A method and system for testing a lithium battery for a vacuum cleaner

Abstract

The present application relates to the technical field of lithium battery testing, and particularly relates to a testing method and system for lithium batteries of vacuum cleaners. The method comprises the following steps: collecting working currents of the vacuum cleaner under various working conditions to obtain target current waveform data; configuring a programmable dynamic load control component based on the target current waveform data, and driving an electronic load unit of the programmable dynamic load control component to output an equivalent current signal; loading the equivalent current signal on the lithium battery of the vacuum cleaner, and collecting multi-point temperature data in real time; constructing a thermal field model based on the multi-point temperature data, simulating and calculating temperature field distribution data through the thermal field model, and generating temperature rise trend prediction results. The present application constructs a high-simulation testing system covering the whole process of electrical load, thermal field evolution, communication interaction and control response, and realizes dynamic, accurate and closed-loop evaluation of the real performance and thermal safety boundary of the lithium battery of the vacuum cleaner.

Description

Technical Field

[0001] This invention relates to the field of lithium battery testing technology, and in particular to a testing method and system for a vacuum cleaner lithium battery. Background Technology

[0002] As the core power supply component of cordless vacuum cleaners, the performance of lithium batteries directly determines the overall battery life, power output, and safety. Currently, most vacuum cleaners use a series-parallel combination structure of high-rate 18650 or 21700 lithium batteries, coupled with a battery management system (BMS) for overcurrent, overheat, and overvoltage protection. During the research and development and quality control stages, it is necessary to evaluate the battery's response capabilities under complex operating conditions such as varying suction power, start-up impact, and stall loads through testing methods.

[0003] However, existing testing methods for vacuum cleaner lithium batteries mostly employ static load testing methods such as constant current and constant power, which are insufficient to simulate the nonlinear and transient current characteristics under real-world operating conditions. This results in significant discrepancies between test data and actual usage conditions. Furthermore, thermal management methods often only monitor the battery casing temperature, lacking real-time awareness of the thermal evolution within the cell, easily overlooking the risk of localized overheating. In addition, existing testing systems cannot actively identify or simulate the communication protocol between the vacuum cleaner controller and the BMS, lacking high-fidelity simulation capabilities for overall machine control behavior, and cannot dynamically adjust load impedance to reproduce the characteristics of the entire system. These limitations restrict a comprehensive assessment of the true performance and safety boundaries of lithium batteries. Summary of the Invention

[0004] Therefore, it is necessary for the present invention to provide a testing method and system for vacuum cleaner lithium batteries to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, a testing method for a vacuum cleaner lithium battery includes the following steps:

[0006] Step S1: Collect the operating current of the vacuum cleaner under various operating conditions to obtain the target current waveform data; configure the programmable dynamic load control component based on the target current waveform data, and drive the electronic load unit of the programmable dynamic load control component to output the equivalent current signal;

[0007] Step S2: Load an equivalent current signal using the vacuum cleaner's lithium battery and collect multi-point temperature data in real time; construct a thermal field model based on the multi-point temperature data, simulate and calculate the temperature field distribution data through the thermal field model, and generate a temperature rise trend prediction result;

[0008] Step S3: Collect lithium battery management communication behavior based on the three-dimensional thermal field distribution map, identify the protocol type and parse the command frame structure to obtain BMS communication protocol parameters;

[0009] Step S4: Configure the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner, and drive the real-time variable impedance component to respond to the load characteristics under each state to obtain variable impedance response data.

[0010] Step S5: Based on the target current waveform data and variable impedance response data, construct the mapping relationship between the vacuum cleaner's operating state and the current waveform, and generate a whole-machine load behavior model;

[0011] Step S6: Based on the temperature rise trend prediction results, determine the risk of thermal runaway in the overall machine load behavior model, send preset protection commands by controlling the virtual vacuum cleaner controller, and then reduce the output current through the programmable dynamic load control component.

[0012] This invention overcomes the limitations of traditional testing methods that cannot reflect real-world operating characteristics by constructing a full-link testing method encompassing dynamic load control, thermal field distribution monitoring, communication behavior analysis, and control response simulation. It achieves a realistic performance reconstruction of vacuum cleaner lithium batteries under nonlinear, highly dynamic, and strongly coupled operating environments. By outputting an equivalent current signal highly matched to actual operating conditions, it effectively reproduces the current change characteristics corresponding to suction adjustment, stall impact, and start-stop transients, improving the representativeness of the test waveforms from the source. At the thermal management level, it integrates multi-point temperature data and spatial gradient processing to establish a three-dimensional thermal field model covering the battery cell, BMS, and casing, and introduces a time dimension to analyze thermal evolution trends, thereby enabling early warning of localized abnormal temperature rises and potential thermal runaway risks. The communication identification mechanism enables the extraction and analysis of the structure of BMS protocols from different manufacturers, breaking down data barriers in the control link and providing a parameter basis for subsequent simulation control. The virtual controller and variable impedance components work together, enabling dynamic linkage between control behavior, power response, and electrical load characteristics, achieving a high-fidelity reproduction of the overall control logic and its feedback to the battery load. The load behavior model integrates timing characteristics, current waveforms, and impedance response, providing data support for subsequent intelligent control and fault analysis based on state recognition. The risk assessment mechanism combines thermal field trends and operating condition responses to establish a multi-level protection strategy triggering system, achieving closed-loop control with real-time judgment and dynamic intervention during testing. This fundamentally improves the scenario fit, safety, reliability, and intelligent response capabilities of the vacuum cleaner lithium battery testing system.

[0013] Preferably, the present invention also provides a testing system for a vacuum cleaner lithium battery, used to perform the above-described testing method for a vacuum cleaner lithium battery, the testing system comprising:

[0014] The dynamic load module is used to collect the operating current of the vacuum cleaner under various operating conditions and obtain the target current waveform data; based on the target current waveform data, the programmable dynamic load control component is configured, and the electronic load unit of the programmable dynamic load control component is driven to output the equivalent current signal.

[0015] The thermal field monitoring module is used to load an equivalent current signal using the vacuum cleaner's lithium battery and collect multi-point temperature data in real time; a thermal field model is built based on the multi-point temperature data, and the temperature field distribution data is calculated through simulation of the thermal field model, and the temperature rise trend prediction results are generated.

[0016] The communication identification module is used to collect lithium battery management communication behavior based on the three-dimensional thermal field distribution map, identify the protocol type and parse the command frame structure to obtain BMS communication protocol parameters;

[0017] The virtual control module is used to configure the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner, and drive the real-time variable impedance component to respond to the load characteristics under each state to obtain variable impedance response data.

[0018] The load modeling module is used to construct a mapping relationship between the vacuum cleaner's operating state and the current waveform based on the target current waveform data and variable impedance response data, and generate a whole machine load behavior model.

[0019] The safety protection module is used to determine the risk of thermal runaway from the overall load behavior model based on the temperature rise trend prediction results. It sends preset protection commands by controlling the virtual vacuum cleaner controller and then reduces the output current through the programmable dynamic load control component.

[0020] This invention constructs a complete test link through the collaborative construction of various modules, covering real-world current reproduction, fine thermal distribution sensing, communication protocol inversion, control logic simulation, multi-dimensional data modeling, and dynamic safety response, which significantly improves the comprehensiveness and accuracy of performance and safety evaluation of vacuum cleaner lithium batteries. In the early stages of testing, the system accurately reproduces the characteristics of nonlinear and transient fluctuating currents, ensuring that the load input is representative of reality. The thermal response process utilizes both spatial distribution and temporal evolution analysis to reveal potential risks such as localized overheating and abnormal hotspots in advance. At the control path level, communication structure analysis and instruction content recognition enable engineering usability in understanding the access and response mechanisms of the BMS control logic, providing logical closed-loop support for control simulation. During the linkage simulation, synchronous feedback of electrical response and control status enables the system to reproduce complex dynamic load behavior, thereby establishing an accurate operating model with time sequence, operating condition, and current mapping relationships. Finally, based on load behavior and thermal trends, multi-level thermal runaway prediction and real-time intervention control are achieved, giving the testing process adaptive response capabilities. This comprehensively enhances the system's ability to identify the true performance boundaries and safety thresholds of lithium batteries under extreme operating conditions, meeting the full-process requirements of R&D verification, quality control, and safety assessment. Attached Figure Description

[0021] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0022] Figure 1 This is a schematic flowchart of the testing method for a vacuum cleaner lithium battery according to the present invention;

[0023] Figure 2 for Figure 1 A detailed flowchart of step S1;

[0024] Figure 3 for Figure 1 A detailed flowchart of step S2. Detailed Implementation

[0025] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0026] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0028] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a testing method for lithium batteries used in vacuum cleaners, the method comprising the following steps:

[0029] Step S1: Collect the operating current of the vacuum cleaner under various operating conditions to obtain the target current waveform data; configure the programmable dynamic load control component based on the target current waveform data, and drive the electronic load unit of the programmable dynamic load control component to output the equivalent current signal;

[0030] Step S2: Load an equivalent current signal using the vacuum cleaner's lithium battery and collect multi-point temperature data in real time; construct a thermal field model based on the multi-point temperature data, simulate and calculate the temperature field distribution data through the thermal field model, and generate a temperature rise trend prediction result;

[0031] Step S3: Collect lithium battery management communication behavior based on the three-dimensional thermal field distribution map, identify the protocol type and parse the command frame structure to obtain BMS communication protocol parameters;

[0032] Step S4: Configure the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner, and drive the real-time variable impedance component to respond to the load characteristics under each state to obtain variable impedance response data.

