Charging control method and system for chargers based on big data

Abstract

This invention discloses a charging control method and system for a charger based on big data. The invention relates to the field of big data technology. It determines corresponding charging state coefficients based on data identification of various big data combinations. A charging twin model of the charger is determined based on each charging state coefficient, the charging progress value displayed by the charger, and the corresponding charging scenario. Multiple abnormal charging data are identified based on the identification of the charging twin model. Data deviation combinations are determined based on each abnormal charging data, the corresponding charging stage, and the charger's charging efficiency value. The corresponding data deviation amount is determined based on the detection of this data deviation combination. Simultaneously, the corresponding optimization progress is determined based on the detection of the charger's charging progress. A corresponding charging control system is determined based on the optimization progress of the charging process, the charger's continuous charging time, and the corresponding charging target, triggering dynamic optimization of the charger's charging data, thereby improving the accuracy of the charging control system.

Description

Technical Field

[0001] This invention relates to the field of big data technology, and in particular to a charging control method and system for a charger based on big data. Background Technology

[0002] With the development of technology, chargers are gradually being applied to people's lives. Charging piles, as large-scale chargers, can charge electric vehicles. They collect multiple working data from the charger, determine the corresponding charging status based on the identification of multiple working data, and perform data management and control on the charger in the charging state. In the existing technology, collecting multiple charging data from the charger and managing and controlling multiple charging data to optimize the charging effect of the charger based on the adjustment of multiple charging data cannot form a corresponding charging twin model, and cannot guarantee the accuracy of data deviation, resulting in low accuracy of the charging control system. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a charging control method and system for a charger based on big data.

[0004] This invention provides a charging control method for a charger based on big data, comprising: Data detection is performed on the charger in the charging state, and multiple charging data are collected. Multiple sets of big data combinations are determined based on the multiple charging data, the corresponding charging progress nodes, and the charger temperature data. The charging state coefficient is determined based on the data identification of each big data combination. The charging twin model of the charger is determined based on each charging state coefficient, the charging progress value displayed by the charger and the corresponding charging scenario. The charging twin model includes a circuit simulation part, a thermodynamic simulation sub-part and an aging prediction sub-part. Based on the identification of the charging twin model, multiple abnormal charging data are identified. Based on each abnormal charging data, the corresponding charging stage and the charging efficiency value of the charger, a data deviation combination is determined. Based on the detection of the data deviation combination, the corresponding data deviation amount is determined. The system collects the charger's autonomous control system, determines the data control task based on the autonomous control system, the data deviation, and the charger's current charging mode, and triggers autonomous adjustment of the data deviation in the data control task to dynamically optimize the charger's charging process. The corresponding optimization progress is determined based on the detection of the charging process of the charger. The corresponding charging control system is determined according to the optimization progress of the charging process, the continuous charging time of the charger, and the corresponding charging target, and the dynamic optimization of the charging data of the charger is triggered. The charging control system includes power transmission dimension, thermal management dimension, chemical protection dimension, and power quality dimension.

[0005] This invention provides a charging control system for a charger based on big data, which is applied to the aforementioned charging control method for a charger based on big data.

[0006] Compared with the prior art, the beneficial effects of the present invention are: (1) The corresponding charging state coefficient is determined based on the data identification of each big data combination. The charging twin model of the charger is determined according to each charging state coefficient, the charging progress value displayed by the charger and the corresponding charging scenario. Multiple big data combinations are introduced to control the charging state coefficient and improve the accuracy of the charging twin model of the charger.

[0007] (2) Based on the identification of the charging twin model, multiple abnormal charging data are identified. Based on each abnormal charging data, the corresponding charging stage and the charging efficiency value of the charger, a data deviation combination is determined. Based on the detection of the data deviation combination, the corresponding data deviation amount is determined. Multiple abnormal charging data are controlled, the detection of the data deviation combination is realized, and the accuracy of the data deviation amount is improved.

[0008] (3) Based on the autonomous control system, the data deviation and the current charging mode of the charger, the data control task is determined, and the autonomous adjustment of the data deviation is triggered in the data control task to dynamically optimize the charging process of the charger; the corresponding optimization progress is determined based on the detection of the charging process of the charger, and the corresponding charging control system is determined according to the optimization progress of the charging process, the continuous charging time of the charger and the corresponding charging target, and the dynamic optimization of the charging data of the charger is triggered. The data control task is introduced, and the autonomous adjustment of the data deviation is triggered. The overall consideration of the optimization progress of the charging process, the continuous charging time of the charger and the corresponding charging target is realized, which improves the accuracy of the charging control system. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the charging control method for a charger based on big data in an embodiment of the present invention. Figure 2 This is a flowchart illustrating step S11 in the charging control method for a charger based on big data in an embodiment of the present invention. Figure 3 This is a flowchart illustrating step S12 in the charging control method for a charger based on big data in an embodiment of the present invention. Figure 4 This is a flowchart illustrating step S13 in the charging control method for a charger based on big data in an embodiment of the present invention. Figure 5This is a flowchart illustrating step S14 of the charging control method for a charger based on big data in an embodiment of the present invention. Figure 6 This is a flowchart illustrating step S15 in the charging control method for a charger based on big data in an embodiment of the present invention. Figure 7 This is a schematic diagram of the structural composition of the charging control system of the charger based on big data in an embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0011] Please see Figures 1 to 7 A charging control method for a charger based on big data, applied to big data scenarios; the charging control method for a charger based on big data includes: Step S11: Perform data detection on the charger in the charging state and collect multiple charging data. Determine multiple sets of big data combinations based on the multiple charging data, the corresponding charging progress nodes and the charger temperature data. Step S12: Determine the corresponding charging state coefficient based on the data identification of each big data combination, and determine the charging twin model of the charger according to each charging state coefficient, the charging progress value displayed by the charger and the corresponding charging scenario. Step S13: Based on the identification of the charging twin model, identify multiple abnormal charging data, determine the data deviation combination based on each abnormal charging data, the corresponding charging stage and the charging efficiency value of the charger, and determine the corresponding data deviation amount based on the detection of the data deviation combination. Step S14: Collect the charger's autonomous control system, determine the data control task based on the autonomous control system, data deviation, and the charger's current charging mode, and trigger autonomous adjustment of the data deviation in the data control task to dynamically optimize the charger's charging process. Step S15: Determine the corresponding optimization progress based on the detection of the charging process of the charger, determine the corresponding charging control system according to the optimization progress of the charging process, the continuous charging time of the charger, and the corresponding charging target, and trigger the dynamic optimization of the charging data of the charger.

[0012] refer to Figure 2 In step S11, the specific steps are as follows: S111: Collect multiple working data of the charger, determine the working status of the charger based on the multiple working data and the feedback signal of the charger, and monitor the working status of the charger in real time. At this time, perform data detection on the charger and mark multiple charging data during the detection process. S112: Determine the temperature data of the thermometer based on the temperature detection of the charger, and mark the dimensional range of the charger at each location. At the same time, collect the charging progress nodes of the charger, and determine multiple sets of big data combinations based on multiple charging data, corresponding charging progress nodes and the temperature data of the charger.

[0013] In the embodiments of this application, the system establishes a high-frequency, multi-channel synchronous sampling mechanism to decode protocol messages in real time, using this as a sensing input. A high-precision analog-to-digital converter (ADC) is used with a sampling period of milliseconds (e.g., 10ms-100ms) to collect analog signals from the charger output, focusing on electrical parameters such as output voltage, output current, and bus voltage. Simultaneously, digital messages uploaded by the charger's internal control unit and battery management system (BMS) are monitored via a controller area network (CAN) or PLC communication bus. These feedback signals cover key information such as estimated IGBT junction temperature, fan speed duty cycle, fault code register status, relay engagement status on the charger side, and the highest / lowest single-cell voltage of the battery pack on the BMS side.

[0014] Based on the collected values ​​and the parsed feedback signals, the system uses state machine logic to determine the precise operating mode of the charger. Through threshold comparison and logical judgment, it compares the collected real-time voltage / current values ​​with preset threshold ranges. For example, if the current value is maintained within ±2% of the set value and the voltage has not reached the upper limit, it is determined to be in "constant current (CC) state". The system maintains a finite state machine (FSM) internally. When a specific feedback signal sequence is received (such as handshake success signal + relay closure feedback + current establishment feedback), the state is triggered to jump from "standby" to "charging". Once the determination is successful, the current operating state tag is locked until a state transition condition is detected.

[0015] A sliding time window is established, and the statistical characteristics (such as variance and slope) of the data within the window are calculated in real time. By monitoring the rate of change of voltage or current, if the instantaneous rate of change exceeds the preset dynamic threshold (such as a sudden increase or decrease), it is identified as a transient event (such as a sudden load change or grid fluctuation). Based on the monitoring results, the system updates the current operating condition description in real time (e.g., from "stable charging" to "dynamic adjustment") to ensure that subsequent data analysis is based on the latest operating condition background.

[0016] Each data point is cleaned and tagged with rich metadata to form an effective big data combination, which is key to data assetization. Through data cleaning, outliers caused by sensor malfunctions and invalid data points caused by communication packet loss are removed, and missing data segments are filled in using interpolation algorithms. At the same time, the timestamps of electrical analog quantity acquisition and communication message reception are unified to the same reference clock (such as UTC time) to solve the time deviation caused by asynchronous sampling. On this basis, multi-dimensional labeling is performed: each data point is labeled with its current working state (such as the label ID representing the constant current stage); feature labels such as "steady state", "fluctuation", and "peak" are added according to data characteristics; and a unique session ID and sequence number are assigned to the data stream to ensure the orderliness of the data on the timeline.

