An adaptive control method and system for a vehicle generator

Abstract

The application discloses a kind of self-adaptive control method and system of vehicle generator.The method is by collecting generator operating data and using spectrum analysis to solve its real-time speed;Subsequently, linear regression fitting is performed based on output current and speed, and the measured regression slope is extracted to preliminarily match the candidate power level. A multi-dimensional feature vector containing maximum current, average power and voltage change rate is constructed, and the fuzzy recognition model is used to lock the final power level attribute information. Based on the recognition result and the state of the battery, an optimization objective function is constructed to couple the energy efficiency and the life characteristics, and the optimal duty cycle is adaptively solved to adjust the charge-off period. In addition, the system realizes the physical no-load isolation of the generator by controlling the DC contactor main contact, and introduces a soft shutdown mechanism to eliminate transient high voltage surge. The application realizes non-intrusive blind recognition, effectively shields random load interference, improves energy efficiency while prolonging battery life, and has high engineering production value.

Description

Technical Field

[0001] This invention relates to the field of vehicle energy management and battery control technology, and in particular to an adaptive control method and system for vehicle generators. Background Technology

[0002] In recent years, with the development of intelligent commercial vehicles (such as heavy-duty trucks) and recreational vehicles (RVs), the popularity of parking air conditioners and high-power vehicle electrical appliances has been increasing year by year. To meet the power demand during parking, vehicles are usually equipped with additional large-capacity parking lithium battery systems. During vehicle operation or generator idling, the original vehicle alternator charges the parking lithium battery through a power regulation unit (such as a commonly used DC-DC charging module or charging relay), which has become the mainstream energy replenishment architecture.

[0003] However, in practical engineering applications, the rated power models (e.g., 80A, 130A, or 150A) of generators in different vehicle models, batches, and even vehicles that have undergone after-sales maintenance often vary significantly. Existing parking battery management systems (BMS) and related charging control equipment face several pressing technical challenges. First, fixed charging strategies lead to a contradiction between "generator overload" and "low overall energy efficiency." Most existing charging actuators use preset fixed current limits or fixed duty cycles (such as fixed charging-stop cycles) for blind charging. If the original vehicle is equipped with a low-power generator (e.g., 80A), fixed high-intensity charging will cause the generator to operate under long-term overload, leading to generator overheating, belt slippage, or even coil burnout. Conversely, if the original vehicle is equipped with a high-power generator (e.g., 150A), a conservative charging strategy will result in power waste and completely sever the linkage between the charging load and the generator's optimal mechanical-to-electrical energy conversion efficiency range, leading to serious fuel waste and low energy efficiency. Second, traditional generator status acquisition methods suffer from strong intrusion or poor anti-interference capabilities. To achieve matching between the generator and battery, some existing technologies obtain generator operating conditions by accessing the original vehicle's CAN bus. However, this method requires cracking the proprietary communication protocols of each automaker, resulting in extremely high adaptation costs and the risk of voiding the original vehicle's wiring harness warranty—a "tampering" solution. Other existing technologies attempt to infer generator speed by detecting voltage fluctuations in the DC bus or extracting voltage ripple. However, pure Fast Fourier Transform (FFT) suffers from significant computational delays and signal loss during rapid acceleration or deceleration. More critically, existing ripple analysis only measures speed; when high-power electrical appliances (such as air conditioning compressors and headlights) are randomly activated, the bus current fluctuates drastically. Existing technologies cannot distinguish whether this current surge originates from the generator's own output characteristics or from external load interference, easily leading to misjudgments of the generator's actual power capacity. Thirdly, intermittent charging control for high-power generators is highly susceptible to actively inducing load dumping high-voltage surges, posing serious safety hazards to traditional hard shutdown or purely software protection mechanisms. When implementing a periodic charging and stopping strategy, if the charging circuit is instantly cut off, the high-power generator (e.g., running at full load of 150A) instantly loses its massive electrical load. The enormous magnetic energy within the generator has nowhere to be released, generating a destructive high-voltage pulse on the DC bus. Existing BMS systems often use a "hard shutdown" method that directly cuts off the contactor, resulting in severe arcing at the main contacts. Furthermore, their protection mechanisms typically rely on purely software-based voltage threshold polling, with response times often in the millisecond (ms) range. Before the charging circuit is cut off, the high-voltage surge has already damaged the core power devices or sensitive electronic components of the original vehicle, posing a serious safety hazard.

[0004] In summary, the industry urgently needs a non-intrusive generator blind identification technology that can effectively filter out external load interference, accurately construct the generator's power profile, and perform adaptive dynamic charging scheduling that balances peak efficiency and power balance, while achieving safe and shock-free soft shutdown and no-load isolation at the underlying physical link. Summary of the Invention

[0005] In order to solve the above-mentioned technical problems in the prior art, the present invention proposes an adaptive control method and system for vehicle generators.

[0006] According to a first aspect of the present invention, an adaptive control method for a vehicle generator is proposed, comprising: S1, acquire the original operating data of the generator in the preset driving cycle, extract the output current of the generator based on the original operating data, and calculate the real-time speed of the generator. S2, based on real-time speed and output current, performs linear regression analysis, extracts the regression slope, matches the regression slope with a preset standard slope library to determine the candidate power level, and generates a trigger signal when the match is successful; S3, in response to receiving the trigger signal, constructs a multi-dimensional feature vector consisting of maximum current characteristics, average power characteristics, and voltage change rate characteristics, and uses an identification model to perform logical verification on the candidate power level based on the multi-dimensional feature vector to determine the final generator power level attribute information. S4. Based on the generator power level attribute information and the real-time status of the battery, an optimization objective function is constructed that couples the energy conversion efficiency, generator operating efficiency and battery life loss characteristics. The optimal turn-on time and turn-off time that match the generator power level are adaptively solved using a dynamic duty cycle adjustment mechanism.

[0007] In the above technical solution, by establishing a closed-loop control architecture from raw signal extraction to multi-objective optimization, it is possible to accurately identify the generator specifications in complex vehicle environments without the need to install additional sensors, and dynamically balance charging efficiency and battery life.

[0008] In some specific embodiments, before calculating the real-time rotational speed in step S1, the physical topology of the generator is identified by performing transient spectrum analysis on the original operating data, and the number of magnetic pole pairs and rectifier pulsation of the generator is locked accordingly. Real-time rotational speed is based on the formula Calculated, where, This refers to the real-time rotational speed of the generator rotor. The ripple fundamental frequency extracted from the raw operating data; The number of magnetic pole pairs; This represents the rectified pulse count.

[0009] In the above technical solution, the mechanical speed is inverted by utilizing the intrinsic electrical characteristic ripple of the generator, eliminating the dependence on vehicle communication protocols and external physical speed sensors, and reducing the hardware cost and deployment difficulty of the system.

[0010] In some specific embodiments, identifying the physical topology category of the generator includes: Step a: Upon detecting the engine start-up moment, continuously collect data at a preset sampling frequency. The data points are divided into several frames and Fourier transforms are performed to obtain the ripple fundamental frequency sequence of the start-up transient process. Step b involves extracting the first segment of the ripple fundamental frequency sequence and performing linear fitting to calculate the ripple fundamental frequency sequence. The dynamic upward slope that changes over time; Step c: Extract the last segment of the ripple fundamental frequency sequence and calculate the mean to obtain the start-up end frequency; Step d involves mapping and matching the dynamic rising slope with multiple preset slope threshold intervals, and using the start-up end frequency for interval correction to determine the physical topology category of the generator.

[0011] In the above technical solution, by capturing the dynamic evolution characteristics of the frequency at the moment of startup, the invisible physical structural parameters inside the generator are automatically identified, ensuring the universality of the control algorithm for different generator topologies.

[0012] In some specific embodiments, the expression for the dynamic ascent slope is:

[0013] In the formula, For the first The timestamp corresponding to the frame; The arithmetic mean of the timestamps of the selected frames; For the first The ripple fundamental frequency of the frame is extracted using FFT; The arithmetic mean of the selected frame fundamental frequency sequence; The expression for the start-up end frequency is:

[0014] In the formula, This represents the total number of frames captured during the startup phase. For cumulative indexing, For the first The ripple fundamental frequency corresponding to the frame, This represents the average number of frames.

[0015] In the above technical solution, a refined mathematical description model of the startup transient process is established, which ensures that high-confidence original feature quantities are obtained within an extremely short startup time, providing reliable data support for hardware category classification.

[0016] In some specific embodiments, the determination logic of step d includes: When identified satisfy At that time, the generator was determined to belong to the 6-pole single-cell category; When the dynamic rising slope is identified satisfy And the start-up end frequency satisfy At that time, the generator was determined to belong to the 8-pole single-cell category; When the dynamic rising slope is identified satisfy And the start-up end frequency satisfy At that time, the generator was determined to belong to the 6-pole double-three category; When the dynamic rising slope is identified satisfy And the start-up end frequency satisfy At that time, the generator was determined to belong to the 8-pole double-triple category; in, The first slope threshold, The second slope threshold, This is the threshold for the start-up end frequency.

[0017] In the above technical solution, the use of a multi-dimensional mapping matrix for topology determination can effectively isolate false frequency fluctuations caused by belt slippage or changes in starting resistance, and significantly improve the accuracy of topology identification.

[0018] In some specific embodiments, determining the physical topology category of the generator further includes: Calculate the ratio of the dynamic rise slope to the start-up end frequency; compare the ratio with a preset judgment threshold to verify the data consistency of the current start-up transient process and eliminate abnormal interference.