[0033] Step S5: Based on the target current waveform data and variable impedance response data, construct the mapping relationship between the vacuum cleaner's operating state and the current waveform, and generate a whole-machine load behavior model;

[0034] Step S6: Based on the temperature rise trend prediction results, determine the risk of thermal runaway in the overall machine load behavior model, send preset protection commands by controlling the virtual vacuum cleaner controller, and then reduce the output current through the programmable dynamic load control component.

[0035] In this embodiment of the invention, reference Figure 1 The diagram shown is a flowchart illustrating the steps of a testing method for a vacuum cleaner lithium battery according to the present invention. In this example, the testing method for the vacuum cleaner lithium battery includes the following steps:

[0036] Step S1: Collect the operating current of the vacuum cleaner under various operating conditions to obtain the target current waveform data; configure the programmable dynamic load control component based on the target current waveform data, and drive the electronic load unit of the programmable dynamic load control component to output the equivalent current signal;

[0037] In this embodiment of the invention, the vacuum cleaner main unit is first run continuously for 300 seconds in standard carpet mode (power level I), floor mode (power level II), and powerful mode (power level III). A high-precision current acquisition module (model: FLUKE A3004 FC) is used to record the current curves under each operating condition at a sampling frequency of 10kHz. The obtained raw current data is then imported into a data processing system (configured with an Intel i7-12700 processor, 32GB of memory, Windows 10 system, and NI LabVIEW installed). Using the 2022 software, pulse characteristics, periodic variation characteristics, and peak current amplitude for each gear level are extracted through Fourier transform and peak analysis methods. The obtained feature data is then incorporated into a current gear level library to construct a three-level mode current feature database. Next, based on the duration and switching frequency of each gear level under actual operating conditions, the timing splicing tool in the data processing system sequentially combines the three gear current feature data in a fixed order (Gear I 120 seconds → Gear II 90 seconds → Gear III 90 seconds) to generate a target current waveform data with a total length of 300 seconds. Cubic spline interpolation is used to eliminate instantaneous jumps between splicing segments of each gear level, ensuring waveform continuity and representativeness of actual operation. Finally, the target current waveform data is imported into a programmable dynamic load control platform (using Chroma). The 63600 series electronic load and 63110A control unit analyze the target waveform data with a time resolution of 0.1ms, calculate the current setpoint corresponding to each moment, and convert the analysis result into a PWM duty cycle signal (control frequency of 20kHz, duty cycle accuracy of 0.1%) through the control software. The PWM signal is output to the control port of the electronic load unit to drive its constant current mode operation, simulating the working current environment of the actual vacuum cleaner lithium battery under composite load conditions. Finally, the equivalent current signal with a maximum of not less than 12A and a variation range of not less than 6A is output at the load end.

[0038] Step S2: Load an equivalent current signal using the vacuum cleaner's lithium battery and collect multi-point temperature data in real time; construct a thermal field model based on the multi-point temperature data, simulate and calculate the temperature field distribution data through the thermal field model, and generate a temperature rise trend prediction result;

[0039] In this embodiment of the invention, the collected and generated equivalent current signal is loaded onto the lithium battery module via a digital control signal drive. The loading frequency of the current signal is set to 1Hz, and the signal amplitude is precisely reconstructed according to the target current waveform data constructed in step S1. The current control error must not exceed ±0.5%. During the loading process, six sets of K-type miniature thermocouple sensors need to be pre-embedded inside the lithium battery module. The placement positions include the top center point of the battery cell, the middle of the battery cell, the middle of the BMS circuit board, the lower left corner of the BMS circuit board, the front of the inner wall of the housing, and the top edge of the housing. The sensor sampling frequency is set to 10Hz, the acquisition duration must not be less than 120 minutes, and the data recording accuracy is 0.1℃. After the acquisition is completed, the temperature data is organized sequentially according to the sensor number. A rectangular coordinate system based on the battery spatial structure is established based on the heat conduction theory. Temperature data is mapped onto three-dimensional coordinates, and cubic spline interpolation is used to calculate the continuous temperature distribution in three-dimensional space, with the interpolation interval limited to 1.0 mm, resulting in a lithium battery spatial thermal field dataset. Based on this dataset, the entire battery pack is divided into 125 cubic regions (5×5×5) using a voxel partitioning method. The average temperature and temperature gradient difference are calculated for each voxel region, and color coding is used for three-dimensional rendering to generate a visualized three-dimensional thermal field distribution map. Simultaneously, the collected temperature change data over time is processed with a first derivative to calculate the temperature rise rate at each sensor point. Linear regression is performed with a 10-minute time window to obtain the temperature rise trend curve. The temperature value for the next 30 minutes is predicted every 30 minutes, and the prediction results are continuously updated using a sliding window. Finally, the temperature rise trend prediction result is output.

[0040] Step S3: Collect lithium battery management communication behavior based on the three-dimensional thermal field distribution map, identify the protocol type and parse the command frame structure to obtain BMS communication protocol parameters;

[0041] In this embodiment of the invention, after generating the three-dimensional thermal field distribution map in step S2, a data acquisition system composed of an oscilloscope and a logic analyzer is immediately activated to continuously acquire signals from the communication interface between the lithium battery management system (BMS) and the main control board. The acquisition clock frequency is fixed at 1MHz, the sampling resolution is set to 8 bits, and the acquisition time window is 300 seconds to capture complete communication cycle data. The obtained communication data stream is then restored to a standard TTL level logic signal using a differential method. After the communication data stream acquisition is completed, a frequency distribution statistical method is used to analyze the probability of occurrence of each data byte, and a byte distribution matrix is ​​constructed based on the frequency of occurrence. The start and end bits of the communication data frame are identified by the first byte and its interval features in the matrix. A byte sequence with a data interval of no more than 5ms is used as a complete command frame for extraction. Subsequently, the extracted command frame is parsed using a fixed field offset method to read the frame header, length field, command byte, data segment, and... The verification field, with the length field set to the 2nd byte, uses the CRC-8 standard for verification, with an error rate not exceeding 0.2%. Based on the numerical range of command bytes and the distribution rules of functional bits, command frames are classified into read, write, and response commands. The command code range for read commands is 0x20–0x2F, for write commands it is 0x30–0x3F, and for response commands it is 0xA0–0xAF. The data segment content structure of each type of command is extracted. Then, based on the command type and time sequence, the response interval and command cycle of each type of command are statistically analyzed to construct a BMS communication behavior time sequence diagram. Physical layer characteristic parameters of the communication interface, such as peak signal voltage (3.3V), signal rise time (not exceeding 20ns), and logic level hold time (greater than 300ns), are extracted. Finally, the physical characteristics of the communication interface, command frame structure, functional command categories, and response timing parameters are integrated to form a set of BMS communication protocol parameters.

[0042] Step S4: Configure the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner, and drive the real-time variable impedance component to respond to the load characteristics under each state to obtain variable impedance response data.

[0043] In this embodiment of the invention, the target current waveform data generated in step S1 and the BMS communication protocol parameters extracted in step S3 are input to the control interface configuration platform. The platform maps the current waveform feature values ​​to the vacuum cleaner's operating status one-to-one based on the state mapping table, and constructs a controller state transition table. The transition table includes three fixed operating speeds (low speed, medium speed, and high speed) and two dynamic states (start and stop). The average current values ​​corresponding to each state are limited to 4.2A, 6.5A, 9.8A, 2.0A, and 0A, respectively, and the state switching logic condition is set to a current slope threshold exceeding ±1.5A / s. According to the command frame structure and instruction cycle requirements in the BMS communication protocol parameters, the communication interface driver circuit is configured, and the interface level standard is UART. The system operates at 3.3V with a fixed baud rate of 115200bps. Each instruction in the control command sequence is limited to 8 bytes in length, with a command cycle of 500ms. A command buffer queue is set up inside the virtual vacuum cleaner controller. After switching states according to the state transition table, the corresponding BMS function instruction set is retrieved from the queue, and control commands are automatically and continuously sent to the BMS control interface. The control commands issued by the controller are transmitted to the BMS system via a signal transfer module. The system acquires BMS response data in real time and determines the command execution result based on the feedback fields in the response content. After determining the compatibility of the response data structure with the protocol, a communication protocol compatibility status code is generated, and the controller's internal control parameters are adjusted based on the status code, such as the delay compensation time (set to 20ms) and the number of command repetitions (set to 2 times). Simultaneously, the real-time current response values, BMS response status, and control commands collected under each state are combined to extract the load characteristic data under the current state. The load characteristic data is input to the impedance regulation control module, and the desired impedance value is calculated using the voltage-current ratio method. The desired impedance value is limited to the range of 0.2Ω to 0.6Ω. Based on the calculation results, impedance regulation control parameters are generated, including the control signal frequency (10kHz) and duty cycle (45%–70%). The control signal drives a real-time variable impedance component composed of a MOS array, and the voltage and current response values ​​at the output of the component are measured and recorded in real time to the impedance response data acquisition module. The recording interval is 50ms, and the continuous acquisition time is 300 seconds. Finally, the variable impedance response data under each operating condition is obtained.