[0017] Specifically, the charging control is performed on a charger that is charging an electric vehicle with a rated voltage of 400V and a capacity of 100kWh. The scenario is set at 15 minutes after the start of charging, when the battery SOC rises from 30% to 35%, and it is in the high-current constant-current charging stage.

[0018] The charger's MCU collects operating data in real time at a 20ms cycle using voltage and current Hall sensors. At time T, the collected output voltage is 395.2V and the output current is 249.8A. At the same time, the MCU reads the feedback signal from the BMS via the CAN bus, where the "charger output enable" bit is True, the "battery state of charge (SOC)" feedback value is 35%, and the "maximum single cell voltage" feedback is 4.15V.

[0019] The control system performs logical judgment on the collected data: it detects that 395.2V is lower than the constant voltage threshold of 400V, and the current of 249.8A is very stable. At the same time, the BMS feedback does not send "stop charging" or "reduce current" commands. Based on this, the state machine logic determines that the charger's current working state is locked as "CC constant current charging mode - high power area".

[0020] The system continued monitoring for the next 5 seconds. At time T+2 seconds, a slight fluctuation in the grid voltage was detected, causing the output current to briefly drop from 249.8A to 245.0A and then recover. The system calculated the rate of change of current and found that it momentarily exceeded the steady-state threshold. Therefore, the current monitoring status was temporarily marked as "Grid Disturbance - Dynamic Adjustment in Progress".

[0021] At the data point at time T+2 seconds, although the current drops, it is not a fault but a real response. The system cleans the data point, retains the data, and adds the following multi-dimensional tags: the basic tag is the ID of this charging session, the status tag is constant current mode, the event tag is the grid disturbance event, and the numerical tag records the voltage and current values ​​at that time. This set of data is encapsulated into a record with a timestamp and holographic tag and stored in the big data buffer.

[0022] Furthermore, the internal heat distribution of a charger is non-uniform, and a single ambient temperature value cannot reflect the true heat dissipation pressure. Therefore, a temperature sensor network is deployed in the charger's critical heat source areas. This includes the power module area monitoring the core junction temperature close to the IGBT / MOSFET heatsink, the transformer / inductor area monitoring the temperature rise of magnetic components, the PCB control area monitoring the motherboard chip temperature, and the air duct / environment area monitoring the temperature difference between the air inlet and outlet to assess heat dissipation efficiency. Each sensor is assigned logical spatial coordinates or area identifiers; for example, Zone_0 is set as the core power area, and Zone_1 as the secondary heat dissipation area. The collected temperature values ​​are bound to the corresponding area IDs, and the temperature gradient between areas is calculated and marked in the data structure.

[0023] Charging progress is a key anchor point on the timeline, determining the battery's current accepting capacity and responsiveness. In addition to the conventional SOC percentage, the charging process is discretized into physical feature nodes, including the trickle pre-charge stage for battery activation, the high-current constant-current stage for main energy injection, the inflection point of current peak reaching constant voltage transition point, and the float / tricrick end stage for energy replenishment and balancing. The SOC value is obtained by parsing BMS messages, and the inflection point is identified by combining the characteristics of locally acquired voltage / current curves. The acquired electrical data is strictly aligned to the current progress node, ensuring that each data packet has a "time phase" attribute, thereby accurately locating the physical stage of the charging process.

[0024] Electrical, temporal, and spatial quantities are combined using a Cartesian product to form a high-dimensional feature vector. Using a unified system clock as a reference, operational data (voltage / current), spatial temperature data (temperature values ​​for each region), and time progress data (SOC / stage identifier) ​​from the same time slice are aggregated. Derivative features are calculated based on the basic data, such as "temperature rise to power ratio" or "estimated internal resistance for the current stage." All these variables are encapsulated into a structured data object, whose combination includes timestamps, operational data sets, progress node identifiers, spatial temperature distribution vectors, and sets of derived features. Each data set records not only voltage and current but also progress nodes and dimensional ranges.

[0025] Specifically, the current time is 25 minutes after charging began; at this time, the battery SOC has increased from 35% to 60%, still in the high-power constant current charging zone, but the internal heat of the charger is gradually increasing, and the cooling fan is running at high speed; the charger's thermal management system scans through the internally integrated NTC thermistor array; the collected temperature data is as follows: the temperature of Zone_0 corresponding to the power module IGBT heat sink is 68.5°C, the temperature of Zone_1 corresponding to the transformer winding is 72.0°C, and the temperature of Zone_2 corresponding to the ambient air inlet is 38.0°C; the system calculates the spatial dimension features, and the temperature difference between the core Zone_0 and the air inlet Zone_2 is 30.5°C. This gradient value is marked as a "high heat load gradient" state; the data is spatially labeled to form a temperature vector and a high load gradient label.

[0026] By parsing the BMS message through the CAN bus, the current SOC is found to be 60%. At the same time, the MCU detects that the output voltage has not yet reached the set constant voltage threshold, and the current remains constant. The system determines that it is currently in the middle to late stage of the "CC high current constant charging stage". A time progress label is added to the current moment to clarify that it is currently in the middle to late stage of constant current and the SOC value is 60%.

[0027] The system integrates electrical data, temperature data, and progress data collected at the same time. At this moment, the electrical data is recorded as an output voltage of 580V and an output current of 300A. The system generates a set of big data composite objects numbered Data_Comb_00X, which encapsulates a precise timestamp, electrical data containing power parameters, progress tags indicating the middle and late stages of constant current and 60% SOC, and thermal dimension information including temperature values ​​of each region and high load gradient determination. In addition, it also includes derived indicators such as estimated energy efficiency indicators and thermal stress index. This set of Data_Comb_00X is stored in the big data cache pool. It is not just a string of voltage and current values, but a digital mapping of the complete physical scenario that "at SOC 60%, the power device temperature reaches 68.5°C with a large temperature difference".

[0028] refer to Figure 3 In step S12, the specific steps are as follows: S121: In multiple sets of big data combinations, data identification is performed on each set of big data combinations, and multiple key charging data of the charger are determined during the identification process. Based on multiple key charging data, corresponding charging behavior and power consumption efficiency coefficient of the charger, the corresponding charging state coefficient is determined. S122: Determine the charging progress value displayed by the charger based on the charger's progress recognition, determine multiple scene features based on the charger's surrounding detection, determine the charging model framework based on the multiple scene features and each charging state coefficient, and determine the charging twin model of the charger based on the charging model framework and the charging progress value displayed by the charger.

[0029] In the embodiments of this application, the system needs to perform pattern matching on multiple sets of input big data combinations, filter out redundant information, and lock in key variables that are sensitive to the control strategy; apply a sliding window algorithm to scan the big data of the time series, identify steady-state segments and transient segments in the data stream, for example, identify the "steady-state operating area" with the smallest current and voltage fluctuations through variance analysis, and extract the average value within this interval as the benchmark data; based on the pattern recognition results, extract representative key physical quantities from the big data combinations, including electrical characteristic extremes such as peak current and voltage ripple coefficient of the current cycle, thermodynamic characteristic values ​​such as the temperature rise rate of core power devices and the highest point of thermal field distribution, and dynamic response characteristics such as voltage drop amplitude and recovery time when the load changes abruptly.

[0030] A mapping relationship between key data and control logic is established. If the key data shows that the current fluctuates slightly around the reference value, the behavior is marked as "constant current maintenance"; if the data shows that the voltage reaches the threshold and the current decreases exponentially, the behavior is marked as "constant voltage trickle flow"; if the data shows that the voltage or current exhibits an unexpected step change, the behavior is marked as "abnormal adjustment". At the same time, a real-time energy flow model is constructed to calculate the instantaneous power conversion efficiency, i.e., the power consumption efficiency coefficient. The input power is calculated based on the voltage and current collected from the grid side, and the output power is calculated based on the voltage and current collected from the battery side, thus obtaining the power consumption efficiency coefficient. To be more accurate, a heat loss correction factor is introduced, and the power consumption of the cooling fan and the power consumption of the control board are included in the denominator to obtain a comprehensive efficiency coefficient that includes auxiliary energy consumption.

[0031] A multi-objective evaluation function is constructed, whose inputs include key charging data reflecting the current load intensity, charging behavior reflecting the current control stage, and power consumption efficiency coefficient reflecting the current energy efficiency level and heat loss. Based on this, the state coefficient is deconstructed and quantified. The charging state coefficient is usually defined as a vector containing multiple components, including thermal stress components based on key temperature data and efficiency loss, electrical stress components based on voltage ripple and current peak, and a comprehensive health index obtained by normalizing and weighting the above components. This index is between 0.0 and 1.0, where 1.0 represents the optimal state and 0.0 represents an impending fault or protection shutdown. Combined with historical data, the currently calculated state coefficient is corrected by Kalman filtering to eliminate instantaneous fluctuations caused by sensor noise.