[0019] In the above technical solution, the cross-verification logic of the slope and frequency ratio eliminates outlier data interference under abnormal operating conditions, thereby enhancing the system's robustness in extreme environments and non-steady-state startup processes.

[0020] In some specific embodiments, step S2 includes: S21. When the generator speed is increasing linearly, the generator output current is synchronously collected in a preset speed step to construct a feature sequence containing several sample points. The least squares method is used to perform univariate linear regression fitting on the feature sequence to extract the regression slope. S22, calculate the Euclidean distance between the regression slope and each nominal standard slope in the preset standard slope library, select the minimum value in the Euclidean distance set, and when the minimum value is less than or equal to the preset allowable error threshold, determine the candidate power level and generate a trigger signal.

[0021] In the above technical solution, the linear coupling relationship between rotational speed and current is used for preliminary screening, and the candidate range is quickly locked through the geometric spatial distance discrimination mechanism, which significantly reduces the computational overhead of the system under high-frequency sampling.

[0022] In some specific embodiments, step S3 is characterized by: sequentially performing fuzzification processing of multi-dimensional feature vectors, fuzzy inference based on a rule base, and defuzzification operation of output variables using an identification model to obtain generator power level attribute information. By introducing a fuzzy identification model to handle uncertainties in the vehicle electrical system, discrete sensor signals are transformed into attribute judgments with probabilistic meaning, achieving soft identification of generator specifications.

[0023] In some specific embodiments, the fuzzification process in step S3 includes: using a membership function combining trapezoidal and triangular bounding boxes to map the maximum current characteristic to a low-value range. Median range High value range Fuzzy sets, and adjacent fuzzy sets are separated by a width of... The linear slope transition zone; Map the average power characteristics to a low value range Median range High value range Fuzzy set, and set the width at the decision boundary. or The membership degree of the slope transition zone.

[0024] In the above technical solution, a membership function with a transition interval is set, which effectively absorbs the numerical fluctuations caused by line voltage drop and measurement drift, and avoids frequent oscillations in the identification results at the critical point of judgment.

[0025] In some specific embodiments, fuzzy inference includes introducing non-steady-state interference interception criteria: When the rate of voltage change is detected When the generator is determined to be in a stable adjustment phase, a consistency check is performed on the candidate power level. When the voltage change rate is monitored to be at When the current sampled data is determined to be affected by external random load access interference, non-steady-state suppression logic is executed to shield non-linear interference by stopping the update of the identification results or performing weight reduction processing.

[0026] In the above technical solution, an unsteady-state interference interception criterion is established, which can automatically identify and shield the nonlinear interference generated by the high-power random load on the vehicle on the voltage signal, ensuring the uniqueness of the identification process under dynamic load environment.

[0027] In some specific embodiments, the defuzzification operation of the output variable includes: The output variable of the identification model is defined as the generator power level, and a single-point membership function is used to correspond to the output fuzzy set. ; Using the center of gravity formula Calculate the output clarity value And based on the principle of proximity, Mapped to the nearest nominal power rating, where, For the fuzzy rule index participating in the summation calculation, This is the single-point nominal value corresponding to the generator's rated current. For the first The membership weights output by the fuzzy rules.

[0028] In the above technical solution, the center of gravity method is used to transform the fuzzy reasoning results into definite nominal decisions, realizing the accurate mapping of mathematical logic to physical execution instructions, and ensuring the determinism and reliability of the decision output.

[0029] In some specific embodiments, the expression for the objective function is: In the formula, , It is a preset fixed period; Duty cycle, , , These are the weighting coefficients. Bus voltage To set the charging current, This is the generator's optimal efficiency power point. This refers to the generator's rated power. For the battery's rated capacity, The maximum allowable SOC fluctuation, This is the charging rate penalty coefficient; The constraints for optimizing the objective function are:

[0030] In the formula, For the storage battery Real-time state of charge at any given moment; This represents the lower limit of the generator's output power within its efficient operating range. This refers to the upper limit of the generator's output power within its high-efficiency operating range. The preset minimum charging on-time; The preset minimum charging shutdown time.

[0031] In the above technical solution, a comprehensive evaluation model of coupling efficiency and lifespan is constructed. By finding the optimal solution for system operation within multiple physical red lines, a dynamic balance between energy utilization efficiency and equipment health is achieved.

[0032] In some specific embodiments, the dynamic duty cycle adjustment mechanism includes retrieving a control strategy mapping table corresponding to power level attribute information to determine the duty cycle adjustment logic: If the generator power rating is 80A, a shallow charge and shallow discharge strategy is implemented: the initial search center of the duty cycle is set to 0.5, and the shutdown time is dynamically reduced according to the real-time change rate of SOC. If the generator power rating is 130A, the balancing strategy is implemented: dynamically adjust the duty cycle optimization step size in the objective function to increase the proportion of conduction time in the cycle; If the generator power rating is 150A, a deep sleep strategy is implemented: a symmetrical period with equal on-time and off-time is configured, and a soft shutdown mechanism that limits the current drop slope is implemented during the switch from on-time to off-time.

[0033] In the above technical solution, a customized charging and stopping strategy is matched to the differences in physical redundancy of different power specifications, realizing differentiated control of "prioritizing low-power charging and prioritizing high-power energy saving", thus optimizing the fuel economy of the whole vehicle.

[0034] In some specific embodiments, the method further includes: The on-time and off-time parameters are encapsulated into control signals in a preset communication protocol format and sent to the vehicle communication terminal; the battery management system parses the control signals and implements intermittent charging operation of the generator through the underlying hardware execution module.

[0035] The above technical solution establishes a closed-loop communication system from cloud strategy to vehicle-mounted underlying hardware, ensuring that complex control decisions can be translated into rapid physical responses and enabling fine-grained segmentation of the charging circuit.

[0036] In some specific embodiments, the underlying hardware execution module includes a DC contactor connected in series between the generator output terminal and the positive terminal of the battery, and a pre-drive circuit electrically connected between the DC contactor coil and the battery management system; the battery management system controls the main contacts of the DC contactor to close during the conduction time and open during the off time through the pre-drive circuit so that the generator enters the no-load standby mode.

[0037] In the above technical solution, the combination of the pre-drive circuit and the DC contactor realizes the generator’s true physical-level “no-load standby”, completely eliminating the useless electromagnetic resistance generated by the generator during non-charging periods.

[0038] In some specific embodiments, calculating the real-time rotational speed includes the following steps: Step 1: Extract the raw voltage signal from the battery terminal using a non-invasive acquisition circuit, suppress the common-mode DC component in the raw voltage signal using a differential conditioning circuit, and extract the AC ripple signal from it. Step 2: The AC ripple signal is biased, conditioned, and converted from analog to digital to obtain digital ripple data. Step 3: Perform a fast Fourier transform on the digital ripple data to extract the ripple fundamental frequency, and combine the number of magnetic pole pairs and the rectified pulsation number to calculate the real-time rotational speed.

[0039] In the above technical solution, a non-intrusive differential signal processing link is used to accurately extract weak AC features in the context of high common-mode noise, ensuring the purity and real-time performance of the underlying input data of the identification algorithm.

[0040] In some specific embodiments, the preset standard slope library includes: a standard slope of approximately 0.012A / rpm for the 80A level, a standard slope of approximately 0.019A / rpm for the 130A level, and a standard slope of approximately 0.025A / rpm for the 150A level.

[0041] According to a second aspect of the invention, a computer-readable storage medium is provided on which one or more computer programs are stored, which, when executed by a computer processor, implement the method described above.

[0042] According to a third aspect of the present invention, an adaptive control system for a vehicle generator is provided, comprising: The data acquisition module is configured to acquire the generator's raw operating data within a preset driving cycle, extract the generator's output current based on the raw operating data, and calculate the generator's real-time speed. The feature generation module is configured to perform linear regression analysis based on real-time rotational speed and output current, extract the regression slope, match the regression slope with a preset standard slope library to determine candidate power levels, and generate a trigger signal when the match is successful. The identification module is configured to construct a multi-dimensional feature vector consisting of maximum current characteristics, average power characteristics, and voltage change rate characteristics in response to a trigger signal. Based on the multi-dimensional feature vector, the identification model is used to perform logical verification on the candidate power level to determine the final generator power level attribute information. The control module is configured to construct an optimization objective function that couples energy conversion efficiency, generator operating efficiency, and battery life loss characteristics based on generator power level attribute information and real-time battery status. It then uses a dynamic duty cycle adjustment mechanism to adaptively solve for the optimal on-time and off-time that match the generator power level.

[0043] This invention proposes an adaptive control method and system for vehicle generators, which has the following technical advantages: 1. This invention constructs a multi-dimensional feature vector by calculating the correlation coefficient between output current and real-time speed. This vector can accurately isolate artifact interference caused by random load connections such as air conditioning and headlights from complex underlying electrical data. Compared to traditional single-threshold detection or invasive solutions that require cracking the CAN protocol, this invention achieves adaptive blind identification of heterogeneous generator models (such as 80A, 130A, and 150A) without compromising the original vehicle wiring harness warranty. This solves the problem of control logic incompatibility caused by different suppliers' generators in the same vehicle model.

