined with the typical fault evolution trajectory in the multi-dimensional feature space, it performs pattern matching classification, thereby replacing the fault judgment logic of single-parameter threshold exceedance with a multi-parameter time-series feature fusion mode. At the same time, it is equipped with a four-level hierarchical progressive response strategy, which fundamentally solves the problems of traditional solutions failing to capture early hidden fault signals and lagging in the prediction of safety hazards. It improves the fault identification accuracy of mobile power supplies in complex portable scenarios, fully covers the blind spots of traditional monitoring solutions, and effectively ensures the safety of battery cell operation and the reliability of equipment use.

[0043] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0045] Figure 1 A schematic diagram of a multi-dimensional battery cell safety status monitoring system for mobile power supplies;

[0046] Figure 2 A flowchart of a multi-dimensional battery cell safety status monitoring system for mobile power supplies;

[0047] Figure 3 This is a flowchart of the time series feature extraction module during the time series feature extraction process. Detailed Implementation

[0048] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0049] This invention provides a multi-dimensional battery cell safety status monitoring system for mobile power supplies. It constructs a system-level framework for multi-parameter time-series feature fusion and fault classification diagnosis. The system synchronously collects raw data streams of four operating parameters—cell voltage, temperature, expansion stress, and internal resistance—through a multi-parameter data acquisition module. A time-series feature extraction module uses a sliding window to extract three types of time-series features: rate of change, fluctuation, variance, high-frequency noise, and energy. A multi-feature fusion decision module maps these multi-dimensional time-series features to a high-dimensional feature space, performs pattern matching with pre-stored typical fault evolution trajectories, calculates the matching degree and fault severity index, and outputs a four-level classification diagnosis result. Finally, a classification response control module executes corresponding progressive protection actions based on the diagnosis level. A storage unit manages the storage of all operating data and configuration parameters.

[0050] The entire system adopts a low-power hierarchical scheduling architecture, replacing the traditional single-parameter threshold-based fault judgment logic with a multi-parameter timing feature fusion mode. This effectively captures latent abnormal signals in the early stages of fault evolution, solving the problems of delayed prediction of safety hazards and insufficient coverage of monitoring blind spots in existing solutions. It is fully adaptable to the complex and ever-changing portable usage scenarios of power banks. A detailed description is provided below with reference to specific embodiments.

[0051] like Figure 1 As shown, the multi-dimensional battery cell safety status monitoring system for mobile power banks used in this embodiment consists of five parts: a multi-parameter data acquisition module, a timing feature extraction module, a multi-feature fusion decision module, a hierarchical response control module, and a storage unit. Each module interacts and transmits commands via an internal integrated circuit bus. The entire system is housed in a 20000mAh polymer lithium-ion battery cell mobile power bank. The battery cell has a rated voltage of 3.7V, a charging cutoff voltage of 4.2V, a discharging cutoff voltage of 2.75V, and a rated maximum charge / discharge rate of 2C. The system adopts a low-power hierarchical scheduling architecture. Under normal operation, only the multi-parameter data acquisition module and the basic decision logic maintain low-power operation, with the overall operating current controlled below 5mA. When a parameter abnormality is detected, triggering a suspected abnormality level, the higher-order computing units of the timing feature extraction module and the multi-feature fusion decision module are sequentially activated, ensuring real-time monitoring while also considering the power bank's battery life requirements.

[0052] The multi-parameter data acquisition module integrates a voltage acquisition unit, a temperature acquisition unit, an expansion stress acquisition unit, and an internal resistance detection unit. Each acquisition unit adopts a synchronous trigger acquisition mechanism to ensure that the timestamp deviation of the four types of parameters at the same time is controlled within 5 milliseconds. This provides a unified time reference for the accurate coupling calculation of subsequent multi-parameter time-series characteristics and eliminates feature matching errors caused by time offset.

[0053] The voltage acquisition unit employs a high-precision voltage divider detection circuit connected in parallel across the positive and negative terminals of the battery cell. It can simultaneously acquire the open-circuit voltage and on-load operating voltage data of the battery cell. Under normal conditions, the sampling frequency is set to 100Hz, and the sampling accuracy can reach 1mV. The temperature acquisition unit contains two independent sets of thermistor elements. The first set is attached to the geometric center of the battery cell body surface to acquire the overall temperature data of the battery cell body. The second set is attached to the welding position of the positive electrode tab of the battery cell to acquire the tab temperature data of the current concentration area. The two sets of sensing elements acquire and transmit data independently, with the sampling frequency consistent with that of the voltage acquisition unit, and the temperature measurement accuracy can reach 0.1℃.