[0032] Specifically, at the 32-minute mark after charging began, the charger remained in the high-power constant current (CC) phase, maintaining a high output power. However, due to rising ambient temperature and prolonged operation, the device began to experience a slight decrease in energy efficiency. The system received 200 sets of big data combinations from the last 10 seconds, including information on voltage, current, and multiple temperatures. Through time-series analysis, it was found that although the output voltage remained relatively stable at 580V, the transformer temperature exhibited a monotonically increasing trend, and the input current showed slight sawtooth fluctuations. Based on this, the system identified the following key data: the key current was 298.5A, slightly lower than the set value of 300A; the key temperature rise rate was 0.8°C / min, identified as a rapid temperature rise characteristic; and the voltage ripple was 1.2V, within the normal range.

[0033] Since the current remains around 298.5A and the voltage is stable, the system determines the current charging behavior as "constant current steady-state operation". In terms of efficiency calculation, the system calculates the input power to be approximately 183.2kW (including auxiliary power consumption) and the output power to be approximately 173.1kW, thus obtaining a power consumption efficiency coefficient of approximately 0.945 (i.e. 94.5%). By comparing with historical benchmarks, the system found that under the same power, the efficiency in the initial stage was 96.5%, but it has now dropped to 94.5%, indicating that heat loss has increased and the efficiency coefficient has decreased.

[0034] The system calls the evaluation function to perform multi-dimensional weighted calculations based on key temperature rise rate of 0.8°C / min (too high), power consumption efficiency coefficient of 0.945 (too low, meaning more energy is converted into heat), and charging behavior of "constant current steady state". The component calculation results show that due to low efficiency leading to heat accumulation, the thermal stress component is calculated to be 0.78, and the energy efficiency component is calculated to be 0.75. After weighted fusion, the system generates the final charging state coefficient vector, which includes thermal stress, energy efficiency level, and a comprehensive health index of 0.76. This set of coefficients of 0.7+ clearly indicates to subsequent steps that although the charger can still maintain "constant current" behavior, the temperature rise is intensified due to the decrease in power consumption efficiency.

[0035] Furthermore, the charging progress value displayed by the charger is determined based on the charger's progress recognition, and multiple scene features are determined based on the charger's surrounding detection. A charging model framework is determined based on multiple scene features and various charging state coefficients. A charging twin model of the charger is determined based on this charging model framework and the charging progress value displayed by the charger. This approach takes into account both the charging model framework and the charging progress value displayed by the charger as a whole, ensuring the accuracy of the charger's charging twin model. At the same time, multiple sets of big data combinations are introduced to control the charging state coefficients, further improving the accuracy of the charger's charging twin model.

[0036] At this point, the charging progress is the main variable on the time axis, determining the model's time step position and boundary conditions during the simulation process. A simple percentage is insufficient to reflect the complex physicochemical processes. The system performs multi-source progress data fusion, not relying solely on the SOC percentage reported by the BMS, but integrating the ampere-hour integral method, open-circuit voltage calibration method, and Kalman filter estimation values ​​to improve data accuracy. Simultaneously, physical stage mapping is performed, mapping the linear SOC value to nonlinear physical charging stages, including the pre-charge stage mainly used for lithium-ion intercalation activation, the high-current constant-current stage mainly for lithium-ion migration with low internal resistance, and the constant-voltage trickle-current stage where polarization increases and the charging current decays exponentially. Finally, a normalized progress value of 0.0 to 1.0 is generated, which not only includes the percentage of charge but also corrects for capacity reduction errors caused by battery aging.

[0037] The system performs multi-dimensional environmental perception, acquiring ambient temperature and humidity through integrated temperature and humidity sensors, monitoring grid voltage deviation, frequency deviation, and voltage harmonic distortion rate at the access point, and detecting the presence of other heat sources around the charger. Based on this, scene features are quantified, mapping the above environmental data into scene feature vectors. The system labels features based on the perceived data; for example, ambient temperature above a certain threshold is labeled as a "high temperature scene," and high harmonic distortion rate is labeled as a "grid pollution scene." At the same time, features are weighted according to sensitivity; for example, for air-cooled chargers, ambient temperature has a higher weight than humidity, thereby constructing a scene feature vector that reflects the actual operating conditions.

[0038] The system pre-configures multiple control logic frameworks, such as the "speed-priority framework," "life-priority framework," and "thermal safety-priority framework," and makes multi-objective decisions based on scenario characteristics and state coefficients. For example, if the scenario characteristic is "high temperature" and the state coefficient shows "high thermal stress," the "thermal safety-priority framework" is automatically matched. The constraints of this framework will strictly limit the temperature rise slope of the power devices. Next, parameter initialization and boundary setting are performed. The previously calculated charging state coefficient is used as the initial parameters of the model, and the charging progress value is used to set the time boundary and objective constraints of the model. Finally, a charging twin model is instantiated in the digital space. This charging twin model includes a circuit simulation part, a thermodynamic simulation sub-part, and an aging prediction sub-part. The charging twin model will receive input data in real time and run in millisecond steps, outputting the predicted voltage, current, and temperature curves to build the corresponding digital mirror.

[0039] Specifically, the timeline advances to the 55th minute after charging begins; at this point, the charger is nearing the end of charging, and it is midday in the height of summer, with the charging station at full capacity; the system performs data fusion, the BMS reports a SOC of 95%, and the system detects that the charging current has naturally decreased to 40A, exhibiting characteristics of the constant voltage stage; based on this, the system determines that it is currently in the "late stage of constant voltage trickle charging"; when determining the normalized progress value, although the SOC is 95%, considering the extremely small current and severe battery polarization, the system generates a progress value of 0.98, which means that in the model logic, this charging task has approached the physical limit.

[0040] Through environmental perception, the ambient temperature sensor reading is 42℃, which is an extremely high temperature. At the same time, it is detected that the adjacent Type B charger is operating at full power, and the hot air it is emitting is blowing directly into the charger's air inlet. Based on this, the system extracts two features: "extreme high temperature environment" and "external thermal interference", and combines them to generate a scene vector containing these two labels.

[0041] The system combines scene vectors and previously calculated state coefficients (assuming high thermal stress) to decide against using the general "fast termination framework," instead matching an "extreme thermal protection framework." This framework relaxes charging time requirements, drastically lowers temperature control thresholds, and forces the cooling system to operate at full speed. During parameter injection and instantiation, the system injects a progress value of 0.98 into the model, sets the model to "tailing control" mode, and slows down the response speed of the current loop to prevent overshoot. Simultaneously, scene features are injected into boundary conditions, the ambient temperature parameter is set to 42℃, and an "external thermal radiation source" item is added to the thermal model. At this point, the digital twin model of the charger is officially generated. During background operation, the model deduces that although the current is only 40A, due to the high air inlet temperature of 42℃ and the influence of thermal radiation from neighboring units, the temperature of the bus capacitor will still slowly rise above the critical value in the next 10 minutes without intervention.

[0042] refer to Figure 4 In step S13, the specific steps are as follows: S131: Dynamically identify the charging twin model, identify multiple abnormal contents during the identification process, determine the corresponding abnormal charging data based on the identification of each abnormal content, collect multiple abnormal charging data, and determine the corresponding charging stage based on the tracing of each abnormal charging data. S132: Collect the charging efficiency value of the charger, determine the first level of deviation based on the charging efficiency value of the charger and each abnormal charging data, determine the second level of deviation based on the charging efficiency value of the charger and the charging stage corresponding to each abnormal charging data, and determine the corresponding data deviation combination based on the first level of deviation and the second level of deviation. S133: Based on the detection of the data deviation combination, multiple data deviation items are determined, and multiple deviation ranges are determined according to the identification of each data deviation item. Based on the multiple deviation ranges, the current charging mode of the charger, and the charging progress value displayed by the charger, the corresponding data deviation amount is determined, and the corresponding charging impact content is marked.

[0043] In the embodiments of this application, the digital twin model is synchronized with the physical charger in time, and the current expected state is simulated in real time in virtual space, and the residual between the observed physical quantity value and the model prediction value is calculated in real time. The system establishes a dynamic confidence interval for each key parameter. This interval is calculated in real time based on the current charging mode, ambient temperature and load conditions, rather than being fixed. In terms of judgment logic, if the residual exceeds the boundary of the confidence interval and lasts for a certain time window, the system determines it as "abnormal". The system will further identify the type of abnormality, including numerical abnormalities of parameter exceeding limits, trend abnormalities of rate of change, and phase abnormalities of AC power factor deviation.

[0044] The system maintains a circular buffer that always caches the raw high-frequency sampling data from the most recent N seconds. When an anomaly is detected, the system immediately locks the buffer to prevent data from being overwritten. In the multi-dimensional anomaly data extraction stage, the system not only collects the variable that triggered the anomaly, but also simultaneously collects all-dimensional data before and after that timestamp, including electrical quantities such as three-phase input voltage / current, DC output voltage / current, and PWM drive duty cycle, thermal quantities such as readings of all temperature sensors and air inlet / outlet wind speed, as well as control quantities such as BMS requested current, actual charger output current, and closed-loop control error value. This locked set of data with precise timestamps constitutes the "abnormal charging data".

[0045] Using the timestamps in the abnormal data set, the system searches the historical database and aligns the time with the main timeline of the current charging process. The system combines BMS messages, charger control state machine, and battery characteristic curves to determine the specific stage at which the abnormality occurred. These stages include the pre-charging stage for low-current activation, the constant-current high-power stage as the main energy transfer period, the constant-voltage / float charging stage with gradually decreasing current, and the shutdown cooling stage after charging is completed to dissipate residual heat. The final output marks the abnormal data as an abnormality of a specific stage.