[0044] 2. This invention abandons the traditional fixed charge-stop ratio mode and introduces an adaptive adjustment mechanism based on a target control function. Under the dual constraints of satisfying the battery's state of charge (SOC) fluctuation range and the generator's peak efficiency range, the optimal dynamic duty cycle is mapped in real time. This not only ensures that the generator always operates in the high-efficiency range of mechanical energy-to-electrical energy conversion, reducing fuel consumption, but also effectively avoids overcharging and deep discharging of the battery through closed-loop feedback on the SOC change rate, significantly extending the cycle life of the parking lithium battery system.

[0045] 3. To address the severe voltage fluctuations generated during intermittent charging switching of high-power generators, this invention specifically designs a symmetrical sleep cycle and a soft-shutdown mechanism that limits the slope of current changes. Through the coordinated operation of the underlying hardware pre-drive circuit and the physical characteristics of the generator's original factory regulator, a smooth power drop is achieved during the transition from the on to the off state by utilizing the generator's own magnetic field attenuation characteristics. Compared to the hysteretic response of traditional software protection, this solution effectively protects the system's power devices and sensitive electronic components of the original vehicle from high-voltage surges.

[0046] 4. This invention utilizes a non-intrusive acquisition circuit and Fast Fourier Transform (FFT) technology to directly calculate the real-time generator speed from the minute ripples at the battery terminal, achieving "zero intrusion" at the sensing layer. Simultaneously, combined with automatic policy domain matching of the identification results by a cloud server, the entire system possesses powerful online evolution and remote optimization capabilities, greatly simplifying after-sales installation and maintenance processes and providing reliable technical support for large-scale commercial applications across vehicle models. Attached Figure Description

[0047] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments and, together with the description, serve to explain the principles of the invention. Many anticipated advantages of the embodiments and other embodiments of the invention will be readily recognized as they become better understood through reference to the following detailed description. Other features, objects, and advantages of this application will become more apparent from reading the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart of an embodiment of the adaptive control method for a vehicle generator according to this application; Figure 2 This is a circuit diagram of a generator voltage ripple acquisition circuit according to a specific embodiment of this application; Figure 3 This is a schematic diagram of the hardware architecture of a non-intrusive generator shutdown control system according to a specific embodiment of this application; Figure 4 This is a framework diagram of an adaptive control system for a vehicle generator, according to a specific embodiment of this application. Figure 5 This is the overall topology of an adaptive control system for a vehicle generator, which is a specific embodiment of this application. Detailed Implementation

[0048] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0049] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0050] Figure 1 A flowchart of an adaptive control method for a vehicle generator according to an embodiment of this application is shown. Figure 1 As shown, the method includes the following steps: S1: Obtain the original operating data of the generator within the preset driving cycle, extract the generator's output current based on the original operating data, and calculate the generator's real-time speed.

[0051] In some specific embodiments, data is collected continuously from the vehicle end. The original operating data of each driving cycle (of which) The raw data includes output current, timestamp, generator voltage, battery SOC value, and information reflecting the vehicle's electrical load status (such as whether high-power electrical appliances such as headlights and air conditioning are turned on).

[0052] In some specific embodiments, frequency characteristics reflecting the real-time operating status of the generator are extracted by sampling the voltage ripple signal at the battery terminal, and real-time speed information is obtained based on the frequency characteristics and the inherent structural parameters of the generator.

[0053] Specifically, this process uses a non-invasive method to extract the AC ripple component, which includes three-phase full-wave rectification, from the positive and negative output terminals of the battery. (Reference) Figure 2 , Figure 2A generator voltage ripple acquisition circuit diagram according to an embodiment of this application is shown. As shown, the raw voltage signal enters the sampling link from connectors J2 (P+) and J3 (P-) at the generator input. This sampling link, through a high-impedance differential input architecture, utilizes J2 and J3 in conjunction with a conditioning circuit composed of a high-precision operational amplifier U1A (model OPA2330AIDR), to achieve non-invasive detection of the generator's operating condition without disassembling it. In the specific extraction process, the raw signal first passes through a filter network (C1, C2, C7) composed of DC blocking capacitors to remove the high-voltage DC component and extract the small AC component. It is then differentially amplified using precision sampling resistors R2 and R4 (1kΩ) and feedback resistors R1 and R5 (100kΩ), thereby adjusting the weak millivolt-level signal to a voltage range recognizable by the analog-to-digital converter. To ensure acquisition accuracy and high dynamic range, the circuit utilizes a voltage reference chip U3 (model TPR3312) to provide a stable reference bias voltage AD_VREF. This design enables the system to simultaneously process low-frequency ripple signals during generator idling and high-frequency ripple signals during high-speed driving. The amplified and conditioned ripple signal AIN_P enters the analog front-end chip U2 (model TPAFE51716H) with high-resolution ADC functionality. Chip U2 communicates with the BMS main control unit (not shown) via the J1 interface. The J1 interface, serving as a communication bridge between the acquisition module and the BMS main control unit, includes an SPI bus composed of pins such as SCLK, MOSI, MISO, and CS, rapidly transmitting the converted digital ripple data to the BMS main control unit for FFT calculations, thereby achieving tracking of the generator ripple frequency characteristics and accurate speed calculation under different operating conditions. Furthermore, this solution employs an integrated design, with most components of the generator voltage ripple acquisition circuit using miniature packages such as 0603 or SOP. This miniaturization demonstrates that the module can be easily integrated into a corner of the BMS mainboard, occupying only approximately 22×18mm of PCB space, thus achieving the integration of intelligent identification functions without altering the original external envelope size of the parking battery.

[0054] In some specific embodiments, after receiving the digital ripple data, the BMS main control unit performs frequency extraction and rotational speed calculation steps. First, in the frequency extraction stage, the BMS main control unit uses a Fast Fourier Transform (FFT) spectral analysis algorithm to process the digital sequence. To ensure a balance between recognition accuracy and real-time performance, a sampling rate is set. for (Preferred) The number of points for FFT calculation was selected as 1024 to achieve the desired frequency resolution. Maintain at the agreement The Hanning window is added to the calculation to reduce spectral leakage, thereby accurately locking the frequency component with the largest amplitude as the ripple fundamental frequency. .

[0055] Specifically, in the signal preprocessing stage, the BMS main control unit performs windowing processing (preferably Hanning window) on the raw sampling sequence acquired by the ADC to suppress spectral leakage and improve the signal-to-noise ratio of frequency components; then executes... Point Fast Fourier Transform, where the sampling frequency , Preferably, the frequency domain is 1024 or 2048 to obtain the frequency domain signal. To improve search efficiency and eliminate invalid interference, a preset physically feasible frequency band is used. (like Peak search is performed within this index range. The main control unit locks the spectral line with the largest amplitude within this index range. And record the corresponding candidate frequencies. With candidate amplitude To prevent higher harmonics from being misidentified as the fundamental frequency, a harmonic consistency check is performed: candidate frequencies are detected. Frequency division position (e.g.) Does the surrounding area contain a significant spectral peak? If the amplitude at that location... and If the ratio exceeds a preset threshold (e.g., 0.7), the current candidate is determined to be a second harmonic, and fundamental frequency backtracking is performed. Update to new candidate frequencies. Furthermore, dual filtering is performed using adaptive amplitude thresholding and time continuity tracking. Specifically, only when the candidate amplitude... A signal is considered valid when it exceeds a preset multiple (e.g., 3 times) of the current frequency band's average amplitude. Simultaneously, the frequency change rate of the current frame is compared to that of historical valid frames. If the change rate exceeds a preset proportion (e.g., 25%), a hold mechanism is activated. Output updates are only performed when the frequency remains stable near the new frequency for several consecutive frames (e.g., 3 frames) to eliminate signal jumps caused by load switching. This mechanism ensures accurate identification of weak ripple fundamental frequency characteristics even under extreme conditions such as heavy load acceleration and high battery current discharge, which lead to increased noise floor.

[0056] Secondly, in the speed calculation and generator type identification stages, the mapping relationship between the generator's physical topology and electrical characteristics is first established. Specifically, utilizing the physical characteristics of an automotive silicon rectifier AC generator after three-phase full-wave rectification, its output ripple frequency is correlated with the real-time rotor speed of the generator. Satisfying the formula: ;in, This represents the number of pole pairs on the generator rotor. This refers to the rectified ripple count. To achieve parameter locking when motor specifications are unknown, and engine speed cannot be obtained, the BMS main control unit executes a physical topology identification program based on the transient ripple frequency characteristics of engine startup to obtain the generator's physical parameters and then calculate the real-time motor rotor speed. Specifically, this includes: Step 1, Start Acquisition and Spectrum Analysis: At the moment of engine startup, trigger ripple signal acquisition at a preset sampling frequency. (like Continuous data acquisition Data points (e.g.) (number of points), and according to the preset number of points (e.g., ... (Points) are used to perform frame-by-frame FFT operations to obtain the ripple frequency sequence corresponding to each time frame. (in ).

[0057] Step 2, Dynamic Slope Extraction: Extract the first segment of the frequency sequence in the speed ramp-up phase of the ripple frequency sequence (the first segment of the frequency sequence). (Frame), and linear fitting was performed using the least squares method to calculate the dynamic rising slope of the ripple frequency over time. In the formula, The slope of the frequency change (Hz / s) represents the "acceleration" characteristic of the generator at the moment of startup. For the first The timestamp (in seconds) corresponding to the frame; The arithmetic mean of the timestamps of the selected frames; For the first The ripple fundamental frequency (Hz) of the frame is extracted by FFT. This is the arithmetic mean of the selected frame fundamental frequency sequence.