[0054] The expansion stress acquisition unit uses a flexible thin-film pressure sensor sheet, which is completely attached to the central area of ​​the cell's encapsulation surface. The detection surface of the sensor sheet faces the cell's encapsulation side, while the back side of the sensor sheet abuts against the inner wall of the power bank's casing. This allows for real-time acquisition of expansion stress data generated during the cell's charging and discharging process. The internal resistance detection unit employs an AC injection detection circuit, periodically injecting a fixed-amplitude AC detection signal into the cell's circuit. It acquires the circuit's voltage and current response signals and calculates the cell's AC internal resistance value using Ohm's law. The AC detection signal frequency is set to 1kHz, which is the characteristic frequency point of the corresponding cell's electrochemical impedance spectrum. This effectively avoids interference from the cell's inter-electrode capacitive reactance characteristics on the internal resistance detection results, improving the accuracy and stability of the AC internal resistance detection. The internal resistance detection cycle is decoupled from the voltage and temperature acquisition cycle. Under normal conditions, the detection cycle is set to 10 seconds, and the detection frequency can be independently adjusted under abnormal operating conditions, balancing detection accuracy and power consumption control.

[0055] After the multi-parameter data acquisition module completes the acquisition of raw data for each parameter, it first performs filtering and noise reduction processing on the raw data through a first-order low-pass filter circuit to filter out high-frequency noise caused by environmental electromagnetic interference. Then, the processed multi-channel raw data stream is time-synchronized and calibrated, and after being uniformly stamped with the corresponding timestamp, it is transmitted to the time-series feature extraction module.

[0056] In the specific implementation of this embodiment, after the expansion stress acquisition unit acquires the stress data, it calculates the actual expansion deformation of the battery cell by combining it with the surface temperature data of the battery cell body. The calculation formula is as follows: , where δ is the actual expansion deformation of the battery cell, σ is the stress value collected by the expansion stress acquisition unit, Ts is the surface temperature of the battery cell body, k is the stress deformation conversion coefficient, and γ is the temperature compensation coefficient.

[0057] The stress-strain conversion coefficient k and temperature compensation coefficient γ were determined through isothermal and pressure calibration tests on battery cell samples. The specific calibration procedure is as follows: Ten sets of brand-new battery cell samples of the same model were selected and placed in a programmable isothermal chamber with a temperature control accuracy of 0.5℃. Temperature gradients were set from 25℃ to 55℃, with a test gradient set every 5℃, for a total of seven temperature gradients. Under each temperature gradient, a gradient pressure was applied to the battery cell using a precision electric pressurization device, gradually increasing the pressure from 0N to 500N, with a test point set every 50N. After each test point was held stably for 30 seconds, the output stress value of the flexible pressure sensor and the surface deformation of the battery cell measured by the laser displacement sensor were simultaneously collected. After completing all temperature and pressure gradient tests, the average value of the 10 sets of sample data was taken, and a multiple linear regression algorithm was used for fitting and solving to obtain the calibration values ​​of the stress-strain conversion coefficient k and the temperature compensation coefficient γ. This fixed calibration value is used uniformly for battery cells of the same model, eliminating the need for individual calibration. This temperature compensation mechanism can eliminate the interference of thermal expansion deformation of the packaging material caused by Joule heating during normal charging and discharging of the battery cell, ensuring that the final deformation only reflects the structural changes of the active material inside the battery cell and fault deformations such as lithium plating and swelling, and greatly improves the ability of expansion deformation data to characterize hidden faults inside the battery cell.

[0058] like Figure 3 As shown, after receiving the raw data streams from each channel transmitted by the multi-parameter data acquisition module, the time-series feature extraction module extracts the time-series features of each parameter using the sliding window analysis method. The specific implementation process is as follows.