[0046] Specifically, at the 22nd minute after charging began, the charger was in the constant current high-power charging phase (approximately 40% SOC), with the output current stable at 300A. However, the power grid where the charging station was located suddenly experienced a millisecond-level voltage drop. The charger's digital twin model performed a model simulation, predicting that under the current grid input voltage, the DC output voltage should be stable at 560.0V, and the ripple coefficient should be less than 1%. At 22 minutes and 00.500 seconds, the system performed residual monitoring and detected that the DC output voltage instantly dropped to 545.0V, while simultaneously detecting that the PWM drive duty cycle instantly reached its saturation value. The system calculated the residual to be 15.0V, while the current dynamic confidence interval was only ±2.0V. Since the residual was far beyond the confidence interval, the system determined the anomaly to be "instantaneous DC output voltage drop anomaly," which was classified as an "input disturbance conduction anomaly."

[0047] The system immediately locked a 2-second time window before and after the event. In the collected data set, it recorded the DC side output voltage dropping to 545.0V and the output current slightly dropping to 295.0A. The AC side data showed that the A-phase input voltage dropped from 380V to 365V. The control data showed that the bus voltage dropped from 700V to 680V and the PID controller output saturated. These data formed a complete abnormal charging data package, which was used to analyze the response characteristics of the control system.

[0048] The system retrieved historical records and confirmed that the current time point was in the middle of the charging process. A status query showed that the BMS reported a SOC of 42%, the charger's state machine indicated constant current charging, and the battery characteristics showed it was in the low internal resistance region with the strongest current acceptance. Based on this, the system precisely defined this anomaly as "mid-stage of constant current high-power charging." Since it was in the middle of the constant current charging cycle (CC), the drop in bus voltage directly affected the maintenance of output power. The system marked this data as a voltage drop anomaly in this stage, indicating that subsequent algorithms have insufficient ability to suppress input disturbances in this stage, requiring optimization of control parameters.

[0049] Furthermore, the charging efficiency value of the charger is collected. Based on the charging efficiency value of the charger and various abnormal charging data, the first level of deviation content is determined. Based on the charging efficiency value of the charger and the charging stage corresponding to each abnormal charging data, the second level of deviation content is determined. Based on the first level of deviation content and the second level of deviation content, the corresponding data deviation combination is determined, which takes into account the overall consideration of the first level of deviation content and the second level of deviation content, and ensures the accuracy of the corresponding data deviation combination.

[0050] At this point, the system calculates the instantaneous charging efficiency value in real time based on high-precision sampling data, which is the ratio of DC-side output power to AC-side input active power. A sliding window averaging algorithm is typically used to smooth instantaneous fluctuations. In the correlation analysis based on efficiency and abnormal data, the system compares the calculated efficiency value with the theoretical benchmark efficiency under this operating condition. If the efficiency value is significantly lower than the benchmark value, and the abnormal data shows an increase in temperature, the first level of deviation is defined as "high heat loss deviation," meaning that the abnormality has led to a large amount of energy being converted into waste heat. If the efficiency value is basically normal, but the abnormal data shows voltage or current waveform distortion, the first level of deviation is defined as "electrical characteristic drift deviation," meaning that the equipment parameters have changed, but have not yet caused significant energy loss.

[0051] The system extracts the charging stage tags corresponding to abnormal data and analyzes the electrical characteristics of that stage. For example, in the high-current constant-current stage, the internal resistance is small and the current is large, making it sensitive to heat accumulation; in the high-voltage constant-voltage stage, the voltage is high and the current is small, making it sensitive to overvoltage and lithium plating; in the trickle-down maintenance stage, the current is extremely small, making it sensitive to current control accuracy. In the deep attribution based on the coupling of stage and efficiency, the system makes specific judgments: if the abnormality occurs in the constant-current stage and the efficiency is low, it indicates that the thermal resistance of the device is abnormal under high current, and the second level of deviation is defined as "heavy load thermal saturation deviation"; if the abnormality occurs in the constant-voltage stage, although the efficiency is normal, the voltage fluctuation is abnormal, indicating that the battery polarization is severe, and the second level of deviation is defined as "polarization internal resistance deviation"; if the abnormality occurs at the constant-current / constant-voltage switching point, manifested as control oscillation, the second level of deviation is defined as "loop regulation hysteresis deviation".

[0052] The system combines the first-level deviation (performance perspective) and the second-level deviation (spatiotemporal perspective) as orthogonal dimensions to construct a multi-dimensional vector. This combination of data deviations is usually in a structured data form. This combination has extremely high application value and directly determines the selection of subsequent control strategies. For example, the combination of "high heat loss deviation" and "heavy load heat saturation deviation" should trigger "forced cooling + power reduction", while the combination of "electrical characteristic drift deviation" and "polarization internal resistance deviation" should trigger "constant voltage threshold correction + current reduction".

[0053] Specifically, at the 25-minute mark after charging began, the charger was in the high-current constant-current (CC) charging phase, with a state of charge (SOC) of approximately 45%. Due to dust accumulation on the cooling fan, its heat dissipation capacity had decreased. Regarding the first deviation, the system collected input power data at 180.5 kW and output power data at 171.0 kW, calculating a real-time charging efficiency of approximately 94.7%. At this power point, the charger's theoretical baseline efficiency should be 96.5%. The system detected abnormal charging data at this time, showing an IGBT temperature as high as 85°C, while the normal temperature should be 70°C. Based on the significant decrease in efficiency accompanied by high temperature, the system determined the first deviation to be "high heat loss deviation," indicating that a large amount of energy was lost due to heat loss.

[0054] Regarding the second deviation, according to the traceability of S131, it was confirmed that the anomaly occurred during the "high current constant current charging stage," with the output current maintained at 300A. In the deep attribution analysis, the system believes that during the CC stage, the high current passing through the bus and power devices will generate significant Joule heat loss. Combining the characteristics of low efficiency (high heat loss) and CC stage (high current), the system analysis concludes that the heat dissipation system is no longer able to offset the heat generated by the high current. Therefore, the system determines the second deviation to be "heavy load thermal saturation deviation," which indicates that the thermal equilibrium point of the equipment has drifted to the danger zone under heavy load conditions.

[0055] The system cross-validates the results from the first level (energy side) and the second level (spatiotemporal side); finally, it determines that in the corresponding data deviation combination, the first level is "high heat loss", which represents the loss in the energy dimension; the second level is "saturation in constant current stage", which represents the state in the spatiotemporal dimension; the combination ID is identified as "thermal risk in constant current mode", which clearly tells the control system: "The charger has caused serious heat loss (efficiency reduction) in the constant current stage due to poor heat dissipation, and the current state is unsustainable".

[0056] Therefore, based on the detection of this data deviation combination, multiple data deviation items are identified. Based on the identification of each data deviation item, multiple deviation ranges are determined. Based on multiple deviation ranges, the current charging mode of the charger, and the charging progress value displayed by the charger, the corresponding data deviation amount is determined, and the corresponding charging impact content is marked. This approach takes into account multiple deviation ranges, the current charging mode of the charger, and the charging progress value displayed by the charger as a whole, ensuring the accuracy of the corresponding data deviation amount. At the same time, it controls multiple abnormal charging data, realizing the detection of this data deviation combination and improving the accuracy of the data deviation amount.

[0057] At this point, the system maintains a deviation feature mapping table. Based on the input deviation combination, it retrieves the core physical quantity that caused the combination. Common data deviation items include thermal resistance deviation item mapping the difference between actual and theoretical temperature rise, voltage regulation deviation item of static error between output voltage and target voltage, phase control deviation item of phase shift between switching transistor drive signal and current, and ripple noise deviation item of high-frequency noise component in DC output. During the identification process, the system will perform correlation analysis on each candidate item. For example, if the deviation combination points to "efficiency loss" and "thermal saturation", the system will prioritize locking "thermal resistance deviation item" and "switching loss deviation item" which are strongly related to power loss, and exclude irrelevant items.

[0058] The system considers the constraints of the current charging mode. For example, in "Fast Mode," to pursue speed, the system allows for larger fluctuations in voltage and current, with a wider deviation range. In "Long-Life Mode" or "Fine Maintenance Mode," to protect the battery, the deviation range is set to be extremely narrow. At the same time, the system considers the constraints of the charging progress value. In the early stage of charging, the battery polarization is small, and the tolerance for current fluctuations is high, with a wide current deviation range. In the late stage of charging, the battery is close to full charge and is prone to overcharging, requiring extremely high voltage accuracy, with an extremely narrow voltage deviation range. Based on the current mode and progress, the system calls the preset rule engine to calculate the "allowable fluctuation threshold range" corresponding to each data deviation item. This range constitutes the reference for determining whether the deviation exceeds the standard.

[0059] The system calculates the absolute deviation between the actual observed value and the benchmark value, and combines this with the determined "allowable fluctuation threshold range" to calculate the normalized data deviation. If the actual deviation is within the allowable range, it indicates a healthy system. If it exceeds the allowable range, the calculated value will indicate the degree of exceeding the limit, which is the key basis for determining the intensity of regulation in subsequent steps. Based on the magnitude of the normalized data deviation and the type of deviation item, the system assigns specific impact labels, clearly informing the control system "what the consequences will be if no intervention is made." These labels include safety impacts such as "insulation breakdown risk" and "thermal runaway risk," lifespan impacts such as "battery lithium plating damage" and "electrolyte drying," performance impacts such as "limited charging power" and "low energy feedback efficiency," and user experience impacts such as "extended charging time" and "range jump."