[0044] Step S5: Based on the target current waveform data and variable impedance response data, construct the mapping relationship between the vacuum cleaner's operating state and the current waveform, and generate a whole-machine load behavior model;

[0045] In this embodiment of the invention, the target current waveform data generated in step S1 and the variable impedance response data obtained in step S4 are used as input data sources. First, the current waveform data is used to extract features using statistical processing tools. The average current, maximum current, minimum current, and rate of change of current within each 5-second time window are calculated using a piecewise integration method, resulting in four sets of feature quantities, forming a current waveform feature vector with a fixed dimension of 4. Then, the variable impedance response data is processed. The voltage and current data points collected within each control cycle (50ms) are used to calculate the instantaneous impedance value in voltage / current form. A moving average method is then used to smooth the 20 adjacent impedance values, extracting the average, maximum, and change amplitude of impedance within each 10-second time window, forming impedance feature data with a dimension limited to 3. The current waveform feature vector and impedance feature data are then synchronously combined according to the time alignment principle to generate joint feature samples. Each sample contains 7 numerical fields, arranged sequentially to form a two-dimensional dataset. The resolution is fixed at 10 seconds. Next, the operating status of each sample in the dataset is labeled according to the operating status classification standard in step S4. Low speed is labeled as status 1, medium speed as status 2, high speed as status 3, start-up as status 4, and stop as status 5. The operating status label is appended to the end field of the joint feature sample to construct a labeled dataset. Then, the dataset is divided into training and validation sets according to time order, with a ratio of 8:2. The mapping relationship between operating status and current waveform is established by sequence feature comparison. The mapping function determines the state correspondence based on the principle of minimizing the Euclidean distance between each field in the joint feature sample and the operating status label. The continuous feature sequence is classified into the corresponding operating status by stepwise comparison. Finally, the current waveform feature data and the correspondence between operating status are summarized to generate the whole machine load behavior model. The model structure is represented in the form of a state sequence and the mapping entries are stored in a table, including four fields: state number, feature vector range, impedance range, and time window number.

[0046] Step S6: Based on the temperature rise trend prediction results, determine the risk of thermal runaway in the overall machine load behavior model, send preset protection commands by controlling the virtual vacuum cleaner controller, and then reduce the output current through the programmable dynamic load control component.

[0047] In this embodiment of the invention, the temperature rise trend prediction result generated in step S2 is synchronously matched with the whole machine load behavior model constructed in step S5 along the time axis. The temperature rise slope in each time period is correlated with the corresponding operating state. The temperature rise slope is calculated using the first-order difference method to process the temperature sequence, with a unit interval of 60 seconds. When the temperature rise slope exceeds 1.2℃ / min for three consecutive periods and the current temperature is higher than 55℃, a Level 1 thermal runaway warning is triggered. If the temperature rise slope exceeds 1.5℃ / min for five consecutive periods and the current temperature is higher than 60℃, a Level 2 thermal runaway warning is triggered. If the temperature rise slope exceeds 1.8℃ / min in any period and the temperature exceeds 65℃, a Level 3 thermal runaway alarm is triggered. Multi-level risk warning signals are generated according to these warning rules. Then, control strategies are set according to the warning level. Under a Level 1 warning, the output current amplitude is reduced by 20% by controlling the programmable dynamic load control component; under a Level 2 warning, the amplitude is reduced by 40%; and under a Level 3 alarm, the load output is immediately disconnected, and a signal synchronization module is used to... The virtual vacuum cleaner controller sends a state change command, forcibly switching to a stop state within the controller. In practice, after executing the command, the controller suspends all BMS communication activities, and the state switch response time is controlled within 500ms. The programmable dynamic load control component uses a digital PID adjustment module to adjust the PWM signal parameters, reducing the duty cycle to 80%, 60%, or 0% of the original value, corresponding to actual load reduction control. The PID control cycle is 10ms, and the adjustment time does not exceed 5 seconds. After completing the above control actions, the temperature changes of each point of the lithium battery pack are monitored in real time and recorded to the temperature feedback channel. The temperature feedback sampling frequency is set to 2Hz, and the continuous recording time is 300 seconds. The system judges whether the temperature drop trend is stable. If the temperature rise slope is below 0.5℃ / min for five consecutive segments, the control is considered effective; otherwise, the above control process is repeated until stability is achieved. Finally, the control commands, temperature change data, state switch records, and execution time during the overload protection process are summarized to generate a protection response data report.

[0048] This invention overcomes the limitations of traditional testing methods that cannot reflect real-world operating characteristics by constructing a full-link testing method encompassing dynamic load control, thermal field distribution monitoring, communication behavior analysis, and control response simulation. It achieves a realistic performance reconstruction of vacuum cleaner lithium batteries under nonlinear, highly dynamic, and strongly coupled operating environments. By outputting an equivalent current signal highly matched to actual operating conditions, it effectively reproduces the current change characteristics corresponding to suction adjustment, stall impact, and start-stop transients, improving the representativeness of the test waveforms from the source. At the thermal management level, it integrates multi-point temperature data and spatial gradient processing to establish a three-dimensional thermal field model covering the battery cell, BMS, and casing, and introduces a time dimension to analyze thermal evolution trends, thereby enabling early warning of localized abnormal temperature rises and potential thermal runaway risks. The communication identification mechanism enables the extraction and analysis of the structure of BMS protocols from different manufacturers, breaking down data barriers in the control link and providing a parameter basis for subsequent simulation control. The virtual controller and variable impedance components work together, enabling dynamic linkage between control behavior, power response, and electrical load characteristics, achieving a high-fidelity reproduction of the overall control logic and its feedback to the battery load. The load behavior model integrates timing characteristics, current waveforms, and impedance response, providing data support for subsequent intelligent control and fault analysis based on state recognition. The risk assessment mechanism combines thermal field trends and operating condition responses to establish a multi-level protection strategy triggering system, achieving closed-loop control with real-time judgment and dynamic intervention during testing. This fundamentally improves the scenario fit, safety, reliability, and intelligent response capabilities of the vacuum cleaner lithium battery testing system.

[0049] Preferably, step S1 includes the following steps:

[0050] Step S11: Perform multi-condition operation tests on the vacuum cleaner main unit to obtain current characteristic data for each setting;

[0051] Step S12: Perform time-series modeling of the vacuum cleaner's operating mode based on the current characteristic data of each gear to obtain the target current waveform data;

[0052] Step S13: Configure the programmable dynamic load component parameters based on the target current waveform data to obtain the PWM control parameters;

[0053] Step S14: Drive and control the electronic load unit in real time based on PWM control parameters to obtain the equivalent current signal.

[0054] In this embodiment of the invention, the vacuum cleaner main unit and a standard test power module are first connected to the experimental platform. The main unit is then set to three operating speeds: low, medium, and high. The continuous operating time for each speed is set to 300 seconds. A high-precision digital current sensor (measurement accuracy ±0.2%, sampling frequency 1kHz) is used to collect the current value of the main current path in real time, recording the instantaneous peak and average current values ​​per second. The current characteristic data for each speed is extracted by the data processing unit. The average current characteristic value for the low-speed speed is 4.0A, for the medium-speed speed it is 6.2A, and for the high-speed speed it is 9.5A. The current fluctuation amplitude must not exceed ±0.5A. After extracting the current characteristics for the three operating conditions, the current data for each speed is constructed into a complete time sequence according to the operating order. The interval calibration method is used to represent the time periods corresponding to different states in segments, with each segment set to a duration of 60 seconds. After splicing, the target current waveform data is formed, with a waveform duration of 6 seconds. The time step is 1 second, and the final output is the target current waveform table. Then, the corresponding PWM duty cycle parameter is set according to the current value at each time point in the target current waveform table. The PWM control parameter is calculated using a linear proportional conversion method. The conversion formula is duty cycle D = I / I_max × 100%, where I is the target current value, I_max is fixed at 10A, the duty cycle range is limited to 20% to 95%, and the PWM signal frequency is set to 20kHz. The PWM control parameter is input to the programmable dynamic load component controlled by the digital signal processor. This component drives the power MOS array in the electronic load unit to adjust the conduction time of the resistor channel in real time, so as to realize the current simulation output, which is equivalent to the working current of the vacuum cleaner host. The current signal output by the electronic load is recorded by the high-precision sampling module and used as the equivalent current signal for the lithium battery loading test in the subsequent step S2.

