Cooperative control method and system for front stage and rear stage of direct-current fast charging module

Abstract

The invention relates to the technical field of charging management, in particular to a front-and-back stage cooperative control method and system for a direct-current fast charging module. Monitoring working parameters and environmental parameters of the front-stage converter and the rear-stage converter in real time, and constructing a multi-dimensional feature matrix in combination with a pre-acquired historical charging mode and user behavior data; performing multi-scale prediction on the charging energy demand by adopting the trained time sequence prediction model, and generating an energy demand curve comprising a short-term prediction window and a long-term prediction window; and dynamically adjusting modulation parameters of the front-stage converter according to the energy demand curve of the short-term prediction window, and adaptively switching a hybrid modulation mode of the rear-stage converter based on the energy demand curve of the long-term prediction window. According to the invention, working parameters and environmental parameters of the front-stage converter and the rear-stage converter can be monitored in real time, and comprehensive analysis is carried out in combination with a historical charging mode and user behavior data, so that more accurate and efficient charging control is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of charging management, and particularly to a method and system for coordinated control of the front and rear stages of a DC fast charging module. Background Art

[0002] With the popularization of electric vehicles and various portable electronic devices, DC fast charging technology has become one of the key technologies to meet the demand for fast charging. However, existing DC fast charging modules still face many challenges in terms of energy efficiency, charging speed, and system stability. In a DC fast charging system, the front-stage converter is mainly responsible for converting the input power supply into a DC voltage suitable for post-stage processing, and the post-stage converter further adjusts the voltage and current to meet the specific requirements of battery charging. Currently, the front-stage converter and the post-stage converter are usually independently controlled, lacking a coordination mechanism, resulting in the entire charging system being difficult to achieve optimal performance. Traditional charging control methods often rely on fixed parameter settings or simple feedback adjustment mechanisms, making it difficult to adapt to complex and variable charging environments and user behavior patterns, thus leading to problems such as low charging efficiency, device overheating, and electromagnetic interference. Summary of the Invention

[0003] Aiming at the defects of the existing technology, the present invention provides a method and system for coordinated control of the front and rear stages of a DC fast charging module, which can real-time monitor the working parameters and environmental parameters of the front-stage converter and the post-stage converter, and perform comprehensive analysis in combination with historical charging patterns and user behavior data to achieve more accurate and efficient charging control.

[0004] To solve the above technical problems, the technical solution adopted by the present invention is: A method for coordinated control of the front and rear stages of a DC fast charging module, comprising: S01. Real-time monitor the working parameters and environmental parameters of the front-stage converter and the post-stage converter, and construct a multi-dimensional feature matrix in combination with the previously obtained historical charging patterns and user behavior data; S02. Based on the multi-dimensional feature matrix, use a trained time series prediction model to perform multi-scale prediction on the charging energy demand, and generate an energy demand curve including a short-term prediction window and a long-term prediction window; S03. Dynamically adjust the modulation parameters of the front-stage converter according to the energy demand curve of the short-term prediction window, and adaptively switch the hybrid modulation mode of the post-stage converter based on the energy demand curve of the long-term prediction window.

[0005] Further, the construction of the multi-dimensional feature matrix includes: Generate a humidity influence factor based on the humidity parameter in the environmental parameters, and calculate the thermal resistance coefficient in combination with the real-time device temperature; Extract the charging period preference, single charging duration, and demand power from the user behavior data, and generate a weighted user behavior feature vector; Perform frequency-domain decomposition on the input voltage fluctuation of the pre-stage converter, extract the harmonic energy ratio of the selected frequency band, and use it as the electromagnetic interference feature; Match the similarity between the user behavior feature vector and the historical charging pattern to generate a scenario matching coefficient; Fuse the humidity influence factor, the thermal resistance coefficient, the electromagnetic interference feature, and the scenario matching coefficient through a preset spatio-temporal alignment engine, and fuse them with the real-time monitored input voltage and real-time device temperature to generate a multi-dimensional feature matrix.

[0006] Further, the time series prediction model uses a two-layer recurrent neural network, where: The first layer is an LSTM network, which is used to process short-term time series data and output the power change rate and load fluctuation period; The second layer is a GRU network, which is used to process long-term trend data and predict the steady-state power range and load peak period; Dynamically fuse the environmental temperature, humidity parameters, and real-time device temperature through an attention mechanism to generate a multi-scale energy demand curve.

[0007] Further, the multi-scale prediction process of the time series prediction model includes: Perform time series alignment and normalization processing on the multi-dimensional feature matrix to generate a standardized input sequence; Input the standardized input sequence into the two-layer recurrent neural network, extract short-term fluctuation features through the LSTM network, and fit long-term trend features through the GRU network; Use the attention mechanism to dynamically allocate the weights of the environmental temperature, humidity parameters, and real-time device temperature to generate an environment-device coupling factor; Fuse the short-term fluctuation features, long-term trend features, and environment-device coupling factor, and output a multi-scale energy demand curve.

[0008] Further, the modulation parameters for dynamically adjusting the pre-stage converter include: Analyze the energy demand curve of the short-term prediction window, and extract the power change rate gradient and load fluctuation frequency; According to the power change rate gradient, calculate the duty cycle adjustment amount of the pre-stage converter through a fuzzy control algorithm; Based on the load fluctuation frequency, dynamically adjust the switching frequency of the pre-stage converter so that it has a negative correlation with the fluctuation frequency; Input the adjusted duty cycle and switching frequency into the pulse width modulation module of the pre-stage converter to generate a real-time drive signal.