[0060] Specifically, the time progresses to the 25th minute after charging begins; the charger is in high-current constant current (CC) mode, and the progress value shows SOC 45%; according to the analysis of S132, there is currently a deviation combination of "heavy load thermal saturation"; the system receives a deviation combination containing high heat loss and constant current stage saturation. Through decoupling and identification, the system identifies that the main problem is the conversion of energy loss into heat energy, so the first deviation item is locked as "thermal resistance deviation item"; at the same time, due to the change in semiconductor carrier mobility caused by the increase in temperature, the stability of the output voltage is affected, so the second deviation item is locked as "voltage steady-state error item".

[0061] The charger is currently in "intelligent temperature control mode," which balances speed and safety. In this mode, the system specifies a strict allowable range for voltage steady-state error, set at ±0.5V. Simultaneously, the current progress is SOC 45% (mid-CC). Based on battery characteristics, the mid-CC period is a high-current input period, requiring high voltage accuracy. Therefore, the system determines the dynamic deviation range of the "voltage steady-state error" to be [-0.5V, +0.5V]. For the "thermal resistance deviation," the system, according to the safety strategy in this mode, sets the allowable temperature rise slope range to ≤0.5°C / min.

[0062] Regarding the deviation calculation, the actual detected output voltage was 0.8V lower than the target value, with an absolute deviation of 0.8V. Since the allowable range is ±0.5V, the normalized deviation is calculated to be 1.6. The actual temperature rise slope of the thermal resistance item is 0.8°C / min, with an allowable range of 0.5°C / min, and the normalized deviation is also calculated to be 1.6. Based on the calculation results that both items exceed the limit by 60%, the system generates charging impact labels, including a thermal impact label of "cumulative thermal risk (requires derating)" and a control impact label of "decreased control stability (loop prone to oscillation)". The system generates a data package containing the voltage error and thermal resistance slope items, clarifying their exceedance multiples and corresponding negative impacts.

[0063] refer to Figure 5 In step S14, the specific steps are as follows: S141: Real-time monitoring of the charger's database, determination of the charger's autonomous control system based on the charger's database and charging scenarios, and determination of multiple autonomous control items based on the identification of the autonomous control system. S142: Determine the first level of data control content based on multiple autonomous control items and data deviation; determine the second level of data control content based on multiple autonomous control items and the current charging mode of the charger; and determine the data control task based on the first level of data control content and the second level of data control content. S143: In the data control task, multiple data control nodes are identified based on the identification of the data control task. The corresponding autonomous control content is determined according to the multiple data control nodes, the charging progress value displayed by the charger, and the data deviation. Autonomous control of the data deviation is triggered, and the charging process of the charger is optimized based on the autonomous control of the data deviation.

[0064] In the embodiments of this application, the system performs a multi-dimensional data pool scan, including hardware status monitoring, real-time reading of junction temperature data of power devices, ESR degradation trend of capacitors, and cumulative operating time of heat dissipation system; electrical parameter monitoring, obtaining current bus voltage stability, AC side power factor, and DC side ripple current statistics; and historical operating condition backtracking, retrieving the historical control performance of the charger under similar scenarios to check for any bad records; and based on the monitoring data, the system constructs a dynamic constraint set for the current moment.

[0065] The system performs charging scenario feature alignment, including environmental scenario identification to determine whether the current environment is high temperature and humidity, extreme cold and low temperature, or high altitude and low pressure; load scenario identification to determine whether the battery is in the high-rate fast charging, trickle maintenance, or dynamic pulse repair stage; and grid scenario identification to determine whether the grid support capability is strong or weak. Based on the above scenario features, the system calls the matching control logic architecture from the cloud or local storage, such as the "thermal safety priority system" which focuses on sacrificing efficiency for low temperature rise in high temperature scenarios, the "grid adaptability system" which focuses on suppressing harmonics and stabilizing input power in weak grid scenarios, or the "fast response system" which focuses on improving bandwidth and fast current tracking in fast charging scenarios.

[0066] The system decouples and categorizes control items, including main power loop control items such as output current regulation, output voltage regulation, and charging curve slope; auxiliary energy management items such as switching frequency and PFC target value; and thermal management control items such as fan speed, liquid cooling pump flow rate, and semiconductor duty cycle. At the same time, the system combines the database constraints in step 1 to determine the availability of items and eliminate unusable items. For example, if the fan is already fully loaded, the "fan speed" item is marked as saturated and is no longer used as an effective regulation variable. Instead, the system relies on the "current reduction" item.

[0067] Specifically, the timeline advances to the 30th minute after charging begins; this is midday in summer, with an ambient temperature as high as 42°C; the charger is operating at high power, and the database shows that its cooling fan has been running for a long time, resulting in slight performance degradation; the system reads the charger's local database and finds that the real-time junction temperature of the key power module has reached 82°C, with only a 3°C margin remaining before reaching the 85°C protection threshold; at the same time, the database records show that when the fan speed is at 90% load, the airflow is 5% lower than the nominal value; based on this, the system constructs an emergency constraint: "The temperature rise rate must be controlled within 0°C / min, and it can no longer rely solely on fan speed increases."

[0068] The system identified the environmental characteristics as "extreme high temperature" and the load characteristics as "continuous high power output". The system determined that the conventional "constant voltage and constant current control system" could not cope with the current heat dissipation challenge. Therefore, it dynamically loaded the "adaptive derating control system based on dynamic feedback of thermal resistance" from the algorithm library. The core logic of this system is "to obtain the maximum thermal safety margin with the minimum power sacrifice".

[0069] Based on the "adaptive derating control system," the system identifies the following available autonomous control items: Item A is output current limiting, which reduces heat generation to lower the temperature; Item B is dynamic adjustment of switching frequency, which slightly reduces switching losses by lowering the frequency; Item C is dead time injection, which balances losses and efficiency by finely adjusting the dead time. In the item availability confirmation stage, the system assesses that the "fan speed" item has approached its physical limit due to performance degradation recorded in the database, so it is marked as "auxiliary / saturated state" and used only as a maintenance item, not as an adjustment item. The system determines that "output current limiting" is the primary control method and "switching frequency adjustment" is the secondary control method.

[0070] Furthermore, the first level of data control content is determined based on multiple autonomous control items and data deviation, the second level of data control content is determined based on multiple autonomous control items and the current charging mode of the charger, and the data control task is determined based on the first and second level of data control content. This approach takes into account both the first and second level of data control content as a whole, ensuring the accuracy of the data control task.

[0071] At this point, the system establishes a mapping relationship between the deviation amount and the control intensity level. For example, for a slight deviation identified as a "warning zone", the first level of content is "fine-tuning parameters"; for a severe deviation identified as an "intervention zone", the first level of content is "changing the topology or significantly reducing the derating". At the same time, the system performs item-deviation coupling analysis to match the deviation type with the autonomous control item. For example, if the deviation amount is "thermal resistance deviation" exceeding the standard, the matching control item "output current limiting" is matched, and the first level of content is set to "negative adjustment of current"; if the deviation amount is "voltage steady-state error" exceeding the standard, the matching control item "duty cycle compensation" is matched, and the first level of content is set to "increase the PI loop integral gain".

[0072] The system reads the currently active charging mode definition file and extracts key constraint parameters, including power / current adjustment step size limits, maximum / minimum threshold values, and dynamic response bandwidth limits. For example, in silent mode, the fan speed adjustment rate is strictly limited to prevent whistling. In the constraint-based control correction stage, the system inputs the first level of data control content into the mode constraint filter. If the first level of content requires a large adjustment but the current mode limits the adjustment step size, the second level of content corrects it to multiple step adjustments. If the first level of content requires the fan to run at full speed but the current mode is a nighttime silent mode, the second level of content forces the fan command to be limited to a specific speed and instead reduces the current to assist in cooling.

[0073] When the correction direction of the first layer of content conflicts with the constraints of the second layer of content, the system resolves the conflict according to the principle of "safety weight > mode weight > efficiency weight". If the deviation endangers safety, the task will ignore the mode constraints and execute the aggressive strategy of the first layer of content. The system generates a structured data control task object, which includes the task type (such as current ramp adjustment, fan PWM setting), the final target value after fusion, the adjustment speed (execution rate) and the priority of the task in the control queue.

[0074] Specifically, the time progresses to the 30th minute after charging begins; the charger is in "intelligent temperature control mode" with a SOC of 50%; the data deviation output by S13 is 1.8 for thermal resistance deviation, which is a serious over-limit, indicating extremely rapid temperature rise; based on the serious deviation of 1.8, the system determines that if no intervention is made, the device will trigger over-temperature protection shutdown in a short time; the system matches the "output current limit" and "fan speed" in the autonomous control project, and in order to quickly reduce the thermal resistance deviation, the first step is aggressively set as follows: immediately reduce the output current by 80A (from 300A to 220A) and increase the fan speed to 100%.