[0055] This invention significantly improves the fit and representativeness of the test current signal to real-world application scenarios by capturing, analyzing, and reconstructing the current behavior of the vacuum cleaner's main unit throughout its actual operation. By collecting current data covering different suction levels, load states, and operating modes, it ensures that the raw data includes key features frequently encountered in real-world operation, such as start-up impacts, current spikes, frequency changes, and stall responses. The timing modeling process not only preserves the switching rhythm and duration between different suction levels but also reconstructs the dynamic load rhythm during vacuum cleaner operation, solving the problem of current curves being disconnected from operating conditions in traditional testing. During parameter configuration, a high-precision PWM duty cycle output is used as the control signal driving method, ensuring that the equivalent current signal remains consistent with actual operation in terms of waveform amplitude, rate of change, and response delay, avoiding deviations in load strain assessment under static constant current conditions. The overall process abstracts the vacuum cleaner's workload from the physical environment into a digitally controllable standard input signal, providing a stable, accurate, and reproducible current input basis for subsequent thermal response, control behavior, impedance changes, and overall machine modeling, thereby enhancing the consistency and reliability of the entire test system in multi-source collaborative simulation.

[0056] Preferably, step S2 includes the following steps:

[0057] Step S21: Install miniature thermocouple sensors inside the lithium battery pack to obtain a multi-point temperature monitoring network;

[0058] Step S22: Based on the multi-point temperature monitoring network, the lithium battery is dynamically loaded using the equivalent current signal to obtain multi-point temperature data;

[0059] Step S23: Perform spatial gradient analysis on the multi-point temperature data to obtain temperature distribution characteristic parameters;

[0060] Step S24: Perform three-dimensional thermal field interpolation calculation on the convolutional neural network based on the temperature distribution feature parameters to obtain continuous thermal field distribution data;

[0061] Step S25: Perform three-dimensional visualization rendering based on continuous thermal field distribution data to obtain a three-dimensional thermal field distribution map;

[0062] Step S26: Calculate the thermal field evolution prediction based on continuous thermal field distribution data to obtain the temperature rise trend prediction result.

[0063] In this embodiment of the invention, after the lithium battery pack casing is disassembled, eight K-type miniature thermocouple sensors are fixedly placed at key thermally sensitive locations within the lithium battery pack using a positioning fixture. Specifically, these placement points are the center of the top of the battery cell, the bottom edge of the battery cell, the middle of the battery pack, the positive busbar, the negative busbar, the center of the BMS circuit board, the inner front wall of the casing, and the top corner of the casing. The sensor installation error must not exceed ±1mm. The sensor leads are uniformly routed through a high-temperature sleeve into the acquisition module, forming a complete multi-point temperature monitoring network. The equivalent current signal output in step S14 is connected, and the signal is loaded onto the lithium battery power interface. The loading process continues for 300 seconds, with the sensor sampling frequency set to 5Hz. The time-series temperature values ​​of all measurement points are recorded using a high-precision temperature acquisition module, forming a multi-point temperature dataset. Subsequently, the positions of each sensor point are modeled using a three-dimensional coordinate mapping method, and the temperature difference and physical distance between each pair of coordinates are calculated based on adjacent measurement points. The central difference method is used to solve for the temperature gradient in each direction. The temperature distribution is measured, and the local temperature change rate is represented by the square root of the sum of the squares of the temperature gradients at each point, generating a temperature distribution characteristic parameter table. Based on this parameter table, the spatial distribution and temperature gradient are jointly fitted. The three-dimensional space is divided into a voxel grid with a spacing of 0.5 cm using an equidistant grid interpolation method. The temperature estimate of each voxel in the grid is calculated using spatial weighted averaging, forming a continuous thermal field distribution data matrix. The matrix data is visualized using a three-dimensional graphics rendering engine, assigning different colors to different temperature ranges (e.g., blue represents less than 30℃, and red represents greater than 60℃), constructing a complete three-dimensional thermal field distribution map, and outputting images in STL and PNG formats. Subsequently, multiple adjacent thermal field distribution matrices are compared along the time axis, and the temperature change at the same spatial coordinate point is processed by difference. A linear extrapolation method is used to predict the temperature evolution trend of each point within the next 300 seconds. Based on the prediction results, a temperature rise trend curve is generated, and the maximum temperature rise slope, average temperature rise slope, and temperature peak time point of each measuring point are included in the prediction results table.

[0064] This invention comprehensively enhances the ability to distinguish and assess the thermal response behavior of lithium battery packs by constructing a refined, multi-node internal temperature sensing network. The deployment of miniature thermocouple sensors covers the core area of ​​the battery cell, BMS control nodes, and the heat accumulation surface of the casing, enabling temperature acquisition with high spatial distribution accuracy and rapid response characteristics, providing multi-source data support for subsequent thermal behavior reconstruction. By dynamically loading equivalent current signals, the thermal response behavior is given a load basis consistent with the actual usage state of a vacuum cleaner, ensuring that the temperature change process can accurately reflect the heat generation path and conduction process under complex operating conditions. Spatial gradient analysis further reveals the direction of heat diffusion and the laws of local temperature difference changes, providing quantitative parameter basis for the calculation of the three-dimensional thermal field structure. During the thermal field calculation and rendering process, through continuum interpolation and visualization processing, the heat distribution is expanded from discrete measurement points to complete spatial coverage, intuitively presenting the internal heat accumulation area and the migration paths of hot and cold spots, solving the problem that traditional external thermal measurement methods cannot perceive internal heat sources. Ultimately, by predicting the thermal trend evolution, the direction and rate of temperature development can be quantitatively extrapolated, potential thermal runaway risk areas can be identified in advance, and key judgment basis can be provided for overall load assessment, risk classification and protection mechanism triggering, effectively improving the predictive ability and management level of lithium battery thermal safety performance.

[0065] Preferably, step S3 includes the following steps:

[0066] Step S31: Monitor the BMS circuit board communication interface signal based on the three-dimensional thermal field distribution map to obtain the original communication data stream;

[0067] Step S32: Extract and analyze protocol features from the original communication data stream to obtain the communication protocol type identification result;

[0068] Step S33: Configure the decoder of the communication identification sub-component based on the protocol type identification result to obtain the protocol decoder parameters;

[0069] Step S34: Parse the frame structure of the original communication data stream based on the protocol decoder parameters to obtain command frame structure data;

[0070] Step S35: Perform functional classification analysis on the command frame structure data to obtain the BMS functional instruction set;

[0071] Step S36: Analyze the response mechanism of the protocol decoding sub-component based on the BMS functional instruction set to obtain the BMS communication response characteristics;

[0072] Step S37: Based on the BMS communication response characteristics, integrate the protocol type, frame format, instruction set, response mechanism, and timing requirements to obtain the BMS communication protocol parameters.

[0073] In this embodiment of the invention, a digital signal acquisition probe is first connected to the CAN communication interface of the BMS circuit board. Data acquisition is initiated when the thermal field distribution constructed in the lithium battery loading step S2 is in the high-temperature equilibrium stage (the cell surface temperature reaches 50℃±2℃). All communication signals are continuously acquired at a time resolution of 100μs within a 300-second test cycle and stored in real time as the raw communication data stream. In step S32, the raw communication data stream is processed by time-domain segmentation, with each segment being at least 10ms in length. The message frame start bit, identifier field, and frame are extracted using a bitstream grouping parsing method. The length field, based on the byte alignment structure between fields and the presence of the cyclic redundancy check field, determines the communication protocol to be CAN2.0B type, and identifies that the standard frame (11-bit identifier) ​​and extended frame (29-bit identifier) ​​distribution ratio in the frame structure of this protocol exceeds 70%; in step S33, based on the CAN2.0B protocol type identification result, by configuring the embedded frame parsing module in the communication identification subcomponent, the frame synchronization byte is set to 0xAA, the start flag bit length is set to 1 bit, and the check field is set to CRC-15, thus initializing the protocol decoder parameters; In step S34, the configured communication identification sub-component is used to parse the original communication data stream by bytes, identifying the frame header, frame body, and frame tail in units of 1 byte, extracting the data field length, command byte, parameter byte, and terminator, and outputting the complete command frame structure data, with the total number of bytes per frame controlled between 8 and 13 bytes. In step S35, all command frame structure data are grouped according to the command byte values, and divided into six functional instruction sets—reading voltage, reading temperature, status query, charging control, discharging control, and fault reset—using the command number matching method, and the parameter field characteristics and trigger response fields of each type of instruction are recorded. In step S36, based on the content of the functional instruction set, the average response delay, response frame byte count, and response position distribution pattern of each type of command frame are statistically analyzed to construct the communication response characteristics corresponding to each functional instruction. For example, the response time of the temperature reading command is within 5ms, and the frame header is consistent with the command byte. In step S37, the acquired communication protocol type, frame format structure, functional instruction set content, response delay characteristics, and time interval patterns of various commands are structurally integrated, and a complete BMS communication protocol parameter set is constructed through a field mapping table.