[0059] The system employs a sliding window segmentation technique. The timing feature extraction module segments the continuous raw data stream in real-time according to a preset window length and sliding step, generating single-parameter data segments for the corresponding time window. The window length and sliding step support adaptive adjustment based on operating conditions. The system determines the current operating condition type based on the real-time collected charging and discharging current values. When a high-rate charging / discharging condition (greater than or equal to 1C) is detected, the window length and sliding step are automatically shortened, setting the window length to 1 second and the sliding step to 0.2 seconds to improve the sensitivity of capturing transient anomalies. When a low-current discharge or standby condition (less than 0.2C) is detected, the window length and sliding step are automatically increased, setting the window length to 10 seconds and the sliding step to 2 seconds to reduce system power consumption and adapt to the needs of different power bank usage scenarios.

[0060] Single-window time-series feature calculation: For each single-parameter data segment within a time window, numerical calculations are performed to obtain three types of time-series features for the corresponding parameter: rate of change, fluctuation variance, and high-frequency noise energy. The rate of change is the ratio of the difference between the first and last values ​​of the data within the window to the window duration, reflecting the overall trend and speed of change of the parameter within the window. The fluctuation variance is the degree of dispersion of all data points within the window relative to the window average, calculated as the sum of the squares of the differences between each data point and the average value within the window divided by the total number of data points, reflecting the severity of the parameter's fluctuation within the window. High-frequency noise energy is calculated by integrating the noise components of a specified frequency band after bandpass filtering. The selected frequency band is from 1kHz to 10kHz, corresponding to the characteristic noise band caused by faults such as micro-short circuits and poor contact within the battery cell. This band can accurately capture early weak fault signals that cannot be identified by conventional amplitude monitoring. Specifically, the data within the window is first bandpass filtered to separate the noise signal of the target frequency band, and then the square of the noise signal's amplitude is integrated over the window duration to obtain the high-frequency noise energy value within that window. The four parameters of voltage, temperature, expansion, stress, and internal resistance are all calculated independently for the above three types of time series characteristics, generating a total of 12-dimensional time series characteristic data within a single window.

[0061] Feature data alignment and packaging: The time-series feature data of all parameters under the same time window are aligned according to a unified timestamp to ensure that all features correspond to the cell status in the same time interval. After alignment, the 12-dimensional time-series feature data is packaged to generate a feature data package with timestamp, which is then transmitted to the multi-feature fusion decision module.

[0062] After receiving the packaged feature data transmitted by the time-series feature extraction module, the multi-dimensional time-series feature module maps the multi-dimensional time-series features to a unified high-dimensional feature space, performs pattern matching with pre-stored typical fault evolution trajectories, and completes fault classification and level determination. The overall determination process is as follows: Figure 2 As shown, the specific implementation process is as follows.

[0063] State feature points and evolution trajectory generation: The 12-dimensional time-series feature data at each moment is mapped to a 12-dimensional high-dimensional feature space to generate the cell state feature points at the corresponding moment. The state feature points of the continuous time series are concatenated in chronological order to form a real-time evolution trajectory of the cell state. The trajectory length can be dynamically adjusted according to the operating conditions. Under normal conditions, the trajectory retention time is 60 seconds, which is extended to 300 seconds under abnormal operating conditions, thus completely recording the evolution process of the fault.

[0064] Fault trajectory matching degree calculation: The pre-stored fault feature trajectory library in the storage unit is called. The trajectory library pre-stores typical feature evolution trajectories corresponding to four types of faults: overcharging, overheating, internal micro-short circuit, and cell aging. Each typical trajectory is obtained through a large number of sample tests and calibrations. After normalizing the features of each dimension of the real-time trajectory and the typical trajectory, the matching degree between the two is calculated.

[0065] In the specific implementation of this embodiment, the matching degree calculation formula is as follows: Where S is the matching degree between the real-time trajectory and the typical fault trajectory, and n is the total number of feature dimensions. In this embodiment, n is taken as 12. This represents the difference between the normalized real-time value of the k-th dimension temporal feature and the corresponding value of the typical trajectory. denoted as the weight coefficient of the k-th dimension feature.