[0058] Furthermore, the identification logic based on the startup transient frequency slope is supported by a deep physical model. During the engine startup transient (typically lasting...), (seconds), engine speed An approximately linear increase occurs, and its slope (i.e., acceleration) is defined as follows: Accordingly, the generator speed satisfies (in (This refers to the transmission ratio between the generator and the engine). Based on the aforementioned relationship between rotational speed and frequency, the variation of the ripple fundamental frequency with time can be expressed as: Regarding time By taking the derivative, we can obtain the dynamic rising slope of the ripple frequency as a function of time. Because in the same vehicle, the transmission ratio and startup acceleration In a single startup loop, the slope can be considered a constant, therefore... Product with the physical parameters of the generator Proportional. See Table 1 for details; the product of different physical topology categories... It exhibits a significant gradient difference (e.g., 18:24:36:48). This embodiment utilizes this physical characteristic by measuring... This allows for operation without needing to know the transmission ratio. Under the premise of precise engine speed, blind identification of generator physical topology category is achieved.

[0059] Table 1 Combinations of physical parameters for different generator types

[0060] Step 3, End-point feature calculation: Extract the latter segment of the frequency sequence at the start-up end of the ripple frequency sequence (such as the latter...). (frames), calculate their arithmetic mean as the start-up end frequency. ;in, In the formula, This represents the total number of frames collected during the system startup phase. For cumulative indexing, For the first The ripple fundamental frequency corresponding to the frame, The average frame count is the number of consecutive data frames used to calculate the end-average value; it is calculated by selecting the end of the sequence. Frames (e.g.) The frequency is averaged to eliminate frequency fluctuations caused by minor engine speed variations at the end of startup, thereby obtaining a stable frequency determination benchmark. Based on this, this application further refines the frequency values ​​extracted by the above algorithm. The physical meaning is further defined. According to the physical model of the rectification characteristics of an AC generator, the starting end corresponds to the generator entering idle condition, and its theoretical frequency value satisfies... Although the transmission ratio and idling speed While there are fluctuations across different vehicle models, the distribution range of their product remains relatively fixed. This is because the idling speed of commercial vehicles is typically maintained at... RPM range, and belt drive ratio Usually in Within the physical constraints, the frequency distributions of generators of different physical topology categories at the start-up and end times have significant envelope boundaries. By combining the dynamic slope with the start-up end frequency, a two-dimensional feature mapping matrix can be constructed, thus providing a reliable auxiliary verification benchmark for the identification logic in the subsequent step four.

[0061] Step 4, Topology determination: Determine the dynamic slope. It is compared with multiple preset slope threshold ranges, and supplemented by the start-up end frequency. Interval correction is performed to determine the specific physical topology category of the generator, including 6-pole single-phase, 8-pole single-phase, 6-pole double-phase, or 8-pole double-phase. Based on this, the number of generator sets is determined. With rectified pulsation number The slope threshold interval is a priori feature interval determined by collecting operational data on the startup process of several calibrated vehicle models in advance and conducting statistical distribution analysis on the slope of the generator under different physical topology categories. The slope threshold interval is pre-stored in the form of a lookup table in the cloud server or on-board local storage.

[0062] Furthermore, based on the significant slope ratio relationships between different physical topology categories, due to the dynamic rising slope being multiplied by the physical parameters... Proportional, theoretically the slope ratio of the four types of generators should satisfy... In the specific identification process, a preliminary division is made using preset prior feature intervals. Referring to Table 2, considering the differences in transmission ratios and starting accelerations among different vehicles, which may lead to local overlap of the above intervals, this application preferably uses a ratio verification method for auxiliary determination. Specifically, the ratio of the dynamic rise slope to the terminal frequency is calculated. : Because of this ratio The ratio of engine starting acceleration to idle speed is the sole determining factor, independent of the generator's transmission ratio. Therefore, it can be used to verify the consistency of currently collected data and eliminate interference from abnormal operating conditions. Ultimately, the BMS main control unit constructs a two-dimensional feature space using the dynamic ramp rate and terminal frequency. Through a pre-defined rectangular decision region or linear discriminant model, it accurately identifies the specific type of generator. This multi-dimensional feature comparison logic ensures that the system maintains extremely high robustness even under blind identification conditions completely detached from speed signals.

[0063] Table 2 shows the distribution range of starting slope obtained from statistical data of typical vehicle models.

[0064] Specifically, the BMS main control unit determines the slope based on a preset first slope threshold. Second slope threshold and the start-up end frequency threshold The following segmented decision logic is executed to lock the physical topology category of the generator: when a dynamic rising slope is detected. satisfy At that time, the generator was determined to belong to the 6-pole single-three category; when the dynamic upward slope was identified... satisfy And the start-up end frequency satisfy At that time, the generator was determined to belong to the 8-pole single-three category; when the dynamic upward slope was identified... satisfy And the start-up end frequency satisfy At that time, the generator was determined to belong to the 6-pole double-three category; when the dynamic upward slope was identified... satisfy And the start-up end frequency satisfy At that time, the generator was determined to belong to the 8-pole double-three category. Through the aforementioned multi-dimensional value cross-validation logic, the slope overlap interference caused by differences in transmission ratios across different vehicle models is effectively eliminated, ensuring that even in blind identification conditions with only instantaneous frequency characteristics, the number of pole pairs of the generator can still be accurately identified. With rectified pulsation number .

[0065] Step 5, Real-time speed calculation: using the formula The real-time speed of the generator is obtained by calculation. .

[0066] S2: Based on real-time speed and output current, perform linear regression analysis, extract the regression slope, match the regression slope with a preset standard slope library to determine the candidate power level, and generate a trigger signal when the match is successful; In some specific embodiments, the process for extracting and identifying the regression slope is as follows: The BMS continuously monitors the real-time speed of the generator (obtained by ripple fundamental frequency inversion). When the speed is within a preset range (e.g., to When increasing linearly within a given range, the speed is incremented in steps (e.g., every interval). Using [a specific condition] as the trigger, the actual output voltage and charging current of the generator are synchronously acquired to construct a feature sequence containing 10 sample points. Subsequently, univariate linear regression analysis is performed on the feature sequence using the least squares method to calculate the regression slope. (Unit is) The main control unit will measure the actual data. The generator's rated power level is locked by Euclidean distance matching with a pre-stored standard slope library. Experimental calibration data show that the regression slopes of generators with different power levels have significant physical differentiation: the standard regression slope of the 80A level is approximately... The standard regression slope for the 130A level is approximately 0.019A / rpm; the standard regression slope for the 150A level is approximately... By calculating the Euclidean distance between the measured value and the aforementioned standard value, the candidate power specification of the current generator can be preliminarily determined. This candidate power specification will be used as a key prior feature or trigger signal input into the subsequent identification model. By combining multi-dimensional feature vectors for deep logic verification in the identification model, accurate and robust identification of the generator's physical properties can be achieved, thereby providing accurate decision support for the dynamic adjustment of subsequent charging strategies.

[0067] Specifically, the measured regression slope The nominal slope of each power level in the pre-stored standard slope library (in Euclidean distance matching is performed, and the matching logic is as follows: First, the measured slopes are calculated respectively. With each nominal slope Euclidean distance between : Subsequently, by iterating and comparing, the set of Euclidean distances was selected. minimum value When this minimum value meets the preset deviation tolerance condition (i.e. , When the threshold value is less than the preset allowable error threshold, a successful match is determined and a trigger signal is generated. Simultaneously, the [missing information] is... Corresponding power level The candidate rated power specification for the current generator set is locked. By introducing the aforementioned Euclidean distance matching mechanism and deviation tolerance verification, robust identification of the generator set's physical specifications can be achieved even with sampling fluctuations, effectively avoiding identification jumps caused by minor measurement errors.

[0068] S3: In response to the acquisition of the trigger signal, a multi-dimensional feature vector consisting of maximum current characteristics, average power characteristics and voltage change rate characteristics is constructed. Based on the multi-dimensional feature vector, the identification model is used to perform logical verification on the candidate power level to determine the final generator power level attribute information.

[0069] In some specific embodiments, in response to receiving a trigger signal, the original data is filtered to ensure the accuracy of subsequent model identification. This effectively removes abnormal data spikes caused by transient physical conditions such as rapid acceleration and deceleration of the vehicle, resulting in cleaned and valid data. Multidimensional feature vectors are then extracted based on the cleaned and valid data. In the formula, This represents the average power output of the generator. The slope of the voltage fluctuation; The goal is to capture the peak output current of the generator within the observation period. A fuzzy logic-based algorithm is employed to accurately identify the generator's attribute information (i.e., its specific model). First, the multi-dimensional feature vectors extracted in the preceding steps are fuzzified using a preset membership function. Then, an internally constructed fuzzy rule base is invoked for logical reasoning. After comprehensive reasoning using the fuzzy rule matrix, the centroid method is finally used for defuzzification, thereby deriving and outputting a precise generator model, providing a unique basis for subsequently developing a safe and efficient charging control strategy.

[0070] In some specific embodiments, fuzzy logic-based algorithms are used to accurately identify generator attribute information (i.e., specific model specifications). The identification model includes fuzzification processing of the input feature vector, a rule-based fuzzy inference engine, and defuzzification of the output variables.