[0074] This invention guides communication behavior acquisition through a thermal correlation triggering mechanism, combined with data structure parsing and response pattern extraction, to achieve complete identification and parameter reconstruction of the communication protocol of a vacuum cleaner lithium battery BMS, effectively breaking down the technical barriers of closed protocols and undisclosed instructions. Triggering communication monitoring in high-heat-risk areas ensures that the collected data has a highly coupled background context of load conditions and thermal environment, improving the relevance and effectiveness of protocol analysis. Structured extraction of the communication data stream avoids redundant waveform interference, achieving accurate segmentation of start bits, identifier fields, data length, and verification mechanisms, thereby clarifying the protocol type. The decoder parameter configuration process ensures complete alignment between the parsing module and the actual protocol format, avoiding parsing failures due to frame structure mismatches. Instruction classification not only reveals the BMS's functional call structure but also its working logic and control boundaries, providing a highly available behavioral template for controller simulation. Response mechanism analysis extracts key communication dynamic characteristics such as delay time, response field rules, and command trigger frequency, achieving a complete restoration of the protocol's time dimension. The final set of communication protocol parameters covers protocol types, structural rules, instruction definitions, and response strategies, forming a highly engineered interface standard that can be directly used for simulation control. This provides an accurate and comprehensive data foundation for the subsequent virtual controller to accurately reproduce the behavior of the BMS.

[0075] Preferably, in step S4, configuring the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters to simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner includes:

[0076] Based on the target current waveform data, the working state of the virtual vacuum cleaner controller is mapped to obtain the controller state transition table;

[0077] Based on the BMS communication protocol parameters, configure the communication parameters of the virtual vacuum cleaner controller to obtain the communication interface driver parameters;

[0078] Based on the controller state transition table and communication interface drive parameters, simulated instructions are generated for the BMS function instruction set to obtain a virtual control instruction sequence.

[0079] The communication control behavior of the lithium battery BMS is executed based on the virtual control command sequence to obtain the real-time response data of the BMS.

[0080] In this embodiment of the invention, the target current waveform data generated in step S1 is first imported into the control logic processing system. Using a current amplitude range division method, the current value at each moment is mapped to the corresponding vacuum cleaner operating condition state based on the current peak range (Level I: 0.4A, Level II: 4.8A, Level III: 8~12A). A state transition relationship is established based on the order of current value changes, forming a state sequence according to the "low → medium → high → low" cycle. A 300-second state transition timetable is further constructed, and the time points are mapped one-to-one with the operating states to form a controller state transition table. Subsequently, the BMS communication protocol parameters obtained in step S3 are imported into the communication parameter parser, where the communication frame format (standard frame), data field length (not exceeding 8 bytes), command header (starting with 0xA0), response frame delay threshold (less than 5ms), and communication interface standard (CAN2.0B) are set. All parameters are then converted into controller interface configuration parameters. The set of bytes, i.e., the communication interface driver parameters, is used to convert the control behavior corresponding to each state transition point into BMS function instructions through command mapping rules, based on the state transition table and communication interface driver parameters. Specifically, this includes sending a voltage read command (command byte 0xA1) in state I, sending a temperature read command (command byte 0xA2) in state II, and sending a charging prohibition command (command byte 0xA4) in state III. A complete virtual control instruction sequence is generated according to the time sequence of the controller state transition table. Each instruction contains a frame header, command byte, parameter byte, and check bit, totaling no more than 13 bytes. Subsequently, the virtual control instruction sequence is transmitted one by one through the CAN communication interface controller to the communication interface port of the lithium battery BMS circuit board at a rate of 500kbps. The response frame data returned by the BMS is recorded in real time. Each response frame contains a response identifier, status byte, and data field, forming a complete BMS real-time response data sequence.

[0081] This invention constructs a virtual control mechanism highly consistent with the operating behavior of a vacuum cleaner, enabling the lithium battery BMS to receive communication commands completely identical to the actual overall machine control logic during testing, thus realistically reflecting its responsiveness and control compatibility. A working condition mapping is established based on the target current waveform, giving the controller state switching clear current criteria and timing characteristics, effectively ensuring synchronized linkage between control behavior and electrical load status. Through parameterized configuration of the BMS communication protocol, the virtual controller possesses complete communication capabilities, sending structurally valid, content-effective, and parseable responses, avoiding command rejection or abnormal response due to protocol deviations during testing. The automatic generation mechanism of simulated commands constructs a complete control trigger chain, covering functions such as charge / discharge start-up, status query, and fault detection, ensuring that the communication content covers all functional modules required for actual operation. The response data of the BMS after receiving commands not only reflects the completeness of its processing logic but also demonstrates its delay characteristics and load feedback capabilities under different states, providing crucial data support for subsequent impedance adjustment and overall machine behavior modeling. The overall process achieves closed-loop simulation and realistic replication of the communication link, significantly improving the accuracy and practicality of the test system in terms of control behavior reproduction, response capability assessment, and protocol compatibility verification.

[0082] Preferably, the load characteristics of driving the real-time variable impedance component to respond to each state in step S4 include:

[0083] The validity of the BMS real-time response data was verified and analyzed to obtain the communication protocol compatibility assessment results.

[0084] Based on the communication protocol compatibility assessment results, the parameters of the virtual vacuum cleaner controller were optimized and adjusted to obtain the optimized controller configuration parameters.

[0085] Extracting state load characteristic data for each operating condition based on BMS real-time response data;

[0086] The impedance regulation control unit is configured based on the state load characteristic data to obtain the impedance regulation control parameters.

[0087] A dynamic impedance adjustment signal is obtained by driving and controlling a real-time variable impedance component based on impedance adjustment control parameters.

[0088] The real-time impedance response characteristics of a variable impedance component are obtained by measuring the response characteristics of the variable impedance component based on the dynamic impedance adjustment signal.

[0089] A quality assessment analysis of the real-time impedance response characteristics is performed to obtain the impedance response accuracy assessment results;

[0090] Based on the impedance response accuracy assessment results, the real-time impedance response characteristics of each operating condition are corrected and integrated to obtain variable impedance response data.

[0091] In this embodiment of the invention, the BMS real-time response data is first imported into the data verification module. Each response frame is then checked sequentially for field comparison, data field format comparison, and timestamp matching. If the response delay does not exceed 5ms, the frame format is consistent with the preset format, the number of data bytes is within 8 bytes, and the CRC check passes, the response is considered valid. Otherwise, the abnormal frame number is recorded and added to the protocol compatibility defect statistics table. Finally, the communication protocol compatibility evaluation result is output. Based on this evaluation result, if the number of abnormal frames exceeds 5%, the communication interval period in the virtual vacuum cleaner controller is adjusted (increased from 20ms to 3ms). The command repetition count was adjusted from 1 to 3 times, and the frame header identification mask parameters were reconfigured. The modified parameters were then saved as optimized controller configuration parameters. Subsequently, based on all valid response data and the operating condition status markers in the target current waveform, the voltage, current, and temperature data returned by the BMS under each operating condition were extracted in chronological order. The average power change rate and response fluctuation amplitude within the state period were calculated to form the state load characteristic data corresponding to that operating condition. Then, the state load characteristic data was input to the impedance adjustment control unit, and the impedance change range was set from 0.1Ω to 2Ω. Within the range of 0Ω, with a variation accuracy of 0.01Ω and a variation frequency of 1Hz to 500Hz, impedance adjustment control parameters are calculated, including base resistance setting, adjustment amplitude, frequency, and phase offset angle. These parameters are transmitted via the control bus to the real-time variable impedance component (composed of a parallel inductor, capacitor network, and MOSFET array), triggering a real-time drive command. The controller generates a dynamic impedance adjustment signal at a 1ms cycle and controls the device to operate according to the set resistance value. Subsequently, a 10mV excitation voltage is applied across the load through the impedance monitoring unit, and the response current is measured. The instantaneous equivalent impedance is calculated according to Ohm's law. The measurement value is recorded every 1ms for 300 seconds to form a complete real-time impedance response characteristic. The maximum error, fluctuation rate and phase lag analysis are performed on all response data. If the error is less than 0.05Ω, the phase lag is less than 15° and the settling time is less than 3 seconds, it is evaluated as a high-precision response, and the impedance response accuracy evaluation result is formed. Finally, the impedance response data that have passed the evaluation under all conditions are corrected according to the operating conditions. The weighted smoothing method is used to adjust the outliers. The impedance response data of each time period and each range are integrated to form variable impedance response data covering the complete operating cycle.