[0066] The normalization process employs extreme value normalization. The extreme value boundaries of each dimension feature are calibrated using the full-condition safety range of the corresponding parameter. For example, the extreme value boundary for the voltage change rate is set to -5V / s to 5V / s, the extreme value boundary for the temperature change rate is set to -10℃ / s to 10℃ / s, and the extreme value boundary for the expansion deformation change rate is set to -0.5mm / s to 0.5mm / s. Through extreme value normalization, features of different dimensions and orders of magnitude can be uniformly mapped to the interval between 0 and 1, eliminating the weight offset influence of different dimensional features on the matching degree calculation results and ensuring the fairness and accuracy of multi-dimensional feature fusion judgment.

[0067] The weight coefficients of each dimension of features were determined through feature contribution analysis of fault samples, specifically as follows: Fifty valid test samples were collected for each of the four types of faults: overcharging, overheating, internal micro-short circuit, and cell aging. Each sample contained complete fault evolution data, and 12-dimensional time-series feature data were extracted from all samples. The ReliefF feature weight algorithm was used to calculate the contribution weight of each feature to the four fault classifications. The algorithm updates feature weight values ​​by randomly selecting nearest neighbor samples to distinguish feature differences between different categories of samples. The feature weight set corresponding to each type of fault was calculated separately, and then a weighted average was performed based on the occurrence probability and risk level of each type of fault to obtain the global 12-dimensional weight coefficients. The sum of the weight coefficients was normalized to 1. Features with high correlation to core faults corresponded to higher weight values. For example, the voltage high-frequency noise energy corresponding to internal micro-short circuit faults and the electrode temperature change rate corresponding to overheating faults were assigned higher weight coefficients to increase the proportion of core fault features in the matching degree calculation.

[0068] Fault severity classification: After calculating the matching degree between the real-time trajectory and the four types of fault trajectories, the fault type with the highest matching degree is selected as the current fault classification type. Further, the fault severity index corresponding to this fault type is calculated. Based on the range of the fault severity index, the diagnostic results are divided into four levels: normal fluctuation, suspected abnormality, clear fault, and emergency danger. Each level corresponds to a different degree of fault severity and evolution trend. The diagnostic results simultaneously carry the corresponding fault type information and level identifier.

[0069] In the specific implementation of this embodiment, the formula for calculating the fault severity index is as follows: Where i is the fault severity index, and S is the matching degree between the real-time trajectory and the corresponding fault type. τ is the rate of change of matching degree per unit time, β is the time reference constant, and β is the evolution rate weighting coefficient.

[0070] The time reference constant τ is determined by statistically analyzing the evolution time of fault samples. The specific statistical method is as follows: The average evolution time from the appearance of the initial abnormal signal to the development of a definite fault is statistically analyzed for 50 samples of each of the four types of faults. Specifically, the average evolution time for overcharge faults is 120 seconds, for overheating faults it is 180 seconds, for internal micro-short circuits it is 300 seconds, and for cell aging it is 7200 seconds. The geometric mean of the average evolution time of the four types of faults is taken to obtain the time reference constant τ. In this embodiment, τ is set to 600 seconds.

[0071] The evolution rate weighting coefficient β is determined through fault evolution characteristic experiments, specifically calibrated as follows: Accelerated aging and fault-induced experiments are conducted for different fault types, controlling different fault evolution rates. Data on corresponding matching degree changes and actual fault severity are collected. The influence of the matching degree change rate on fault severity is analyzed through polynomial fitting, and the optimal weighting coefficient β for the corresponding fault type is obtained. Different fault types correspond to independently calibrated β values. In this embodiment, the β value is 0.6 for overcharge faults, 0.5 for overheating faults, 0.4 for internal micro-short circuits, and 0.2 for cell aging. For fault types with rapid evolution and high risk, a higher β value is assigned to increase the weight of the evolution rate in severity calculation, achieving early warning of rapidly deteriorating faults.

[0072] The severity index is divided into the following levels: when i is less than 0.2, it is classified as a normal fluctuation level, indicating that the cell is in normal operation and the parameter fluctuations are within a safe range; when i is greater than or equal to 0.2 and less than 0.5, it is classified as a suspected abnormal level, indicating that the cell has shown early abnormal signals and the risk of failure has increased; when i is greater than or equal to 0.5 and less than 0.8, it is classified as a clear failure level, indicating that the cell has already shown a clear failure and there is a high safety risk; when i is greater than or equal to 0.8, it is classified as an emergency danger level, indicating that the cell failure has developed to a serious stage and may cause serious accidents such as thermal runaway at any time.