[0071] Specifically, the input feature vector of the identification model is defined as follows: For each feature variable, corresponding fuzzy linguistic values ​​were defined, and a membership function combining trapezoidal and triangular bounds was used to map the measured feature values ​​to the fuzzy set space, where: Maximum current characteristics This is used to characterize the short-time overload capacity and peak output limit of the generator. Based on the physical characteristics of generators of different power levels, this embodiment defines three fuzzy subsets: Low (L), Medium (M), and High (H), corresponding to generators of 80A, 130A, and 150A levels, respectively. Their corresponding membership function expressions are shown below: Low power rating (80A) membership function

[0072] Membership function for medium power rating (130A)

[0073] High power rating (150A) membership function

[0074] In the formula, This represents the maximum output current characteristic value of the generator after real-time acquisition and processing within a preset sampling period. By setting the aforementioned 20A fuzzy transition band, the algorithm can effectively absorb current fluctuations caused by sensor sampling errors and line impedance, thus enhancing the model's fault tolerance.

[0075] Average power characteristics This is used to characterize the continuous load-carrying capacity and energy output density of a generator. In a 24V power supply system, based on the rated output power of generators of different power ratings and typical line losses, the following fuzzy discrimination criteria are established: Low power level membership function

[0076] Membership function of medium power level

[0077] High power level membership function

[0078] The membership function design for average power characteristics fully incorporates the theoretical steady-state output characteristics of generators with different power levels under a 24V power system environment. Considering that the rated power of 80A, 130A, and 150A generators is approximately 1920W, 3120W, and 3600W, respectively, differentiated physical boundary constraints are set: 2000W is used as the threshold for low power level determination, 2600W to 3000W is used as the saturation mapping range for medium power level determination, and 3400W is used as the starting threshold for entering high power level determination. To ensure the robustness of the identification model in complex electrical environments, a membership slope transition range with a width of 400W or 600W is set at each power determination boundary. Through this smooth evolution of fuzziness, measurement uncertainty interference caused by differences in battery internal resistance, ambient temperature fluctuations, and voltage drop in charging lines is effectively absorbed, thereby achieving accurate extraction and locking of generator power attribute information.

[0079] Voltage change rate characteristics The membership function distinguishes between the generator's own excitation regulation process and voltage fluctuations caused by sudden increases in external random loads in real time through dynamic response characteristics. Under parking charging conditions, the absolute value of the voltage change rate caused by the generator regulator's closed-loop control is usually small and regular, generally maintaining a range of... Within the range; however, the instantaneous connection of a high-power external load (such as a parking air conditioner, air pump, etc.) will produce irregular and significant voltage jumps, with a rate of change typically exceeding [a certain threshold]. Therefore, this application defines a membership function with small rate of change. Membership function with large rate of change Their mathematical expressions are shown in the following formulas:

[0080] .

[0081] Based on the membership function design described above, the identification model can evaluate the reliability of input features in real time. When the system detects high... Features (i.e.) When the value approaches 1), it is determined that there is serious external interference in the current operating condition, and the calculation weight of the data in that time period is reduced during the fuzzy rule reasoning process; conversely, when... In When the generator is in the saturation range, it is determined that the generator is in a stable adjustment phase, which gives higher confidence to the current and power characteristics, thereby ensuring that the exact generator physical topology and power level can still be locked in the environment of complex load fluctuations.

[0082] Specifically, the output variable of the identification model is defined as the generator power level. The corresponding output fuzzy set is defined as Considering the high requirements for real-time performance and computational efficiency of the in-vehicle embedded main control unit, the output membership function of this application preferably adopts a singleton function form. This design greatly simplifies the integral calculation in the subsequent defuzzification process by mapping the fuzzy inference results to discrete power values. The mathematical expressions of the output membership function are shown in the following formulas:

[0083]

[0084]

[0085] In the formula, The output variable represents the measured / calculated value. , , This represents the degree of membership.

[0086] After the fuzzy inference engine completes rule matching and obtains the output weights corresponding to each rule, the output variables are finally defuzzified using either the centroid method or the weighted average method. By calculating the sum of the products of each individual output value and its corresponding membership weight, the determined generator power rating is derived. This output result, as a unique attribute parameter, not only locks in the physical topology specifications of the generator but also provides a precise control basis for subsequently developing safe and efficient charging current allocation and limit protection strategies.

[0087] Specifically, the identification model employs a Mamdani-type fuzzy inference mechanism to perform logical verification based on multi-dimensional feature vectors. Its built-in fuzzy rule base utilizes maximum current... Average power and voltage change rate Multidimensional feature coupling determination is used to ensure the authenticity of candidate power levels, where the maximum current is considered. Average power and voltage change rate Each is mapped to a corresponding fuzzy set through a preset membership function, if the maximum current With average power After membership mapping, all values ​​are in the low range of the corresponding power level and the voltage change rate is... When the generator set is within a preset steady-state range, determine its power level. The maximum current is 80A. With average power After membership mapping, all values ​​fall within the median range of their corresponding power levels, and the voltage change rate... When the generator set is within a preset steady-state range, determine its power level. The maximum current is 130A. With average power After membership mapping, all values ​​are in the high range of the corresponding power level and the voltage change rate is... When the generator set is within a preset steady-state range, determine its power level. The current current is 150A, and this application specifically introduces a non-steady-state interference interception criterion, namely, when the voltage change rate is detected... When the current data is in a pre-defined non-steady-state range, it is determined that the current data is frequently affected by non-linear interference from external random load access or transient physical conditions of the vehicle. The system executes non-steady-state suppression logic to stop updating the current identification results or performs weight reduction processing by adjusting the contribution of the current frame in the time series weighting operation. Through the synergistic effect of the above consistency verification and dynamic interception mechanism, robust identification of the generator's electrical physical properties is finally achieved.

[0088] Furthermore, to provide a quantifiable basis for the execution of the aforementioned fuzzy sets, this application, based on the typical output characteristic curves of 80A, 130A, and 150A generators, precisely defines the range of the membership function as follows: In the current dimension, low-value fuzzy sets... The complete membership interval (i.e., the interval with a membership degree of 1) is: Median fuzzy set The complete membership interval is High-value fuzzy sets The complete membership interval is In the power dimension, low-value fuzzy sets The complete membership interval is Median fuzzy set The complete membership interval is High-value fuzzy sets The complete membership interval is In terms of voltage stability, the steady-state interval The complete membership interval is ; and non-steady-state large range The complete membership interval is It should be noted that the aforementioned fuzzy sets achieve soft switching through linear transition segments. This piecewise linear function design allows the identification model to achieve a stable transition of recognition results through smooth changes in membership degrees when operating conditions fluctuate drastically or sampling errors exist, thus exhibiting higher engineering fault tolerance. In practical applications, those skilled in the art can adaptively adjust the boundary thresholds and slopes of the membership functions for each interval based on the line voltage drop compensation values ​​and load dynamic response characteristics of different vehicle electrical systems to ensure optimal adaptability of the identification model under specific hardware environments.

[0089] Specifically, after completing logical reasoning, the identification model uses the centroid method to perform defuzzification processing in order to calculate the output sharpness value. Its calculation formula is In the formula, Determine the numerical value for the final generator model, such as 130A; The fuzzy rule index participating in the summation calculation has a value range of 1. These correspond to three generator nominal power ratings: 80A, 130A, and 150A, respectively. These are the single-point nominal values ​​corresponding to the generator's rated current, taken as 80, 130, and 150 respectively. For the first The membership weights output by each fuzzy rule, i.e., the activation strengths obtained from fuzzy rule inference, take values ​​ranging from... In between, obtaining the output clarity value Then, based on the principle of proximity, the calculation result of this continuous type is mapped to the nearest nominal power level value, thereby determining the final generator power level, such as when the calculated output clear value... When the value is within the preset range of 125 to 135, the generator power level is determined to be 130A. This determined generator power level attribute is used as the physical reference for executing subsequent dynamic charging strategies, current distribution, and limit protection.

[0090] S4: Based on the generator power level attribute information and the real-time status of the battery, construct an optimization objective function that couples the energy conversion efficiency, generator operating efficiency and battery life loss characteristics, and use a dynamic duty cycle adjustment mechanism to adaptively solve for the optimal on-time and off-time that match the generator power level.

[0091] In some specific embodiments, the optimal intermittent cycle is calculated by constructing an optimization objective function. The goal is to maximize DC-DC conversion efficiency and ensure battery life while meeting constraints such as battery SOC fluctuation range and generator peak efficiency range. In the specific cycle calculation logic, a dynamic duty cycle adjustment mechanism is introduced, defining the duty cycle as the ratio of conduction time to the cycle. Specifically, a multi-dimensional control strategy initial value logic is constructed based on the identified generator model and real-time SOC change rate, providing the basic constraint boundaries for the optimization objective function. In the execution rules of the control strategy, when an 80A low-power generator is identified, a "shallow charge and shallow discharge" strategy is adopted, setting the search center of the duty cycle to around 0.5 (e.g., corresponding to 30s for both on and off times). If the battery SOC is detected to be dropping too quickly in real time, the algorithm will automatically adjust the duty cycle to shift towards the upper limit to enter boost mode. When a 130A medium-power generator is identified, a "balancing strategy" is adopted, increasing the proportion of on time in the cycle. Simultaneously, the algorithm will combine future road condition predictions based on navigation data to further dynamically adjust the optimization step size of the duty cycle. When a 150A high-power generator is identified, a "deep sleep strategy" is adopted, utilizing its large power redundancy to execute a symmetrical duty cycle distribution to maximize energy saving, and employing a soft turn-off mechanism during off-peak periods to limit the current slope. Furthermore, the generator operating efficiency model comprehensively considers both mechanical energy conversion efficiency and power output stability. When dynamically solving for the optimal duty cycle, the algorithm will include the period when the generator's real-time speed falls within the peak efficiency range as a weighting factor in the objective function. This will prioritize the allocation or alignment of conduction time when the generator speed is in the high-efficiency range, ensuring that the vehicle's fuel economy is maximized while maintaining battery power.