[0092] This invention achieves high-precision reproduction and evaluation of the load characteristics of a vacuum cleaner's lithium battery under real-world control behavior by constructing a dynamic response mechanism that integrates communication-driven, state-linked, and impedance-adjusted features. Communication effectiveness verification ensures complete consistency in the protocol structure of command transmission and feedback data during testing, improving control link stability and data availability. The parameter optimization mechanism enables the virtual controller to adapt to different types of BMS, thereby enhancing the protocol compatibility and practical access capabilities of the test platform. The close integration of state load feature extraction and impedance adjustment control achieves dynamic matching of impedance amplitude, response speed, and fluctuation frequency under different operating conditions, allowing the variable impedance component to realistically simulate the complex electrical behavior of the motor, main control board, and power components in the entire system. The impedance response formed under dynamic adjustment signal control possesses temporal continuity and state discrimination, providing high-density, multi-dimensional raw data for subsequent behavioral modeling. The quality assessment of response characteristics ensures that the impedance output meets the requirements of system modeling and operating condition playback in terms of accuracy, stability, and timing synchronization. Ultimately, through multi-state data integration and error calibration, the generated variable impedance response data comprehensively reflects the transient load impact and electrical load characteristics of the vacuum cleaner on the lithium battery under different control modes, providing a reliable foundation for the accurate construction of the whole machine load behavior model and thermal risk prediction.

[0093] Preferably, step S5 includes the following steps:

[0094] Step S51: Perform feature engineering on the target current waveform data to obtain the current waveform feature vector;

[0095] Step S52: Normalize the variable impedance response data to obtain standardized impedance characteristic data;

[0096] Step S53: Construct a dataset based on the current waveform feature vector and standardized impedance feature data to obtain a time series training dataset;

[0097] Step S54: Classify and label the operating status of the vacuum cleaner based on the time series training dataset to obtain the operating status label data;

[0098] Step S55: Construct a labeled training dataset based on the running status label data;

[0099] Step S56: Select time series models based on the labeled training dataset, predict and verify the vacuum cleaner load behavior, and generate a whole machine load behavior model.

[0100] In this embodiment of the invention, firstly, in step S51, feature engineering processing is performed on the acquired target current waveform data. A fixed-length sliding window method is used to divide the current waveform within 300 seconds into 300 windows per second. For each current data point within a window (sampling frequency of 10kHz, totaling 3000 points), the maximum value, minimum value, mean, standard deviation, rise time duration, fall time duration, number of cycles, and maximum rate of change are calculated sequentially, extracting a total of 8 feature indicators to form a current waveform feature vector of length 300×8. Subsequently, in step S52, normalization processing is performed on the integrated variable impedance response data, using the maximum and minimum value standard... The normalization method scales the resistance range of each impedance change curve to between 0 and 1, keeping the relative impedance change ratio constant within each operating cycle. Simultaneously, it segments each normalized impedance curve and statistically analyzes its maximum impedance, minimum impedance, average impedance, change amplitude, number of oscillations, and average rate of change per cycle, constructing six types of impedance characteristic indicators to form standardized impedance characteristic data of length 300×6. In step S53, the current waveform feature vector and the standardized impedance characteristic data are aligned and concatenated along the time dimension, forming a 14-dimensional feature vector corresponding to the features of each second, and then combined in chronological order to form a time series training dataset of length 300×14. In step S5... In step 4, based on the three-state switching sequence and time control strategy set in step S1 during vacuum cleaner operation, state I is labeled "0", state II is labeled "1", and state III is labeled "2". The operating state is then labeled for each second within a 300-second cycle according to the state transition table, obtaining the corresponding operating state label data. In step S55, the 14-dimensional features in the time-series training dataset are combined with the operating state labels to form supervised training samples. Each sample contains 14-dimensional input data within one second and the corresponding operating state label, constructing a complete labeled training dataset with a total of 300 samples. In step S56, a structure depth of 3 is selected. The time-series recognition structure of the layer serves as the operating state prediction module. The training process is controlled with a batch size of 16 and a learning round of 100. It inputs a labeled training dataset and calculates the classification error between the state prediction and the actual label in real time. The connection weights are continuously adjusted according to the principle of minimum error, and the structural parameters when the error is minimum are saved. Finally, in the validation data segment, the current operating state of the vacuum cleaner is predicted second by second at an input frequency of once per second. The predicted state is re-encoded into a state sequence in chronological order. A state-current-impedance ternary mapping table is established through the time correspondence between the state sequence and the original current waveform and impedance change data, generating a whole machine load behavior model covering the complete operating cycle.

[0101] This invention achieves high-precision modeling and intelligent identification of the electrical response behavior of vacuum cleaner lithium batteries under complex dynamic operating conditions by constructing a load modeling process centered on feature extraction, time-series alignment, state labeling, and behavior prediction. Feature vector extraction of the current waveform enhances the ability to express current details such as startup abrupt changes, load switching, and steady-state variations, providing a rich foundation of input information for subsequent state identification. Impedance data normalization enhances the comparability of impedance response amplitudes under different operating conditions, giving electrical load characteristics a unified dimension and range, facilitating fusion analysis. The time-series structure after feature data integration preserves the temporal logic and operating condition change rhythm during vacuum cleaner operation, providing continuous support for operating condition classification and behavior reconstruction. The establishment of state labels establishes a direct mapping relationship between abstract data and physical operating states, realizing data-driven visualization and controllability of operating conditions. The model screening and behavior prediction mechanism optimizes the identification capabilities of various time structures by judging the accuracy of operating conditions. The final generated whole machine load behavior model has the ability to fuse three-dimensional information of current amplitude, impedance characteristics and operating status, providing data support and logical basis for subsequent thermal runaway risk assessment, control strategy matching and intelligent scheduling decision-making.

[0102] Preferably, step S56 includes the following steps:

[0103] Step S561: Design a long short-term memory network architecture based on the labeled training dataset and obtain the LSTM network structure parameters;

[0104] Step S562: Train and optimize the time series based on the LSTM network structure parameters to obtain the parameters of the trained LSTM model;

[0105] Step S563: Construct a mapping relationship between the operating state and the current waveform based on the trained LSTM model parameters to obtain the state-current mapping function;

[0106] Step S564: Based on the state-current mapping function, compare and analyze the Transformer architecture to obtain the Transformer model performance data;

[0107] Step S565: Evaluate the performance of the LSTM model and the Transformer model based on the comparison of LSTM network structure parameters and Transformer model performance data, and obtain the optimal time series model;

[0108] Step S566: Based on the optimal time series model, predict and verify the vacuum cleaner load behavior to obtain load behavior prediction accuracy data;

[0109] Step S567: Based on the load behavior prediction accuracy data integration and optimization mapping relationship, obtain the whole machine load behavior model.

[0110] In this embodiment of the invention, firstly, in step S561, a Long Short-Term Memory (LSTM) architecture is set based on the labeled training dataset constructed in step S55. The input dimension is set to 14, corresponding to the number of time features of current and impedance. The number of hidden units is set to 128, the network depth is set to 2 layers, and the time step is set to 10, that is, data is used as an input sequence every 10 seconds, and the output is the running state label corresponding to the current moment. The weight matrices of the forget gate, input gate, and output gate in the gating mechanism are initialized with a normal distribution, and the bias value is set to a zero vector to obtain the LSTM network structure parameters. In step S562, the training process is executed based on these structure parameters. The first 80% of the samples in the labeled training dataset are used as training data, and the last 20% are used as training data. To validate the data, the training batch size was set to 16, and the maximum number of training epochs was set to 100. The cross-entropy function was used to evaluate the prediction error. After each training epoch, the connection weights were adjusted based on the validation set accuracy. The weight matrix with the smallest output error was used as the parameters of the trained LSTM model. In step S563, the trained parameters were embedded in the recognition module. This module inputs a 14-dimensional feature vector per second, outputs a state number, and constructs a mapping table with historical state records. A state-current mapping function is obtained according to the time series order, and this function serves as a quantitative expression of the relationship between the vacuum cleaner's operating state and the current waveform. In step S564, a Transformer structure was constructed using the same training dataset. The attention heads are set to 4, the hidden layer dimension to 128, and the feedforward network output dimension to 256. The sequence length remains the same as LSTM, and the time step is still 10. The same input data is used for training, and the accuracy, average prediction latency, training time, and parameter size on the validation set are recorded to obtain the Transformer model performance data. In step S565, the validation accuracy (92.7%), prediction latency (0.8 seconds), and training time (56 minutes) under the LSTM structure are compared with those under the Transformer structure (90.2%), prediction latency (1.3 seconds), and training time (73 minutes). Prioritizing prediction accuracy, the LSTM structure is selected for the current stage. The optimal time series identification method serves as the basic framework for operational state prediction. In step S566, based on the trained LSTM structure, the operational state sequence is predicted by sliding across the complete test data and compared with the real state labels hourly. The number of prediction hits and errors is counted at each time point, and the overall accuracy, state transition identification error, and prediction lag time are calculated to obtain the load behavior prediction accuracy data. In step S567, the prediction accuracy data is used as a weighting coefficient to fuse the state-current mapping function with the impedance response sequence. Periods with large prediction deviations are eliminated, and the boundary times are linearly smoothed to output a whole-machine load behavior model with clear state boundaries, continuous time mapping relationship, and multi-dimensional load characteristics.