[0075] The system reads the constraints of the "intelligent temperature control mode," which stipulates that to prevent drastic current fluctuations from damaging the battery's capacity, the maximum current adjustment rate must not exceed 10A / s. The first requirement is to "immediately reduce by 80A," which implies an instantaneous drop and violates the mode's rules. Therefore, the second requirement intervenes, correcting the adjustment action to a "smooth, ramp-down" decrease. Specifically, the current adjustment decreases by 80A at a rate of -10A / s. Although the fan adjustment allows for a fast response, to extend the motor's lifespan, the ramp rate is limited to 20% / s from 100%.

[0076] The system binds the "target value" of the first level of content with the "execution rate" of the second level of content to generate the final data control task. This task issues action commands to the DC-DC converter current loop and cooling fan: start the current ramp function, the target value drops from 300A to 220A, and the ramp rate is -10A / s; start the fan ramp function, the target value rises from 75% to 100%, and the ramp rate is 20% / s. This task is marked as Critical priority and is ready to be executed immediately.

[0077] Therefore, in the data control task, multiple data control nodes are identified based on the identification of the data control task. The corresponding autonomous control content is determined based on the multiple data control nodes, the charging progress value displayed by the charger, and the data deviation. Autonomous control of the data deviation is triggered, and the charging process of the charger is optimized based on the autonomous control of the data deviation. This takes into account the overall consideration of multiple data control nodes, the charging progress value displayed by the charger, and the data deviation, ensuring the accuracy of the corresponding autonomous control content.

[0078] At this point, the system analyzes the charger's hardware control topology and maps tasks to nodes at different levels, including application layer nodes responsible for issuing overall strategies (such as BMS communication protocol parsing nodes and charging strategy decision nodes), control layer nodes responsible for algorithm calculations (such as PID controller nodes, PLL phase-locked loop nodes, and Park conversion nodes), and drive layer nodes responsible for physical execution (such as PWM wave generator nodes, IGBT / MOSFET drive signal nodes, and relay control nodes). Based on the type of data control task, the system selects the set of active nodes. For example, if the task involves "current regulation," the current loop PI controller node is activated; if the task involves "heat dissipation," the fan speed control node is activated.

[0079] The system constructs a multi-dimensional control function that comprehensively considers the impact of charging progress value and data deviation. The charging progress value determines the granularity of control. In the early stage of charging, the battery internal resistance is small and the polarization is weak, allowing for a larger dynamic range in the control content, and the node response speed is set to "fast". In the late stage of charging, the battery is close to full charge and is prone to lithium plating, so the control content is set to "fine-tuning", and the node gain parameter needs to be appropriately reduced. The data deviation determines the adjustment step size. The larger the deviation, the larger the adjustment step size of the node, but it needs to be limited within the maximum safe slope. The system generates new current reference values ​​for current node nodes, switching frequency or phase shift angle for PWM nodes, and fan PWM duty cycle for auxiliary source nodes, among other autonomous control content.

[0080] The calculated control parameters are encapsulated into real-time control messages and sent to each data control node via the internal bus to drive hardware actions. During the deviation reduction and process optimization phase, the system continuously monitors new data deviations. If the deviation gradually converges to a safe range, the system determines that the control is effective and enters the "maintain steady state" phase. If the deviation decreases too quickly, it indicates that the adjustment has been overdone, and the system will activate the "callback mechanism" to slightly increase the current or decrease the fan speed in order to find the Pareto optimal solution for energy consumption and speed. Through this continuous fine-tuning, the charger no longer rigidly executes the preset curve, but instead follows an adaptive optimal trajectory that fits the current battery health, ambient temperature, and power grid conditions.

[0081] Specifically, the time progresses to the 32nd minute after charging begins; the charger executes the task issued by S142, which needs to deal with a severe thermal resistance deviation (value of 1.8), and the current charging progress value is 0.55 (SOC is about 55%, in the mid-to-late stage of CC); the system analyzes that the control tasks are "ramp-down current at a rate of -10A / s" and "improving heat dissipation"; based on this, the system locks the main control node to the digital current loop PI regulator node inside the DSP control chip, the execution node to the PWM wave generator node of the DC / DC converter, and the auxiliary node to the motor drive IC node of the cooling fan.

[0082] In the parameter fusion calculation, considering that the progress value is 0.55, which is in the mid-to-late stage of CC, the system determines that the battery can still withstand a large current but is sensitive to heat accumulation. Therefore, the adjustment strategy is set to "medium-speed current reduction". At the same time, given that the thermal resistance deviation value of 1.8 is a serious deviation, the system decides to make full use of the adjustment authority. For the current node, the system calculates the target current as 220A, sets the slope register to -10A / s, and fine-tunes the proportional gain to 90% of the original value to prevent voltage overshoot during the current reduction process. For the fan node, based on the temperature rise slope deviation, the target duty cycle is calculated to be 100%, and a soft-start time of 2 seconds is set to prevent excessive mechanical stress.

[0083] After the control command was issued, the charger's output current began to decrease linearly at a rate of 10A / s, and the fan speed climbed to its maximum speed within 2 seconds. At the 33rd minute, the current dropped to 220A, and the system detected that the rate of increase in IGBT junction temperature had slowed significantly, with the new thermal resistance deviation dropping to 0.9, entering the safe zone. At this point, the system found that although the thermal risk had been eliminated, the charging power loss was significant. To optimize the user's charging time experience, the system initiated "fine-tuned callback" while ensuring safety. The system fine-tuned the current loop, raising the output current to 230A at an extremely slow rate of 2A / s and maintaining it stable. The charger successfully avoided high-temperature protection shutdown and stabilized the charging power at the maximum allowable value of the device's thermal limit, achieving optimal safety and charging speed.

[0084] refer to Figure 6 In step S15, the specific steps are as follows: S151: Mark the charging process of the charger, determine the corresponding charging progress table based on the identification of the charging process of the charger, determine the corresponding node to be optimized according to the traversal of the charging progress table, and mark the optimization progress of the node to be optimized. S152: Collect the continuous charging time of the charger, determine the charging control framework based on the continuous charging time of the charger and the optimization progress of the node to be optimized, and mark the charging target corresponding to the charger. Construct the corresponding charging control system based on the charging target and the charging control framework. S153: Based on the identification of the charging control system, multiple control measures are determined, and the corresponding dynamic optimization content is determined according to the multiple control measures, the current charging mode of the charger, and the charging progress value displayed by the charger, so as to trigger the dynamic optimization of the charger's charging data.

[0085] In the embodiments of this application, the system constructs a current charging process feature vector based on a three-dimensional space of "time-energy-thermodynamics". This vector not only marks the SOC (State of Charge), but also includes the time dimension of the charging time and estimated remaining time, the energy dimension of the charged capacity, charged energy and current power, the thermodynamic dimension of the battery pack's highest temperature and temperature rise integral, and the chemical dimension of the estimated internal polarization of the battery. The system maps the feature vector to a predefined process state machine for semantic labeling, such as marking it as Phase_Init for the pre-charging and wake-up stage, Phase_CC_High for the high-current constant current fast charging stage, Phase_CC_Mid for the constant current mid-stage thermal equilibrium establishment stage, Phase_CV_Entry for the constant current to constant voltage transition stage, or Phase_CV_Tail for the trickle charging stage.

[0086] Based on the marked process state, the system retrieves the corresponding standard charging progress table from the cloud knowledge base or local storage. This progress table is not a simple voltage / current curve, but a sequence list containing discrete key control points, defining the standard values ​​of various parameters that should be present at a specific time point or SOC point. The system traverses the progress table with pointers, comparing the currently collected real-time data with each milestone node in the table. It performs time axis comparison to determine whether the current SOC has been reached within the predetermined time window, state axis comparison to determine whether the current temperature rise conforms to the standard thermal model of this stage, and energy axis comparison to determine whether the current internal resistance change rate conforms to aging characteristics.

[0087] The system calculates the Euclidean distance or weighted deviation between the actual parameters and the standard parameters of the schedule. If the deviation of a node exceeds a preset tolerance threshold, the node is judged as abnormal. Among all abnormal nodes, the system selects the node with the greatest impact on the final goal (such as charging speed and battery life) and marks it as a node to be optimized. Common node types include thermal equilibrium nodes with excessively rapid temperature rise and insufficient heat dissipation, lithium plating suppression nodes with excessive terminal current and risk of lithium plating, and high-current maintenance nodes that cannot maintain the theoretically high current due to grid or equipment limitations. In addition, the system generates a normalized optimization progress flag for the identified nodes to be optimized. If the node state is perfect, no optimization is needed. If there is a deviation but it is within a controllable range, it is marked as moderate. If the node deviates significantly from the standard, it is marked as urgently needing optimization. This flag determines the aggressiveness of the control framework in subsequent steps.

[0088] Specifically, the time has progressed to 40 minutes since charging began; the charger previously implemented current reduction protection due to high ambient temperature, and the temperature has now dropped, but the charging speed has slowed down, with the current SOC at 62%; the system reads real-time data, showing that 40 minutes of charging time has elapsed, the output current is 200A and is in a recovery state, and the battery temperature is 38°C; the system judges that the current is still high and has not yet entered the constant voltage stage, so it marks the current charging process as "late stage of constant current charging". The characteristics of this stage are that battery polarization begins to increase, the temperature rise inertia is large, and it is the key window period that determines the total charging time.