[0092] Specifically, the soft shutdown mechanism is achieved by controlling the duty cycle of the DC-DC converter to decrease linearly at a preset slope. When the identification result is 150A and the shutdown phase begins, the system does not immediately set the duty cycle to zero, but instead controls it to... to During the buffer period, the current is decreased to zero frame by frame according to the step value. Through this controlled current decay path, the rate of change of the charging circuit current can be limited within a safe threshold, thereby eliminating the impact of induced voltage spikes on the circuit.

[0093] Specifically, optimize the objective function Based on the definition of continuous charging cycles, it determines the optimal duty cycle by maximizing overall efficiency gains. objective function The mathematical expression is defined as In the formula, , , The weighting coefficients are satisfied. The constraints (such as taking values ​​respectively) , , ); A DC-DC conversion efficiency model; For generator operating efficiency model; This represents the battery life loss factor. Substituting it into the specific sub-model expression, its expanded form is: In the formula, , For a preset fixed period (such as) ), , , These are the weighting coefficients. Bus voltage To set the charging current, This is the generator's optimal efficiency power point. This refers to the generator's rated power. For the battery's rated capacity, The maximum allowable SOC fluctuation, This represents the charging rate penalty coefficient. Furthermore, the solution to the above multi-objective optimization function is subject to strict physical constraints, defined as follows: To ensure that the state of charge is within the safe operating range; To limit the generator output power to the high-efficiency bandwidth range; and to meet the minimum switching time constraint. and and the range of duty cycle values In the formula, For the storage battery Real-time state of charge at any given moment (or state of charge at the start of a charging cycle) ); This represents the lower limit of the generator's output power within its efficient operating range. This refers to the upper limit of the generator's output power within its high-efficiency operating range. The preset minimum charging on-time (to avoid frequent operation of power devices); This is the preset minimum charging off time. Since the objective function is about... For a quadratic function with a squared penalty term, the system can obtain the optimal duty cycle using the golden section search or analytical differentiation. And based on this, map back to the specific conduction time. With shutdown time To implement a dynamic charging strategy.

[0094] For the dynamic adjustment and execution of power control parameters, the cloud platform combines the identified attribute information and the real-time battery status to calculate the optimal intermittent cycle (dynamic duty cycle) and encapsulates it into control commands to be sent to the vehicle system. Specifically, the system employs differentiated charging / stop strategy mapping logic at the execution layer, and its underlying code logic implementation is as follows: def calculate_duty_cycle(self, current_soc, soc_delta): """ Duty cycle dynamic calculation logic current_soc: Current state of charge soc_delta: Rate of change of SOC (% / min) """ if self.model_type == 80: # 80A Generator Control Strategy: Shallow Charge and Discharge if soc_delta<-0.5: # Discharge too fast, enter boost mode Ton, Toff = 120, 60 else: Ton, Toff = 120, 120 elif self.model_type == 130: # 130A Generator Control Strategy: Efficiency First if current_soc < 50: # Low power state, hard threshold switching Ton, Toff = 240, 60 else: Ton, Toff = 240, 90 elif self.model_type == 150: # 150A Generator Control Strategy: Deep Sleep and Symmetrical Cycle Ton, Toff = 180, 180 else: Ton, Toff = 180, 120# Default security policy return Ton, Toff # (Continuing logic from the main execution function of run_cloud_strategy) # 2. Obtain real-time SOC status current_soc = vehicle_data['real_time']['soc'] soc_delta = vehicle_data['real_time']['soc_delta'] # 3. Calculate the on and off cycle parameters Ton, Toff = self.calculate_duty_cycle(current_soc, soc_delta) # 4. Encapsulate control parameter instruction packages and send them out. command = { 'timestamp': vehicle_data['real_time']['timestamp'], 'generator_model': self.model_type, 'DC_DC_Ton': Ton, 'DC_DC_Toff': Toff, # If it is a 150A generator, a soft shutdown action flag should be added synchronously to the instruction. 'soft_shutdown_flag': True if self.model_type == 150 else False, 'strategy_version': 'v2.5' } return command Specifically, the above code logic works in conjunction with the optimization objective function: when the system is identified as an 80A generator, a "shallow charge and shallow discharge" strategy is executed, dynamically reducing the on-time and off-time by monitoring the SOC change rate soc_delta (e.g., switching from a 120s / 120s boost mode to a 120s / 60s boost mode); when identified as a 130A generator, an "efficiency priority" strategy is executed, and the on-time percentage is extended by hard threshold switching in a low-power state (e.g., current_soc < 50); when identified as a 150A generator, a "deep sleep" symmetrical cycle (e.g., 180s / 180s) is executed, and a "soft shutdown" action flag soft_shutdown_flag is simultaneously added to the control command packet to achieve energy saving and suppress current ripple by utilizing high-power redundancy. This logic ensures that complex mathematical optimization results can be transformed into discretized time parameters that can be directly executed by the vehicle's underlying controller.

[0095] Furthermore, considering the potential latency or blind spots in the vehicle communication network, the control module of this application incorporates an offline backup and fault prevention mechanism at the underlying architecture. Specifically, during the system cold start initialization phase or before receiving dynamic duty cycle parameters from the cloud, the battery management system defaults to a strongly compatible conservative charging strategy. Under this conservative strategy, without relying on any identification results, the intermittent cycle is directly and forcibly set to 3 minutes on and 1 minute off, and the charging current limit threshold of the DC-DC converter is hard-clamped at 80A. This ensures basic energy replenishment while absolutely preventing the low-power generator (such as the 80A model) from falling into a dangerous state of long-term full-load overload operation due to cloud disconnection, significantly improving the system's robustness.

[0096] This invention provides an end-to-end closed-loop control execution link. Specifically, after the cloud platform performs dynamic calculation of the duty cycle, it calculates the conduction time... With shutdown time Control parameters are encapsulated into a standard CAN signal format (or a command packet in a pre-defined communication protocol format such as automotive Ethernet) and transmitted to the vehicle's T-Box (in-vehicle remote communication terminal) via a 4G / 5G network. Upon receiving the signal, the T-Box transmits it through the vehicle's CAN bus to the vehicle's BMS. The BMS, acting as the final execution agent, parses the specific cycle time and directly sends high / low level control signals to the EN (enable) pin of the in-vehicle DC-DC converter via a hard-wired physical connection, thereby achieving precise intermittent generator operation control on the physical circuitry. This architecture, which directly controls the EN pin, not only has extremely low response latency (milliseconds) but also completely avoids intrusive modifications to the complex excitation logic at the generator's underlying level, greatly improving the compatibility and safety of system retrofits.

[0097] Furthermore, the BMS receives the shutdown time notification from the cloud ( Following the instruction, physical isolation of the generator is achieved through a specific hardware execution module. For example... Figure 3 As shown, the hardware execution module is topologically divided into a main loop (very thick line) and a control loop (medium thick line). Specifically, the main loop includes a circuit connected in series to the positive output terminal of the generator (B). + ) and the positive terminal of the battery (B) +The main contact K1 of the DC contactor between the generator and the battery is rated to carry a current greater than or equal to the peak current of the generator (e.g., a 200A DC contactor is selected). The control circuit includes the DC contactor coil and a freewheeling diode D1 connected in reverse parallel with the coil, as well as a pre-drive circuit as the underlying drive actuator. The positive power supply terminal of the coil is connected to the positive terminal of the battery to obtain a constant +24V power, and its negative return terminal (GND) is electrically connected to the low-side control output terminal of the BMS (which is essentially composed of the switching transistor of the pre-drive circuit). At the same time, the generator and the negative terminal of the battery (B) are connected to the negative terminal of the battery. - Maintain common ground connection. During dynamic execution, when entering the conduction cycle ( When the BMS triggers the pre-drive circuit to conduct on the low side, the DC contactor coil is grounded and energized, generating electromagnetic attraction, which in turn attracts the main contact K1, and the generator charges the battery normally through the main circuit; when the shutdown cycle begins ( When the BMS sends a cutoff signal to the pre-drive circuit to cut off the low-side output, the coil loses its grounding circuit and is de-energized. The main contact K1 opens. At this instant, the freewheeling diode D1 provides a discharge circuit for the reverse electromotive force released by the coil, protecting the underlying circuitry of the BMS from high-voltage breakdown. After the main contacts open, the generator is in a completely physical no-load state. Its internal original voltage regulator, upon detecting an increase in the output voltage, will automatically reduce the excitation current, allowing the generator to smoothly enter the no-load standby state. Thus, safe and reliable intermittent shutdown control is achieved without altering the original vehicle's internal generator circuitry.