[0111] This invention, by introducing a timing recognition structure and mapping function construction mechanism, achieves the extraction, comparison, and refined modeling of the complex nonlinear relationship between the vacuum cleaner's operating state and current waveform, comprehensively improving the accuracy and generalization ability of the whole machine load behavior model in dynamic prediction and practical adaptation. The Long Short-Term Memory (LSTM) structure possesses the ability to extract time-dependent features, enabling the model to capture remote influencing factors and state transition patterns in load changes, effectively compensating for the neglect of dynamic response details in static modeling. Optimization of structural parameters during training enhances the network's perception of the switching rhythm and current amplitude fluctuations of different operating conditions, improving the sensitivity of identifying key operating conditions (such as stalled operation and high-speed gears). The establishment of the mapping function directly binds the model output to electrical behavior, providing a callable and quantifiable basis for state recognition for the system controller and protection mechanism. Performance comparison analysis of the Transformer structure broadens the selection range of timing structures, providing structural references and selection standards for model deployment in subsequent system integration. Optimal structures are selected through multi-dimensional performance evaluation of accuracy, latency, and complexity, ensuring that the load prediction function achieves a balance between response speed and system resource consumption. Ultimately, the prediction accuracy index on the verification data forms an evaluation feedback loop, which helps to complete the correction and reconstruction of the mapping relationship, enabling the whole machine load behavior model to have dynamic update, structural self-consistency and multi-source fusion capabilities, providing a basic guarantee for the simulation accuracy, response time and intelligent control of the vacuum cleaner battery testing system.

[0112] Preferably, step S6 includes the following steps:

[0113] Step S61: Based on the temperature rise trend prediction results, conduct a thermal runaway risk assessment on the overall load behavior model to obtain multi-level risk warning signals;

[0114] Step S62: Design a thermal runaway early warning mechanism based on multi-level risk warning signals, and select a protection strategy to obtain a graded protection control strategy;

[0115] Step S63: Perform load reduction control on the programmable dynamic load control component based on the hierarchical protection control strategy to obtain the load adjustment execution signal;

[0116] Step S64: Perform coordinated protection control on the virtual vacuum cleaner controller based on the load adjustment execution signal to obtain the system protection execution result;

[0117] Step S65: Perform real-time effect monitoring and analysis on the system protection execution results to obtain protection response performance evaluation data.

[0118] In this embodiment of the invention, firstly, in step S61, the temperature rise trend prediction curve constructed in step S2 and the whole machine load behavior model formed in step S56 are time-aligned. The operating state, current amplitude, impedance characteristics, and predicted temperature value corresponding to each second are selected. A three-stage risk judgment method is used to set the thermal runaway risk threshold. Specifically, when the predicted temperature is below 60℃, it is a low-risk zone; 60℃ to 75℃ is a medium-risk zone; and above 75℃ is a high-risk zone. Combined with the temperature rise rate threshold (2℃ / min), a temperature rise anomaly judgment standard is set. Based on the above, green, yellow, and red risk warning signals are output respectively, forming a multi-level risk. A warning signal sequence is generated. In step S62, the warning signals are input to the protection strategy allocation module, and the mapping relationship between the warning level and the response action is set: the green signal maintains normal operation without intervention, the yellow signal corresponds to load limiting control and communication interval extension, and the red signal corresponds to forced load reduction control and communication interruption protection. These are then combined into a hierarchical protection control strategy. Each strategy includes a load limiting value, controller pause time, communication prohibition flag, and recovery judgment time point. In step S63, the hierarchical protection control strategy is output to the programmable dynamic load control component, and the control system recalculates the duty cycle sequence of the PWM control signal according to the limiting value set in the strategy. The system executes a load regulation signal by updating the new PWM sequence to the electronic load control terminal in real time at a frequency of 20kHz. Specifically, the maximum current amplitude is controlled below 8A, and the waveform fluctuation rate does not exceed 2A / s. In step S64, the load regulation execution signal is synchronously transmitted to the virtual vacuum cleaner controller via the control bus. The controller suspends the control state transition corresponding to the high load condition, blocks the generation of the charging start command, and extends all state switching cycles to 1.5 times the original cycle. In the actual communication execution module, the transmission of high-frequency communication commands is temporarily suspended through command masking. Ultimately, the controller operates in three categories: load limiting, frequency limiting, and voltage reduction. The system responds to the action and outputs the system protection execution result. In step S65, the real-time performance monitoring module is activated to monitor and record three indicators: load response current, battery temperature change rate, and communication data packet transmission success rate. The temperature rise and fall rate threshold is set to 1.5℃ / min, current stability fluctuation is not more than ±0.5A, and communication packet loss rate is less than 1%. All indicators are evaluated continuously using a sliding window with each window period of 10 seconds. If all the above requirements are met within three consecutive windows, the protection response effect is evaluated as qualified. Otherwise, it is marked as a protection failure section, and the event number and parameter offset are recorded. Finally, the protection response performance evaluation data is generated.

[0119] This invention deeply integrates temperature rise prediction results with load behavior modeling results to construct a dynamic intervention system for thermal runaway with hierarchical judgment, proactive response, and real-time feedback capabilities. This achieves closed-loop protection and controllable process management for the operational safety of vacuum cleaner lithium batteries. The generation of multi-level risk warning signals eliminates reliance on a single temperature threshold for thermal runaway judgment, instead integrating multiple factors such as temperature level, heating rate, and load status to achieve early quantitative identification of potential thermal anomalies. The hierarchical protection strategy enables orderly progression of risk response measures, allowing load reduction control, communication throttling, and functional shielding to be activated as needed, avoiding excessive intervention or response delays. The dynamic load control module precisely adjusts the amplitude and frequency of the output waveform, effectively reducing the intensity of heat source input and providing buffer time for heat dissipation and system self-cooling. Under the coordination of the strategy, the virtual controller synchronously limits the issuance of high-load operating condition commands, ensuring that the battery operates within a controllable range. The system protection execution results are continuously evaluated through multi-dimensional monitoring indicators, which enables the thermal control strategy to not only have triggering capabilities, but also result verification and self-correction capabilities. This improves the system's temperature control regulation capability, operational stability and test safety under sudden operating conditions, and constructs a lithium battery thermal runaway protection mechanism characterized by prediction-driven, data-supported and linkage response.

[0120] Preferably, the present invention also provides a testing system for a vacuum cleaner lithium battery, used to perform the above-described testing method for a vacuum cleaner lithium battery, the testing system comprising:

[0121] The dynamic load module is used to collect the operating current of the vacuum cleaner under various operating conditions and obtain the target current waveform data; based on the target current waveform data, the programmable dynamic load control component is configured, and the electronic load unit is driven to output the equivalent current signal.

[0122] The thermal field monitoring module is used to load the equivalent current signal of the vacuum cleaner's lithium battery to collect the temperature of the battery cell surface, BMS circuit board and shell in real time, and obtain multi-point temperature data; based on the multi-point temperature data, a thermal field model is constructed to generate a three-dimensional thermal field distribution map and temperature rise trend prediction results.

[0123] The communication identification module is used to collect lithium battery management communication behavior based on the three-dimensional thermal field distribution map, identify the protocol type and parse the command frame structure to obtain BMS communication protocol parameters;

[0124] The virtual control module is used to configure the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner, and drive the real-time variable impedance component to respond to the load characteristics under each state to obtain variable impedance response data.

[0125] The load modeling module is used to construct a mapping relationship between the vacuum cleaner's operating state and the current waveform based on the target current waveform data and variable impedance response data, and generate a whole machine load behavior model.

[0126] The safety protection module is used to determine the risk of thermal runaway based on the temperature rise trend prediction results of the whole machine load behavior model, and to perform overload protection processing by controlling the programmable dynamic load control component and the virtual vacuum cleaner controller.

[0127] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is not limited by the foregoing description. Thus, all changes falling within the meaning and scope of the equivalents of the application are intended to be included within the scope of the invention.