[0089] The system retrieves the standard charging schedule for "charger plus this type of lithium battery" under "summer high temperature conditions". While traversing the schedule, the system focuses on checking the expected parameters of the constant current / constant voltage transition point. According to the standard data, the standard SOC should reach 68% at the 40th minute, the standard current should be maintained at 240A, and the standard battery temperature should be controlled at 40°C. However, the actual data shows an SOC of 62%, lagging by 6%, a current of 200A, which is low, and a temperature of 38°C, which is slightly low but still acceptable. The system analysis reveals that the actual progress is about 4 to 5 minutes behind the standard schedule on the timeline.

[0090] The system identified the main cause of the lag as low current and identified the "high current maintenance node" in the schedule as a node to be optimized. The original task of this node was to maintain a high current input as much as possible under the premise of safety. The system calculated the difference ratio between the current current and the standard current and combined it with the time lag to determine that the execution efficiency of this node was poor. Finally, the optimization progress mark was generated as 0.65, which means that the node is currently only performing at 65% of its capacity, with 35% room for improvement, and is in the "urgently need optimization" range.

[0091] Furthermore, the continuous charging time of the charger is collected, and the charging control framework is determined based on the continuous charging time of the charger and the optimization progress of the node to be optimized. At the same time, the charging target corresponding to the charger is marked, and the corresponding charging control system is constructed based on the charging target and the charging control framework. This takes into account the overall consideration of the continuous charging time of the charger and the optimization progress of the node to be optimized, ensuring the accuracy of the charging control framework.

[0092] At this point, the system performs a dimensional analysis of the continuous charging time, including obtaining the cumulative running time from the start of charging to the present, estimating the remaining charging time required to reach the target SOC in real time by combining the current power curve and battery model, and calculating the difference between the remaining time and the user's expected deadline. In the time- and optimization progress-based framework decision-making, the system performs a two-dimensional mapping between the continuous charging time, representing time pressure, and the progress of the nodes to be optimized, representing the current performance bottleneck, to determine the charging control framework. For example, in a scenario where timeout is imminent and progress is poor, it is determined to be an "extreme catch-up framework," which allows temporary breakthroughs of some non-critical safety thresholds in exchange for speed. In a scenario where time is ample but progress is poor, it is determined to be an "adaptive repair framework," which focuses on dynamically adjusting parameters. In a scenario where time is ample and progress is good, it is determined to be an "optimal energy efficiency framework," which prioritizes improving charging efficiency.

[0093] The system performs multi-objective decoupling and labeling, going beyond a single, vague "fully charged" objective. Instead, it labels specific control vector objectives, including a time objective of reaching the target SOC before a specific time, a health objective of minimizing battery aging during this charge, and a safety objective of ensuring that all physical quantities do not trigger hard protection. Based on the defined control framework, the system dynamically rearranges these objectives. For example, under the "extreme catch-up framework," the time objective is labeled with the highest priority, while the health objective's weight is downgraded. In contrast, under the "optimal energy efficiency framework," the safety and health objectives have the highest priority, while the time objective's weight is downgraded.

[0094] The system defines an architecture that includes algorithm parameters for the perception layer, control logic for the decision layer, and constraints for the execution layer. During dynamic configuration generation, the system loads the corresponding PID parameter set, Kalman filter gain coefficient, and feedforward compensation coefficient according to the framework, and loads fuzzy control rules that match the target. At the same time, it reshapes the constraint boundaries, setting the dynamic boundary envelopes of each physical quantity. For example, under the "limit catch-up framework," the maximum allowable temperature is temporarily increased or the current change rate limit is relaxed. After the system is built, it performs a rapid logic self-check to ensure that there are no conflicts within the newly generated charging control system, and then activates the system to take over the charging control.

[0095] Specifically, the timeline advances to 40 minutes after charging begins; the charger is in the later stages of constant current operation, and the optimization progress marker for the "high current maintenance node" is 0.65, indicating insufficient performance and lagging behind the plan; the system collects the continuous charging time as 40 minutes and identifies the user-set "estimated departure time" as 75 minutes, meaning the remaining time window is 35 minutes; based on the current optimization progress, if the status quo is maintained, the system predicts it will take 42 minutes to fully charge, and the total estimated time will exceed the user's expectations; given the critical state of "time pressure and performance lag," the system determines the charging control framework to be a "dynamic acceleration framework with temperature rise constraints," which aims to utilize the remaining thermal safety margin to make a small power boost to address the time pressure.

[0096] Based on the above framework, the system re-marks the control objectives; the primary objective is set as "ensuring 100% SOC is reached within 75 minutes"; secondary objectives include "maintaining a thermal safety margin of at least 2°C to prevent triggering over-temperature protection" and "maintaining cell consistency, allowing for slight relaxation". This means that in order to meet the primary time objective, the system allows for the sacrifice of some cell consistency objectives, but must strictly adhere to the bottom line of thermal safety margin.

[0097] The system activates the "Dynamic Acceleration Control System" and performs parameter reconfiguration. The system increases the proportional gain of the current loop PI controller by 15% to improve the response speed to current commands and introduces a feedforward thermal model to predict the temperature rise after the current increase in real time and adjust the cooling strategy in advance. In terms of boundary reshaping, the system slightly adjusts the trigger voltage threshold of "constant current to constant voltage" by 0.5V from the standard value, thereby extending the time window for large current injection, and temporarily relaxes the "power derating trigger point" from the standard temperature of 50℃ to 52℃. The new control system starts running, guiding the charger to operate with a tighter thermal boundary and faster response speed in the next 35 minutes to ensure that it meets the user's charging time target.

[0098] Therefore, based on the identification of this charging control system, multiple control measures are determined, and corresponding dynamic optimization content is determined according to multiple control measures, the current charging mode of the charger, and the charging progress value displayed by the charger. This triggers the dynamic optimization of the charger's charging data, taking into account the overall consideration of multiple control measures, the current charging mode of the charger, and the charging progress value displayed by the charger, ensuring the accuracy of the corresponding dynamic optimization content. At the same time, a data control task is introduced to trigger the autonomous adjustment of data deviation, realizing the overall consideration of the optimization progress of the charging process, the continuous charging time of the charger, and the corresponding charging target, thereby improving the accuracy of the charging control system.

[0099] At this point, the system analyzes the logical structure of the currently activated charging control system, which includes power transmission, thermal management, chemical protection, and power quality dimensions. For each dimension, the system matches specific control measures, which are the concrete manifestations of the algorithm. Common types of measures include reference trajectory reshaping measures to modify the voltage / current target curve, closed-loop parameter correction measures to adjust PID controller parameters online, modulation strategy optimization measures to change the PWM carrier frequency or phase shift angle to suppress losses, pulse excitation measures to eliminate battery concentration polarization, and auxiliary energy management measures to dynamically adjust the heat dissipation airflow or liquid cooling flow rate.

[0100] The system analyzes the constraint mechanisms of the current charging mode. The mode defines the "domain" and "boundary" of the measures. For example, in the "constant current mode," the optimization mainly focuses on fine-tuning the current reference value, while in the "constant voltage mode," it shifts to controlling the slope of the current decrease rate. At the same time, different modes also have different definitions for safety constraints such as the ripple current threshold. In addition, the system performs nonlinear mapping of the charging progress value. The charging process is highly nonlinear. In the low SOC range, the focus is on the ability to inject large currents; in the medium SOC range, the focus is on heat loss suppression; and in the high SOC range, the focus is on polarization elimination. Based on the above constraints and characteristics, the system generates a dynamic optimization parameter set for each control measure and calculates specific values, including the adjustment amount of the current reference value, the PID gain adjustment coefficient, or the depolarization pulse width.

[0101] The system encapsulates dynamic optimization content into real-time control messages, which are then sent to the underlying controller via the internal communication bus. The triggering process employs seamless integration technology to ensure that changes in control parameters do not cause transient jumps in output voltage or current. After the physical-level response occurs, the system updates the data, including direct data changes in output voltage and current values, as well as real-time updates of state data such as estimated SOC, internal resistance, and polarization voltage values ​​within the battery. By continuously monitoring the optimized data, such as whether the temperature rise slope decreases or the charging speed increases, the system confirms that the optimization strategy has taken effect.

[0102] Specifically, the time has progressed to 42 minutes after the start of charging; S152 has constructed a "dynamic acceleration control system with temperature rise constraints"; the charger is currently in "intelligent temperature control mode", and the charging progress value shows SOC 65%, which is in the late stage of constant current CC and is about to enter the constant voltage CV stage; the system analyzes the "dynamic acceleration control system" and determines that it is necessary to tap the charging potential while ensuring thermal safety. Therefore, the following key control measures are locked: segmented current reference value increase to increase the target current within the thermal allowable range, dynamic fine adjustment of CV cutoff voltage to delay the constant voltage point to extend the CC stage time, and switching frequency thermal optimization to dynamically adjust the PWM frequency to reduce IGBT switching losses.

[0103] In the mode constraint analysis, the "intelligent temperature control mode" specifies a temperature upper limit of 55℃, and the current temperature is 45℃, leaving a safety margin of 10℃. In the progress value mapping analysis, SOC 65% is in the critical critical region of CC / CV switching, and battery polarization begins to increase. Simply increasing the current will lead to a sharp drop in efficiency and a surge in heat generation. Based on this, the system calculates and generates optimization content: For the current increase measure, considering that the progress is close to the CV stage, it is not advisable to increase it significantly. The calculation result is to fine-tune the current reference value from 210A to 215A. For the CV point fine-tuning measure, in order to save time, the trigger voltage threshold for constant current to constant voltage is increased from 400.0V to 401.5V using the 10℃ temperature rise margin. For the frequency optimization measure, based on the current and temperature, the optimal switching frequency is calculated, and the PWM frequency is reduced from 20kHz to 18kHz, sacrificing a little ripple performance to significantly reduce switching losses.