[0098] In addition, the pre-drive circuit can also adopt a high-side drive topology (such as a P-MOSFET or a small relay drive scheme). Specifically, the pre-drive circuit mainly includes a high-voltage P-channel MOSFET (such as an IRF9540 or equivalent), a small-signal N-channel MOSFET (such as a 2N7002), corresponding gate pull-up / pull-down resistors, and a freewheeling diode. In this topology, the negative terminal of the DC contactor coil is fixed to the system ground (GND). When entering the conduction cycle ( When the BMS outputs a high-level logic signal, it first drives the underlying small-signal N-MOSFET to conduct, and then pulls down the gate potential of the P-MOSFET to make it conduct, which is equivalent to connecting the +24V constant power to the positive terminal of the coil, so that the coil is energized and the main contacts are attracted; when entering the turn-off cycle ( When the P-MOSFET is turned off, the BMS outputs a low level (or a high impedance state), the N-MOSFET is cut off, and the gate of the P-MOSFET is reset to a high potential by the pull-up resistor, thus turning off. The coil loses positive power supply and releases the main contacts. Similarly, a freewheeling diode is connected in reverse parallel across the coil to discharge the reverse electromotive force released by the coil when the P-MOSFET is turned off. This high-side drive architecture is often used in safety and fault-prevention designs where the DC contactor coil needs to be normally de-energized. Those skilled in the art will understand that whether low-side drive or high-side drive is used, it is an optional electronic hardware means to implement the substantive control logic of "achieving generator no-load shutdown by controlling the main circuit of the contactor" in this invention, and both should fall within the protection scope of this invention.

[0099] It is worth noting that, in addition to the active logic control for the preset charging and stopping cycle, the underlying hardware execution module also incorporates a microsecond-level emergency load dump suppression mechanism to cope with sudden physical failures. When a vehicle experiences a generator unexpectedly shutting down or a drive belt slipping and breaking, the input voltage of the DC-DC converter will rapidly collapse. At this time, the BMS, through its real-time high-frequency sampling underlying hardware, will detect the sudden drop in ripple amplitude or extremely high negative voltage change rate caused by the aforementioned fault. This will trigger a hardware interrupt instantly. The BMS will respond within an extremely short time (e.g., less than 50 microseconds). The system forces the bypass upper-level software logic to immediately send a latching signal to the DC-DC converter to shut down the high-voltage output and simultaneously disconnects the pre-drive circuit to release the main contactor at the input. This underlying emergency suppression mechanism can completely cut off the circuit before the residual magnetic field energy inside the generator can generate a high-voltage pulse, thus perfectly avoiding catastrophic breakdown of the parking lithium battery and expensive loads inside the vehicle caused by a kilovolt-level high-voltage surge.

[0100] To further ensure the high reliability and durability of the aforementioned hardware execution module under harsh automotive conditions (such as wading, high temperature, and high vibration), in a preferred engineering deployment embodiment, the hardware execution module also includes the following specific physical protection and electrical wiring design: In terms of electrical safety, a fuse with a preset amperage (e.g., 5A to 10A) is connected in series in the positive power supply circuit of the DC contactor coil to prevent partial short circuits in the coil from causing thermal runaway of the entire vehicle. Furthermore, the freewheeling diode is physically positioned to be mounted close to the pins of the DC contactor coil to minimize electromagnetic interference caused by parasitic inductance when interrupting large currents. Additionally, to ensure compatibility with the underlying BMS hardware of different manufacturers, if the BMS control output signal is open-drain, a pull-up resistor is connected in parallel in the pre-drive circuit to stably match the gate turn-on threshold voltage of the MOSFET.

[0101] In terms of physical wiring and thermal management, the main circuit carrying the high current of the generator uses large-section copper core cables that meet temperature rise requirements (e.g., 25mm² cables with a temperature resistance of 105℃ for 150A continuous current carrying conditions). Its terminals are hydraulically crimped and insulated with heat-shrink tubing. The metal casing of the DC contactor is connected to the negative terminal of the battery for common grounding, and its installation posture within the vehicle body is configured to be upright or horizontal to prevent water accumulation in the cavity.

[0102] In terms of protection level, the DC contactor and pre-drive circuit components are encapsulated in a sealed junction box that meets the preset dustproof and waterproof level (e.g., IP5X or above). The junction box cover is coated with sealant, and all external cable leads are equipped with waterproof tail connectors to physically block the capillary intrusion of moisture.

[0103] In a preferred structural layout and physical integration embodiment, to achieve high-level intelligent charging control without altering the original external envelope dimensions of the parking battery, the hardware execution module of this application adopts a highly integrated space reuse design. Specifically, the DC-DC converter uses a flat package structure (e.g., a height dimension less than or equal to 45mm), and its physical location is configured to be embedded in the handle groove of the parking battery housing or a reserved cavity in the battery cover; the BMS motherboard is arranged vertically along the reserved space on the side of the parking battery. In addition, the generator voltage ripple acquisition circuit and the vehicle communication antenna for cloud communication are highly integrated into the BMS motherboard or the internal interlayer of the battery housing (e.g., utilizing the redundant interlayer space of approximately 40mm in height inside a standard 270mm×190mm×220mm parking battery), avoiding the creation of additional protruding structures on the outer surface of the battery. At the same time, the electrical connectors inside the system preferentially adopt a sink-board surface mount process to maintain the flatness of the overall circuit board outline. After the aforementioned three-dimensional spatial optimization layout, the exterior of the parking battery casing ultimately retains only DC power output terminals (such as 24V terminals) and an optional flexible patch antenna. This spatial configuration achieves seamless integration of the generator adaptive control system without encroaching on any additional physical space in the vehicle compartment or altering the original battery mounting location.

[0104] In terms of thermal management and extreme environment adaptability, considering the highly integrated and confined internal space, the DC-DC converter abandons the easily damaged active fan and adopts a passive cooling architecture that combines shell conduction with natural convection. This architecture, together with the aforementioned hardware selection, ensures that the control system can still operate stably in a harsh wide-temperature range of -20℃ to +65℃. Furthermore, the integrated parking battery system meets an IP54 or higher protection rating, its front-end input electrical interface is widely compatible with alternator voltage fluctuations from 20V to 32V, and its maximum charging current threshold can be flexibly configured via BMS software based on identified power level attribute information (e.g., configured to a maximum continuous current of 130A), thus providing a high degree of physical tolerance for lossless adaptation to different vehicle models. The soft shutdown mechanism specifically controls the gate drive slope of the power switch in the pre-drive circuit, thereby controlling the rate of change of current during the switching from the on to the off state.

[0105] In summary, the generator shutdown intermittent control scheme provided by this invention not only achieves non-intrusive soft shutdown and no-load standby in terms of logic control, but its underlying hardware execution module also fully meets the stringent requirements of automotive-grade IP5X protection standards and continuous current carrying capacity of over 150A (e.g., maintaining contact temperature rise within a safety margin of 50K under full-load continuous operation). This system architecture features a minimalist physical topology and controllable overall cost, demonstrating high feasibility for mass production and significant industrial practical value.

[0106] Continue to refer to Figure 4 As an implementation of the above method, this application provides an embodiment of a framework diagram of an adaptive control system 400 for a vehicle generator. This system embodiment is similar to... Figure 1 Corresponding to the illustrated method embodiment, this system can be specifically applied to various electronic devices. The system 400 includes a data acquisition module 401, a feature generation module 402, an identification module 403, and a control module 404 that are interconnected, wherein: The data acquisition module 401 is configured to acquire the original operating data of the generator within a preset driving cycle, extract the generator's output current based on the original operating data, and calculate the generator's real-time speed. The feature generation module 402 is configured to perform linear regression analysis based on real-time rotational speed and output current, extract the regression slope, match the regression slope with a preset standard slope library to determine the candidate power level, and generate a trigger signal when the match is successful. The identification module 403 is configured to construct a multi-dimensional feature vector consisting of maximum current characteristics, average power characteristics and voltage change rate characteristics in response to a trigger signal, and to perform logical verification on candidate power levels based on the multi-dimensional feature vector using an identification model to determine the final generator power level attribute information. The control module 404 is configured to construct an optimization objective function that couples energy conversion efficiency, generator operating efficiency and battery life loss characteristics based on generator power level attribute information and real-time battery status, and adaptively solve for the optimal on-time and off-time that match the generator power level using a dynamic duty cycle adjustment mechanism.

[0107] Continue to refer to Figure 5 , Figure 5 This paper illustrates the overall topology of an adaptive control system for a vehicle generator according to a specific embodiment of this application. This system achieves deep integration between the adaptive control system for an external generator and the parking battery casing. Without altering the battery's external dimensions, the casing integrates a fuse, a DC-DC charger with a bidirectional controllable discontinuous mode, a 24V 260Ah lithium iron phosphate battery, and a BMS main control unit. The system's operation flow is as follows: The BMS main control unit extracts the generator's operating characteristics through a ripple acquisition line and transmits the acquired ripple signal to an internal ripple ADC for detection. Subsequently, the data is uploaded to a cloud platform via a 5G communication link for data accumulation and power assessment. After calculating the current generator's power level attributes, the cloud platform issues control command signals. The BMS main control unit receives these commands and adjusts the DC-DC charger's duty cycle (e.g., executing a dynamic adjustment logic of charging for 3 minutes and stopping for 1 minute), achieving precise control of the main circuit's charging and stopping cycle. Through this system topology, this invention achieves closed-loop control of perception, decision-making, and execution, ensuring stable power supply to the parking load while significantly improving system integration and energy management efficiency.

[0108] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method of adaptive control of a vehicle alternator, characterized in that, include: S1, acquire the original operating data of the generator within the preset driving cycle, extract the output current of the generator based on the original operating data, and calculate the real-time speed of the generator; S2, perform linear regression analysis based on the real-time rotation speed and the output current, extract the regression slope, match the regression slope with a preset standard slope library to determine the candidate power level, and generate a trigger signal when the match is successful; S3. In response to obtaining the trigger signal, a multi-dimensional feature vector consisting of maximum current characteristics, average power characteristics, and voltage change rate characteristics is constructed. Based on the multi-dimensional feature vector, a recognition model is used to perform logical verification on the candidate power level to determine the final generator power level attribute information. S4. Based on the generator power level attribute information and the real-time status of the battery, construct an optimization objective function that couples the energy conversion efficiency, generator operating efficiency and battery life loss characteristics, and use a dynamic duty cycle adjustment mechanism to adaptively solve for the optimal on-time and off-time that match the generator power level.