[0128] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A testing method for a lithium battery in a vacuum cleaner, characterized in that, Includes the following steps: Step S1: Collect the operating current of the vacuum cleaner under various operating conditions to obtain target current waveform data; configure the programmable dynamic load control component based on the target current waveform data, and drive the electronic load unit of the programmable dynamic load control component to output an equivalent current signal; Step S1 includes the following steps: Step S11: Perform multi-condition operation tests on the vacuum cleaner main unit to obtain current characteristic data for each setting; Step S12: Perform time-series modeling of the vacuum cleaner's operating mode based on the current characteristic data of each gear to obtain the target current waveform data; Step S13: Configure the programmable dynamic load component parameters based on the target current waveform data to obtain the PWM control parameters; Step S14: Drive and control the electronic load unit in real time based on PWM control parameters to obtain the equivalent current signal; Step S2: Load an equivalent current signal using the vacuum cleaner's lithium battery and collect multi-point temperature data in real time; construct a thermal field model based on the multi-point temperature data, simulate and calculate the temperature field distribution data through the thermal field model, and generate a temperature rise trend prediction result; Step S3: Collect lithium battery management communication behavior based on the three-dimensional thermal field distribution map, identify the protocol type and parse the command frame structure to obtain BMS communication protocol parameters; Step S4: Configure the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner, and drive the real-time variable impedance component to respond to the load characteristics under each state to obtain variable impedance response data. Step S5: Based on the target current waveform data and variable impedance response data, construct the mapping relationship between the vacuum cleaner's operating state and the current waveform to generate a whole-machine load behavior model; wherein, step S5 includes the following steps: Step S51: Perform feature engineering on the target current waveform data to obtain the current waveform feature vector; Step S52: Normalize the variable impedance response data to obtain standardized impedance characteristic data; Step S53: Construct a dataset based on the current waveform feature vector and standardized impedance feature data to obtain a time series training dataset; Step S54: Classify and label the operating status of the vacuum cleaner based on the time series training dataset to obtain the operating status label data; Step S55: Construct a labeled training dataset based on the running status label data; Step S56: Based on the labeled training dataset, select time series models and predict and verify the vacuum cleaner load behavior to generate a whole machine load behavior model. Specifically, input the labeled training dataset and calculate the classification error between the state prediction and the actual label in real time. Continuously adjust the connection weights according to the minimum error principle and save the structural parameters when the error is minimum. In the verification data segment, predict the current operating state of the vacuum cleaner every second at an input frequency of once per second, and re-encode the predicted state into a state sequence in chronological order. Establish a state-current-impedance ternary mapping table through the time correspondence between the state sequence and the original current waveform and impedance change data to generate a whole machine load behavior model covering the complete working cycle. Step S6: Based on the temperature rise trend prediction results, determine the risk of thermal runaway in the overall machine load behavior model, send preset protection commands by controlling the virtual vacuum cleaner controller, and then reduce the output current through the programmable dynamic load control component.

2. The testing method for a vacuum cleaner lithium battery according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Install miniature thermocouple sensors inside the lithium battery pack to obtain a multi-point temperature monitoring network; Step S22: Based on the multi-point temperature monitoring network, the lithium battery is dynamically loaded using the equivalent current signal to obtain multi-point temperature data; Step S23: Perform spatial gradient analysis on the multi-point temperature data to obtain temperature distribution characteristic parameters; Step S24: Perform three-dimensional thermal field interpolation calculation on the convolutional neural network based on the temperature distribution feature parameters to obtain continuous thermal field distribution data; Step S25: Perform three-dimensional visualization rendering based on continuous thermal field distribution data to obtain a three-dimensional thermal field distribution map; Step S26: Calculate the thermal field evolution prediction based on continuous thermal field distribution data to obtain the temperature rise trend prediction result.

3. The test method for a vacuum cleaner lithium battery according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Monitor the BMS circuit board communication interface signal based on the three-dimensional thermal field distribution map to obtain the original communication data stream; Step S32: Extract and analyze protocol features from the original communication data stream to obtain the communication protocol type identification result; Step S33: Configure the decoder of the communication identification sub-component based on the protocol type identification result to obtain the protocol decoder parameters; Step S34: Parse the frame structure of the original communication data stream based on the protocol decoder parameters to obtain command frame structure data; Step S35: Perform functional classification analysis on the command frame structure data to obtain the BMS functional instruction set; Step S36: Analyze the response mechanism of the protocol decoding sub-component based on the BMS functional instruction set to obtain the BMS communication response characteristics; Step S37: Based on the BMS communication response characteristics, integrate the protocol type, frame format, instruction set, response mechanism, and timing requirements to obtain the BMS communication protocol parameters.

4. The test method for a vacuum cleaner lithium battery according to claim 1, characterized in that, Step S4 involves configuring the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulating the communication control behavior of the lithium battery under actual vacuum cleaner operating conditions, including: Based on the target current waveform data, the working state of the virtual vacuum cleaner controller is mapped to obtain the controller state transition table; Based on the BMS communication protocol parameters, configure the communication parameters of the virtual vacuum cleaner controller to obtain the communication interface driver parameters; Based on the controller state transition table and communication interface drive parameters, simulated instructions are generated for the BMS function instruction set to obtain a virtual control instruction sequence. The communication control behavior of the lithium battery BMS is executed based on the virtual control command sequence to obtain the real-time response data of the BMS.

5. The test method for a vacuum cleaner lithium battery according to claim 1, characterized in that, The load characteristics of the real-time variable impedance component in each state during step S4 include: The validity of the BMS real-time response data was verified and analyzed to obtain the communication protocol compatibility assessment results. Based on the communication protocol compatibility assessment results, the parameters of the virtual vacuum cleaner controller were optimized and adjusted to obtain the optimized controller configuration parameters. Extracting state load characteristic data for each operating condition based on BMS real-time response data; The impedance regulation control unit is configured based on the state load characteristic data to obtain the impedance regulation control parameters. A dynamic impedance adjustment signal is obtained by driving and controlling a real-time variable impedance component based on impedance adjustment control parameters. The real-time impedance response characteristics of a variable impedance component are obtained by measuring the response characteristics of the variable impedance component based on the dynamic impedance adjustment signal. A quality assessment analysis of the real-time impedance response characteristics is performed to obtain the impedance response accuracy assessment results; Based on the impedance response accuracy assessment results, the real-time impedance response characteristics of each operating condition are corrected and integrated to obtain variable impedance response data.

6. The test method for a vacuum cleaner lithium battery according to claim 1, characterized in that, Step S56 includes the following steps: Step S561: Design a long short-term memory network architecture based on the labeled training dataset and obtain the LSTM network structure parameters; Step S562: Train and optimize the time series based on the LSTM network structure parameters to obtain the parameters of the trained LSTM model; Step S563: Construct a mapping relationship between the operating state and the current waveform based on the trained LSTM model parameters to obtain the state-current mapping function; Step S564: Based on the state-current mapping function, compare and analyze the Transformer architecture to obtain the Transformer model performance data; Step S565: Evaluate the performance of the LSTM model and the Transformer model based on the comparison of LSTM network structure parameters and Transformer model performance data, and obtain the optimal time series model; Step S566: Based on the optimal time series model, predict and verify the vacuum cleaner load behavior to obtain load behavior prediction accuracy data; Step S567: Based on the load behavior prediction accuracy data integration and optimization mapping relationship, obtain the whole machine load behavior model.

7. The test method for a vacuum cleaner lithium battery according to claim 1, characterized in that, Step S6 includes the following steps: Step S61: Based on the temperature rise trend prediction results, conduct a thermal runaway risk assessment on the overall load behavior model to obtain multi-level risk warning signals; Step S62: Design a thermal runaway early warning mechanism based on multi-level risk warning signals, and select a protection strategy to obtain a graded protection control strategy; Step S63: Perform load reduction control on the programmable dynamic load control component based on the hierarchical protection control strategy to obtain the load adjustment execution signal; Step S64: Perform coordinated protection control on the virtual vacuum cleaner controller based on the load adjustment execution signal to obtain the system protection execution result; Step S65: Perform real-time effect monitoring and analysis on the system protection execution results to obtain protection response performance evaluation data.

8. A testing system for a vacuum cleaner lithium battery, characterized in that, For performing the test method for a vacuum cleaner lithium battery as described in claim 1, the test system for the vacuum cleaner lithium battery includes: The dynamic load module is used to collect the operating current of the vacuum cleaner under various operating conditions and obtain the target current waveform data; based on the target current waveform data, the programmable dynamic load control component is configured, and the electronic load unit of the programmable dynamic load control component is driven to output the equivalent current signal. The thermal field monitoring module is used to load an equivalent current signal using the vacuum cleaner's lithium battery and collect multi-point temperature data in real time; a thermal field model is built based on the multi-point temperature data, and the temperature field distribution data is calculated through simulation of the thermal field model, and the temperature rise trend prediction results are generated. The communication identification module is used to collect lithium battery management communication behavior based on the three-dimensional thermal field distribution map, identify the protocol type and parse the command frame structure to obtain BMS communication protocol parameters; The virtual control module is used to configure the virtual vacuum cleaner controller based on the target current waveform data and BMS communication protocol parameters, simulate the communication control behavior of the lithium battery under the actual working state of the vacuum cleaner, and drive the real-time variable impedance component to respond to the load characteristics under each state to obtain variable impedance response data. The load modeling module is used to construct a mapping relationship between the vacuum cleaner's operating state and the current waveform based on the target current waveform data and variable impedance response data, and generate a whole machine load behavior model. The safety protection module is used to determine the risk of thermal runaway from the overall load behavior model based on the temperature rise trend prediction results. It sends preset protection commands by controlling the virtual vacuum cleaner controller and then reduces the output current through the programmable dynamic load control component.

Images