[0104] The system sends an instruction set to the charger's DSP core, setting the current reference value to 215A and using a slow ramp to prevent sudden current changes, setting the CV trigger voltage to 401.5V, and setting the switching frequency to 18kHz. After the optimization takes effect, the charger's output voltage slowly climbs to 401.5V before current limiting begins, extending the high-current charging time by about 3 minutes. At the same time, due to the frequency dropping to 18kHz, although the current increases by 5A, the temperature rise rate remains stable at 46℃. Through the above dynamic optimization, the charger successfully squeezes out critical charging time while ensuring safety margins.

[0105] Please see Figure 7 The charging control system of the charger based on big data includes: The big data module 21 is used to detect data of the charger in the charging state and collect multiple charging data. Based on the multiple charging data, the corresponding charging progress nodes and the charger temperature data, multiple sets of big data combinations are determined. The charging twin model module 22 is used to determine the corresponding charging state coefficient based on the data identification of each big data combination, and to determine the charging twin model of the charger based on each charging state coefficient, the charging progress value displayed by the charger and the corresponding charging scenario. The data deviation module 23 is used to identify multiple abnormal charging data based on the identification of the charging twin model, determine the data deviation combination based on each abnormal charging data, the corresponding charging stage and the charging efficiency value of the charger, and determine the corresponding data deviation amount based on the detection of the data deviation combination. The data control module 24 is used to collect the charger's autonomous control system, determine the data control task based on the autonomous control system, the data deviation, and the charger's current charging mode, and trigger autonomous adjustment of the data deviation in the data control task to dynamically optimize the charger's charging process. The charging control system module 25 is used to determine the corresponding optimization progress based on the detection of the charging process of the charger, determine the corresponding charging control system according to the optimization progress of the charging process, the continuous charging time of the charger, and the corresponding charging target, and trigger the dynamic optimization of the charging data of the charger.

[0106] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A charging control method for a charger based on big data, characterized in that, include: Data detection is performed on the charger in the charging state, and multiple charging data are collected. Multiple sets of big data combinations are determined based on the multiple charging data, the corresponding charging progress nodes, and the charger temperature data. Based on the data identification of each big data combination, the corresponding charging state coefficient is determined, and the charging twin model of the charger is determined according to each charging state coefficient, the charging progress value displayed by the charger and the corresponding charging scenario. The charging twin model includes a circuit simulation component, a thermodynamic simulation sub-component, and an aging prediction sub-component. Based on the identification of the charging twin model, multiple abnormal charging data are identified. Based on each abnormal charging data, the corresponding charging stage and the charging efficiency value of the charger, a data deviation combination is determined. Based on the detection of the data deviation combination, the corresponding data deviation amount is determined. The system collects the charger's autonomous control system, determines the data control task based on the autonomous control system, the data deviation, and the charger's current charging mode, and triggers autonomous adjustment of the data deviation in the data control task to dynamically optimize the charger's charging process. The corresponding optimization progress is determined based on the detection of the charging process of the charger. The corresponding charging control system is determined according to the optimization progress of the charging process, the continuous charging time of the charger, and the corresponding charging target, and the dynamic optimization of the charging data of the charger is triggered. The charging control system includes power transmission dimension, thermal management dimension, chemical protection dimension, and power quality dimension.

2. The charging control method for a charger based on big data according to claim 1, characterized in that, The process involves data detection of the charger during charging, collecting multiple charging data points, and determining multiple sets of large data combinations based on these data points, corresponding charging progress nodes, and charger temperature data. These combinations include: Collect multiple working data from the charger, determine the charger's working status based on the multiple working data and the charger's feedback signal, and monitor the charger's working status in real time. At this time, perform data detection on the charger and mark multiple charging data during the detection process. The temperature data of the thermometer is determined based on the temperature detection of the charger, and the dimensional range of the charger at various locations is marked. At the same time, the charging progress nodes of the charger are collected, and multiple sets of big data combinations are determined based on multiple charging data, corresponding charging progress nodes and the temperature data of the charger.

3. The charging control method for a charger based on big data according to claim 1, characterized in that, The process involves identifying corresponding charging state coefficients based on data combinations of various big data sets, and determining the charging twin model of the charger based on each charging state coefficient, the charging progress value displayed by the charger, and the corresponding charging scenario. This includes: In multiple sets of big data combinations, data identification is performed on each set of big data combinations, and multiple key charging data of the charger are determined during the identification process. Based on multiple key charging data, corresponding charging behavior and power consumption efficiency coefficient of the charger, the corresponding charging state coefficient is determined. The charging progress value displayed by the charger is determined based on the charger's progress recognition, and multiple scene features are determined based on the charger's surrounding detection. The charging model framework is determined based on the multiple scene features and various charging state coefficients. The charging twin model of the charger is determined based on the charging model framework and the charging progress value displayed by the charger.

4. The charging control method for a charger based on big data according to claim 1, characterized in that, The process involves identifying multiple abnormal charging data points based on the charging twin model, determining data deviation combinations based on each abnormal charging data point, the corresponding charging stage, and the charger's charging efficiency value, and determining the corresponding data deviation amount based on the detection of these data deviation combinations. This includes: The charging twin model is dynamically identified, and multiple abnormal contents are identified during the identification process. Based on the identification of each abnormal content, the corresponding abnormal charging data is determined to collect multiple abnormal charging data. The corresponding charging stage is determined by tracing each abnormal charging data. Collect the charging efficiency value of the charger, determine the first level of deviation based on the charging efficiency value of the charger and various abnormal charging data, determine the second level of deviation based on the charging efficiency value of the charger and the charging stage corresponding to each abnormal charging data, and determine the corresponding data deviation combination based on the first level of deviation and the second level of deviation.

5. The charging control method for a charger based on big data according to claim 4, characterized in that, The process of identifying multiple abnormal charging data points based on the charging twin model, determining data deviation combinations based on each abnormal charging data point, the corresponding charging stage, and the charger's charging efficiency value, and determining the corresponding data deviation amount based on the detection of the data deviation combination, further includes: Based on the detection of this data deviation combination, multiple data deviation items are identified. Based on the identification of each data deviation item, multiple deviation ranges are determined. Based on the multiple deviation ranges, the current charging mode of the charger, and the charging progress value displayed by the charger, the corresponding data deviation amount is determined, and the corresponding charging impact content is marked.

6. The charging control method for a charger based on big data according to claim 1, characterized in that, The autonomous control system of the charger, based on the autonomous control system, data deviation, and the charger's current charging mode, determines a data control task, and triggers autonomous adjustment of the data deviation within this data control task to dynamically optimize the charger's charging process, including: The system monitors the charger's database in real time, determines the charger's autonomous control system based on the database and the charger's charging scenario, and identifies multiple autonomous control items based on the identification of the autonomous control system.

7. The charging control method for a charger based on big data according to claim 6, characterized in that, The autonomous control system for the charger, based on the autonomous control system, data deviation, and the charger's current charging mode, determines a data control task and triggers autonomous adjustment of the data deviation within this data control task to dynamically optimize the charger's charging process. It also includes: The first level of data control content is determined based on multiple autonomous control items and data deviation. The second level of data control content is determined based on multiple autonomous control items and the current charging mode of the charger. The data control task is determined based on the first level of data control content and the second level of data control content. In the data control task, multiple data control nodes are identified based on the identification of the data control task. The corresponding autonomous control content is determined according to the multiple data control nodes, the charging progress value displayed by the charger, and the data deviation. Autonomous control of the data deviation is triggered, and the charging process of the charger is optimized based on the autonomous control of the data deviation.

8. The charging control method for a charger based on big data according to claim 1, characterized in that, The process involves determining the corresponding optimization progress based on the detection of the charging process of the charger, determining the corresponding charging control system based on the optimization progress of the charging process, the continuous charging time of the charger, and the corresponding charging target, and triggering dynamic optimization of the charger's charging data, including: The charging progress of the charger is marked, and the corresponding charging progress table is determined based on the identification of the charging progress of the charger. The corresponding node to be optimized is determined by traversing the charging progress table, and the optimization progress of the node to be optimized is marked.

9. The charging control method for a charger based on big data according to claim 8, characterized in that, The process of determining the corresponding optimization progress based on the detection of the charging process of the charger, determining the corresponding charging control system based on the optimization progress of the charging process, the continuous charging time of the charger, and the corresponding charging target, and triggering dynamic optimization of the charger's charging data, also includes: Collect the continuous charging time of the charger, determine the charging control framework based on the continuous charging time of the charger and the optimization progress of the node to be optimized, and mark the charging target corresponding to the charger. Construct the corresponding charging control system based on the charging target and the charging control framework. Based on the identification of the charging control system, multiple control measures are determined, and corresponding dynamic optimization content is determined according to the multiple control measures, the current charging mode of the charger, and the charging progress value displayed by the charger, so as to trigger the dynamic optimization of the charger's charging data.

10. A charging control system for a charger based on big data, characterized in that, The charging control system of the charger based on big data is applied to the charging control method of the charger based on big data as described in any one of claims 1-9.

Images