2. The adaptive control method for a vehicle generator according to claim 1, characterized by, Before calculating the real-time rotational speed in step S1, the method further includes performing transient spectrum analysis on the original operating data to identify the physical topology category of the generator and thereby lock the number of magnetic pole pairs and rectifier pulsation number of the generator. The real-time rotational speed is calculated according to the formula , wherein, is the real-time rotational speed of the generator rotor; is the fundamental frequency of the ripple extracted from the original operating data; is the number of magnetic pole pairs; is the number of commutation pulses.

3. The adaptive control method for a vehicle generator according to claim 2, characterized by, Identifying the physical topology category of the generator includes: Step a: Upon detecting the engine start-up moment, continuously collect data at a preset sampling frequency. The data points are divided into several frames and Fourier transforms are performed to obtain the ripple fundamental frequency sequence of the start-up transient process. b step, extracting a front sub-sequence of the ripple fundamental frequency sequence to perform linear fitting, and calculating to obtain the ripple fundamental frequency sequence a dynamic rising slope changing over time; Step c: Extract the last segment subsequence of the ripple fundamental frequency sequence and calculate the mean to obtain the start-up end frequency; Step d involves mapping and matching the dynamic rising slope with multiple preset slope threshold intervals, and using the starting end frequency for interval correction to determine the physical topology category of the generator.

4. The adaptive control method for a vehicle generator according to claim 3, characterized by, The expression for the dynamic upward slope is: wherein is the first is the time stamp corresponding to the frame; is the arithmetic mean of the selected frame time stamps; is the first is the ripple fundamental frequency extracted from the frame by FFT; is the arithmetic mean of the selected frame fundamental frequency sequence; The expression for the start-up end frequency is: In the formula, is the total number of frames collected in the start-up phase, is the cumulative index, is the first frame corresponding to the ripple fundamental frequency, is the average number of frames.

5. The adaptive control method for a vehicle generator according to claim 3, characterized by, The decision logic for step d includes: When identified satisfy At that time, it was determined that the generator belonged to the 6-pole single-three category; When the dynamic rising slope is identified satisfy And the startup end frequency satisfy At that time, it was determined that the generator belonged to the 8-pole single-three category; When the dynamic rising slope is identified satisfy And the startup end frequency satisfy At that time, it was determined that the generator belonged to the 6-pole double-triple category; When the dynamic upward slope is identified satisfy And the startup end frequency satisfy At that time, it was determined that the generator belonged to the 8-pole double-triple category; wherein, is a first slope threshold value, is a second slope threshold value, is the end of run frequency threshold value.

6. The adaptive control method for a vehicle generator according to claim 3, characterized by, The determination of the physical topology category of the generator also includes: Calculate the ratio of the dynamic rising slope to the start-up end frequency; compare the ratio with a preset judgment threshold to verify the data consistency of the current start-up transient process and eliminate abnormal interference.

7. The adaptive control method for a vehicle generator according to claim 1, characterized by, Step S2 includes: S21, when the generator speed is increasing linearly, the generator output current is synchronously collected in a preset speed step to construct a feature sequence containing several sample points, and the least squares method is used to perform univariate linear regression fitting on the feature sequence to extract the regression slope. S22, calculate the Euclidean distance between the regression slope and each nominal equal standard slope in the preset standard slope library, select the minimum value in the Euclidean distance set, and when the minimum value is less than or equal to the preset allowable error threshold, determine the candidate power level and generate the trigger signal.

8. The adaptive control method for a vehicle generator according to claim 1, characterized by, The S3 step includes: using the identification model to sequentially perform fuzzification processing of the multidimensional feature vector, fuzzy inference based on the rule base, and defuzzification operation of the output variable to obtain the generator power level attribute information.

9. The adaptive control method for vehicle generators according to claim 8, characterized in that, The fuzzification process in step S3 includes: using a membership function combining trapezoidal and triangular shapes to map the maximum current characteristic to a low-value range. Median range High value range Fuzzy sets, and adjacent fuzzy sets are separated by a width of... The linear slope transition zone; Map the average power characteristic to a low value range Median range High value range Fuzzy set, and set the width at the decision boundary. or The membership degree of the slope transition zone.

10. The adaptive control method for vehicle generators according to claim 8, characterized in that, The fuzzy reasoning includes the introduction of non-steady-state interference interception criteria: When the voltage change rate is detected When the generator is determined to be in a stable adjustment phase, a consistency check is performed on the candidate power level. When the voltage change rate is monitored to be at When the current sampled data is determined to be affected by external random load access interference, non-steady-state suppression logic is executed to shield non-linear interference by stopping the update of the identification results or performing weight reduction processing.

11. The adaptive control method for a vehicle generator according to claim 8, characterized by, The defuzzification operation of the output variable includes: The output variable of the identification model is defined as the generator power level, and a single-point membership function is used to correspond to the output fuzzy set. ; Using the center of gravity formula Calculate the output clarity value And based on the principle of proximity, Mapped to the nearest nominal power rating, where, To determine the values ​​for the final generator model; For the fuzzy rule index participating in the summation calculation, ; This is the single-point nominal value corresponding to the generator's rated current; For the first The membership weights output by the fuzzy rules.

12. The adaptive control method for a vehicle generator according to claim 1, characterized by, The expression for the optimization objective function is: In the formula, , It is a preset fixed period; Duty cycle, , , These are the weighting coefficients. Bus voltage To set the charging current, This is the generator's optimal efficiency power point. This refers to the generator's rated power. For the battery's rated capacity, The maximum allowable SOC fluctuation, This is the charging rate penalty coefficient; The constraints of the optimization objective function are: In the formula, For the storage battery Real-time state of charge at any given moment; This represents the lower limit of the generator's output power within its efficient operating range. This refers to the upper limit of the generator's output power within its high-efficiency operating range. The preset minimum charging on-time; The preset minimum charging shutdown time.

13. The adaptive control method for a vehicle generator according to claim 1, characterized by, The dynamic duty cycle adjustment mechanism includes retrieving a control strategy mapping table corresponding to the power level attribute information to determine the duty cycle adjustment logic: If the generator power rating is 80A, a shallow charge and shallow discharge strategy is implemented: the initial search center of the duty cycle is set to 0.5, and the shutdown time is dynamically reduced according to the real-time change rate of SOC. If the generator power rating is 130A, the balancing strategy is executed: dynamically adjust the duty cycle optimization step size in the optimization objective function to increase the proportion of conduction time in the cycle; If the generator power rating is 150A, a deep sleep strategy is implemented: a symmetrical period with equal on-time and off-time is configured, and a soft shutdown mechanism that limits the current drop slope is implemented during the switch from on-time to off-time.

14. The adaptive control method for a vehicle generator according to claim 1, characterized by, The method further includes: The on-time and off-time parameters are encapsulated into a control signal in a preset communication protocol format and sent to the vehicle communication terminal; the battery management system parses the control signal and implements intermittent charging operation of the generator through the underlying hardware execution module.

15. The adaptive control method for a vehicle generator according to claim 14, characterized by, The underlying hardware execution module includes a DC contactor connected in series between the generator output terminal and the positive terminal of the battery, and a pre-drive circuit electrically connected between the DC contactor coil and the battery management system. The battery management system controls the main contacts of the DC contactor to close during the on-time and open during the off-time through the pre-drive circuit so that the generator enters the no-load standby mode.

16. The adaptive control method for a vehicle generator according to claim 2, characterized by, Calculating the real-time rotational speed includes the following steps: Step 1: Extract the raw voltage signal from the battery terminal using a non-invasive acquisition circuit, suppress the common-mode DC component in the raw voltage signal using a differential conditioning circuit, and extract the AC ripple signal therein. Step 2: The AC ripple signal is biased and conditioned and then converted from analog to digital to obtain digital ripple data; Step 3: Perform a fast Fourier transform on the digital ripple data to extract the ripple fundamental frequency, and calculate the real-time rotational speed by combining the number of magnetic pole pairs and the number of rectified pulsations.

17. An adaptive control system for a vehicle alternator, comprising: include: The data acquisition module is configured to acquire the original operating data of the generator within a preset driving cycle, extract the output current of the generator based on the original operating data, and calculate the real-time speed of the generator. The feature generation module is configured to perform linear regression analysis based on the real-time rotational speed and the output current, extract the regression slope, match the regression slope with a preset standard slope library to determine the candidate power level, and generate a trigger signal when the match is successful. The identification module is configured to construct a multi-dimensional feature vector consisting of maximum current characteristics, average power characteristics, and voltage change rate characteristics in response to receiving the trigger signal, and to perform logical verification on the candidate power level based on the multi-dimensional feature vector using an identification model to determine the final generator power level attribute information. The control module is configured to construct an optimization objective function that couples energy conversion efficiency, generator operating efficiency, and battery life loss characteristics based on the generator power level attribute information and the real-time status of the battery, and adaptively solve for the optimal turn-on time and turn-off time that match the generator power level using a dynamic duty cycle adjustment mechanism.

Images