[0073] After receiving the hierarchical diagnosis results output by the multi-feature fusion decision module, the hierarchical response control module executes the corresponding level of system response action according to the diagnosis level. The specific implementation process is as follows.

[0074] When the diagnostic result is at the normal fluctuation level, the graded response control module does not interfere with the normal operation of the charging and discharging circuit, but only writes the current operating data and status information to the storage unit to maintain the low power consumption monitoring state of the system.

[0075] When the diagnostic result is a suspected abnormality level, the graded response control module sends a current-limiting command to the charge / discharge management circuit and a frequency-increase command to the multi-parameter data acquisition module. The current-limiting command restricts the charge / discharge current to 0.5 times the rated current, reducing the workload of the battery cell and slowing down the fault evolution rate. The frequency-increase command increases the sampling frequency of voltage and temperature from 100Hz to 500Hz and shortens the internal resistance detection cycle from 10 seconds to 1 second, increasing the data acquisition density and more accurately capturing the evolution details of the fault, providing richer data support for subsequent fault level determination. The current-limiting and power reduction process adopts a smooth and gradual adjustment method, linearly adjusting the current to the target value within 5 seconds, avoiding battery cell voltage surges and power interruptions caused by power surges, and ensuring the stability of the power bank's external power supply process.

[0076] When the diagnostic result indicates a clear fault level, the graded response control module sends a disconnection command to the main circuit protection switch, disconnecting the main charging and discharging circuit, stopping the charging and discharging operation of the battery cell, and preventing further deterioration of the fault. The main circuit protection switch uses a low-resistance MOSFET array, with a disconnection response time of less than 1 millisecond, enabling rapid circuit disconnection.

[0077] When the diagnosis indicates an emergency danger level, the graded response control module triggers the physical isolation device in the circuit, disconnecting the battery cell from the external circuit and achieving complete isolation of the battery cell. Simultaneously, it triggers an external warning mechanism to issue a warning signal. The physical isolation device uses a fuse-type disconnect switch, which, upon triggering, achieves complete electrical isolation, preventing the battery cell from continuing to discharge through the circuit and causing risks under fault conditions. The warning mechanism includes a buzzer and a red indicator light. The buzzer emits intermittent warning sounds, and the red indicator light flashes continuously, alerting the user to the safety risk.

[0078] The fault trigger thresholds for each level can be dynamically adjusted based on the cumulative number of battery cell cycles and real-time health status. The system counts the cumulative number of charge-discharge cycles of the battery cell. When the number of cycles reaches 500, the trigger thresholds for each level are narrowed simultaneously. The narrowing ratio corresponds to the degree of battery cell capacity decay, adapting to the lower safety tolerance boundary of aging battery cells and improving the monitoring and protection sensitivity during the aging stage.

[0079] While executing response actions, the graded response control module synchronously sends the response action information and current status data to the storage unit for log recording. At the same time, it continuously receives subsequent graded diagnostic results and dynamically adjusts the response level according to changes in the fault level. When the fault level decreases, it gradually restores the charging and discharging function and normal acquisition frequency.

[0080] The storage unit uses non-volatile Flash memory chips, storing system operation logs, fault feature trajectory database data, and system configuration parameters. The storage unit employs a partitioned storage architecture, divided into a read-only protected partition and a rewritable partition. The fault feature trajectory database and system configuration parameters are stored in the read-only protected partition, which can only be modified during authorized firmware upgrades. Under normal operating conditions, it only supports reading, ensuring the security of core configuration data and preventing corruption due to accidental operations or abnormal interference. The system operation logs are stored in the rewritable partition, recording the original collected data, time-series characteristics, fault diagnosis results, and system response actions for each parameter. Each log entry carries a precise timestamp. When the rewritable partition's storage space is full, it automatically overwrites the oldest log data, achieving long-term cyclic recording of operation logs without requiring manual storage space cleanup. The storage unit supports external devices reading log data through a dedicated interface for fault analysis and product optimization.

[0081] In summary, this embodiment constructs a cell safety monitoring system covering the entire chain of data acquisition, feature extraction, fault diagnosis, and graded response through the complete system deployment and operation process described above. It replaces the traditional single-parameter threshold judgment method with a multi-parameter time-series feature fusion judgment logic, effectively capturing weak signals of latent faults such as early lithium plating in micro-short circuits within the cell, significantly advancing the prediction of potential safety hazards. The accompanying four-level graded progressive response strategy matches corresponding protection actions according to the severity of the fault, ensuring cell operation safety while avoiding over-protection that could affect normal device use. The low-power hierarchical scheduling architecture and adaptive parameter adjustment mechanism can adapt to different usage scenarios and operating conditions of power banks, balancing monitoring accuracy and battery life requirements. The entire system comprehensively covers the blind spots of traditional monitoring solutions, effectively improving the safety and reliability of power bank cell operation, and can be widely applied to safety monitoring scenarios of various portable energy storage devices.

[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A multi-dimensional battery cell safety status monitoring system for mobile power banks, characterized in that, The system includes: a multi-parameter data acquisition module, a time-series feature extraction module, a multi-feature fusion decision module, a hierarchical response control module, and a storage unit. Each module achieves data interaction and command transmission through an internal bus. The multi-parameter data acquisition module is used to acquire the raw data streams of voltage, temperature, expansion stress and internal resistance of the battery cell and transmit them to the time-series feature extraction module. The time-series feature extraction module is used to perform sliding window analysis on the raw data stream of each channel, extract the time-series features of each parameter, and transmit them to the multi-feature fusion decision module. The multi-feature fusion decision module is used to map multi-dimensional temporal features to the feature space, perform pattern matching with pre-stored typical fault trajectories, and output the hierarchical diagnosis results to the hierarchical response control module. The graded response control module is used to execute system response actions corresponding to the diagnostic level; The storage unit is used to store operation logs and system configuration data.

2. The mobile power bank multi-dimensional cell safety status monitoring system according to claim 1, characterized in that, The multi-parameter data acquisition module integrates a voltage acquisition unit, a temperature acquisition unit, an expansion stress acquisition unit, and an internal resistance detection unit. The voltage acquisition unit is connected in parallel to the positive and negative terminals of the battery cell to acquire the open-circuit voltage and on-load operating voltage data of the battery cell. The temperature acquisition unit is attached to the surface of the battery cell and the tab position to acquire the temperature data of the battery cell body and the tab. The expansion stress acquisition unit is set at the gap between the battery cell encapsulation surface and the inner wall of the outer shell to acquire the expansion stress and deformation data of the battery cell. The internal resistance detection unit is connected to the battery cell circuit to acquire the AC internal resistance data of the battery cell.

3. The mobile power bank multi-dimensional cell safety status monitoring system according to claim 2, characterized in that, The internal resistance detection unit adopts an AC injection detection circuit, which periodically injects a fixed amplitude AC detection signal into the cell circuit, collects the voltage response signal and current response signal of the circuit, and calculates the AC internal resistance value of the cell. After the multi-parameter data acquisition module completes the acquisition of the raw data of each parameter, it first performs filtering and noise reduction processing on the raw data, and then synchronizes the processed multi-channel raw data stream in time and transmits it to the timing feature extraction module.

4. The mobile power bank multi-dimensional cell safety status monitoring system according to claim 2, characterized in that, The expansion stress acquisition unit uses a flexible thin-film pressure sensing sheet, which is completely attached to the central area of ​​the battery cell packaging surface. The detection surface of the sensing sheet faces the battery cell packaging side, and the back side of the sensing sheet abuts against the inner wall of the power bank casing. The temperature acquisition unit includes two sets of independent temperature sensing elements: one set is attached to the geometric center of the battery cell body surface, and the other set is attached to the welding position of the positive electrode tab of the battery cell. The two sets of temperature sensing elements independently acquire and transmit data. After acquiring the stress data, the expansion stress acquisition unit combines it with the surface temperature data of the battery cell body to calculate the actual expansion deformation of the battery cell. The calculation formula is as follows: Where δ is the actual expansion deformation of the battery cell, and σ is the stress value collected by the expansion stress acquisition unit. γ is the surface temperature of the battery cell, k is the stress-deformation conversion coefficient, and γ is the temperature compensation coefficient.

5. A multi-dimensional battery cell safety status monitoring system for mobile power banks according to claim 1, characterized in that, The temporal feature extraction module performs the following operations when extracting temporal features: It receives the raw data streams from each channel transmitted by the multi-parameter data acquisition module, and performs real-time segmentation of the continuous data stream according to a fixed window length and sliding step size to generate single-parameter data segments corresponding to the time window. Numerical calculations were performed on the single-parameter data segments within each time window to obtain three types of time-series characteristics: the rate of change of the corresponding parameter, the fluctuation variance, and the high-frequency noise energy. The time-series features of all parameters within the same time window are timestamped and packaged, and then transmitted to the multi-feature fusion decision module.

6. The mobile power bank multi-dimensional cell safety status monitoring system according to claim 1, characterized in that, The multi-feature fusion decision module performs the following operations when performing fault mode matching: Receive multi-dimensional time-series feature data transmitted by the time-series feature extraction module, map all time-series features to a unified high-dimensional feature space, and generate cell state feature points at the corresponding time. The state feature points of the continuous time series are connected in series to form a real-time state evolution trajectory, and the internally stored fault feature trajectory library is called. After normalizing the features of each dimension, the matching degree between the real-time state evolution trajectory and each typical fault trajectory in the trajectory database is calculated. Based on the matching degree results, fault classification is completed, and the corresponding diagnostic results are output. The matching degree calculation formula is as follows: Where S is the matching degree between the real-time trajectory and the typical fault trajectory, and n is the total number of feature dimensions. This represents the difference between the normalized real-time value of the k-th dimension temporal feature and the corresponding value of the typical trajectory. denoted as the weight coefficient of the k-th dimension feature.

7. A multi-dimensional battery cell safety status monitoring system for mobile power banks according to claim 6, characterized in that, The fault feature trajectory database pre-stores typical feature evolution trajectories corresponding to four types of faults: overcharging, overheating, internal micro-short circuit, and cell aging. Each typical trajectory is obtained through sample testing and calibration. After the multi-feature fusion decision module calculates the matching degree, it further calculates the fault severity index. Based on the range of the fault severity index, the diagnostic results are divided into four levels: normal fluctuation, suspected abnormality, clear fault, and urgent danger. Each level corresponds to a different fault severity and evolution trend. The diagnostic results carry the corresponding fault type information and level identifier. The formula for calculating the fault severity index is: Where i is the fault severity index, and S is the matching degree between the real-time trajectory and the corresponding fault type. τ is the rate of change of matching degree per unit time, β is the time reference constant, and β is the evolution rate weighting coefficient.

8. A multi-dimensional battery cell safety status monitoring system for mobile power banks according to claim 7, characterized in that, Under the normal fluctuation level, the graded response control module does not interfere with the normal operation of the charging and discharging circuit, but only writes operating data to the storage unit. Under the suspected abnormal level, the graded response control module sends a current limiting command to the charging and discharging management circuit and a frequency increase command to the multi-parameter data acquisition module. Under the clear fault level, the graded response control module sends a cut-off command to the main circuit protection switch to disconnect the main charging and discharging circuit. Under the emergency danger level, the graded response control module triggers the physical isolation device of the circuit to disconnect the connection between the battery cell and the external circuit, and triggers the external early warning mechanism to issue a warning signal.

9. A multi-dimensional battery cell safety status monitoring system for mobile power banks according to claim 1, characterized in that, The hierarchical response control module performs the following operations when executing a system response: Receive the hierarchical diagnosis results output by the multi-feature fusion decision module, identify the level identifier and fault type information in the diagnosis results, and call the corresponding level response strategy; According to the response strategy, control commands are sent to the corresponding actuators to adjust the charging and discharging status of the power bank and trigger the corresponding level of protection action. The response action information and current status data are synchronously sent to the storage unit for logging, while continuously receiving subsequent diagnostic results to adjust the response level.

10. A multi-dimensional battery cell safety status monitoring system for mobile power banks according to claim 1, characterized in that, The storage unit uses a non-volatile storage medium and stores system operation logs, fault feature trajectory database data, and system configuration parameters. The system operation logs record the original acquisition data, time-series feature data, fault diagnosis results, and system response actions of each parameter. The storage unit supports data reading and writing, as well as updating the fault feature trajectory database and modifying the system configuration parameters.

Images