[0009] Further, the rule base of the fuzzy control algorithm includes: When the power change rate gradient exceeds a preset first threshold, increase the duty cycle; When the load fluctuation frequency is in the high-frequency band, reduce the switching frequency.

[0010] Further, the hybrid modulation mode of the adaptive switched post-stage converter includes: Analyze the energy demand curve of the long-term prediction window to obtain the upper and lower limits of the steady-state power interval and the load peak period; When the predicted steady-state power interval exceeds a preset second threshold, switch to the pulse width modulation mode; When the duration of the load peak period exceeds a set value, switch to the pulse frequency modulation mode; Based on the dynamic relationship between the thermal resistance coefficient and the real-time device temperature, adjust the switching hysteresis width between the pulse width modulation mode and the pulse frequency modulation mode.

[0011] The present invention also discloses a control system for the front and rear stages of a DC fast charging module, including: A multi-dimensional matrix construction module for real-time monitoring of the operating parameters and environmental parameters of the front-stage converter and the post-stage converter, and constructing a multi-dimensional feature matrix in combination with the previously obtained historical charging modes and user behavior data; A multi-scale prediction module for multi-scale prediction of the charging energy demand based on the multi-dimensional feature matrix by using a trained time series prediction model, and generating an energy demand curve including a short-term prediction window and a long-term prediction window; A dynamic adjustment module for dynamically adjusting the modulation parameters of the front-stage converter according to the energy demand curve of the short-term prediction window, and adaptively switching the hybrid modulation mode of the post-stage converter based on the energy demand curve of the long-term prediction window.

[0012] Further, the dynamic adjustment of the modulation parameters of the front-stage converter includes: Analyze the energy demand curve of the short-term prediction window to extract the power change rate gradient and the load fluctuation frequency; According to the power change rate gradient, calculate the duty cycle adjustment amount of the front-stage converter through a fuzzy control algorithm; Based on the load fluctuation frequency, dynamically adjust the switching frequency of the front-stage converter so that it has a negative correlation with the fluctuation frequency; Input the adjusted duty cycle and switching frequency into the pulse width modulation module of the front-stage converter to generate a real-time drive signal.

[0013] Further, the hybrid modulation mode of the adaptive switched post-stage converter includes: Analyze the energy demand curve of the long-term prediction window to obtain the upper and lower limits of the steady-state power range and the peak load period; When the predicted steady-state power range exceeds a preset second threshold, switch to the pulse width modulation mode; When the duration of the peak load period exceeds a set value, switch to the pulse frequency modulation mode; Based on the dynamic relationship between the thermal resistance coefficient and the real-time device temperature, adjust the switching hysteresis width between the pulse width modulation mode and the pulse frequency modulation mode.

[0014] Advantages of the present invention: By constructing a multi-dimensional feature matrix, the present invention can comprehensively reflect the real-time state of the charging system and the user behavior characteristics, providing basic data for subsequent prediction and control. Using the trained time series prediction model to perform multi-scale prediction on the charging energy demand, an energy demand curve including short-term and long-term prediction windows can be generated, providing a basis for the dynamic adjustment of the front-stage converter and the rear-stage converter.

[0015] In terms of dynamic adjustment, according to the energy demand curve of the short-term prediction window, the modulation parameters of the front-stage converter can be dynamically adjusted to adapt to the changes in the power change rate and the load fluctuation frequency, thereby improving the charging efficiency and system stability. At the same time, based on the energy demand curve of the long-term prediction window, the hybrid modulation mode of the rear-stage converter can be adaptively switched to meet the requirements of different charging stages and further optimize the charging performance. Brief Description of the Drawings

[0016] Figure 1 It is a flowchart of a front-stage and rear-stage collaborative control method for a DC fast charging module provided by the present invention; Figure 2 It is a module structure diagram of a front-stage and rear-stage collaborative control system for a DC fast charging module provided by the present invention. Detailed Embodiments

[0017] The following further describes the present invention with reference to the drawings and specific embodiments.

[0018] The following further elaborates on the present invention with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only for explaining the present invention and not for limiting the present invention. Additionally, it should be noted that for the sake of description, only the parts related to the present invention are shown in the drawings rather than all the structures. Furthermore, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0019] It should also be noted that, for ease of description, only parts related to the present invention rather than all contents are shown in the drawings. Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but there can also be additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, and so on.

[0020] Embodiment 1 The present invention provides a method for coordinated control of the front and rear stages of a DC fast charging module. Referring to Figure 1 as shown, the method includes the following steps: S01: Real-time monitor the operating parameters and environmental parameters of the front-stage converter and the rear-stage converter, and construct a multi-dimensional feature matrix by combining the previously obtained historical charging modes and user behavior data.

[0021] In step S01, the construction of the multi-dimensional feature matrix is achieved through the following process: First, according to the environmental humidity parameter and the surface temperature of the device monitored in real time, a humidity influence factor is generated by means of weighted combination, where the weight coefficient is determined through historical data analysis and is used to quantify the comprehensive influence of humidity and temperature on the device performance. At the same time, by collecting the junction temperature and environmental temperature data of the power device in real time and combining the power loss calculation results, the thermal resistance coefficient is dynamically calculated to characterize the heat dissipation efficiency of the device.

[0022] For the user behavior data, extract the charging time period preference, single charging duration, and required power from the historical records, count the charging frequency, average duration, and power fluctuation conditions in each time period, and assign weights through the entropy method to generate a normalized user behavior feature vector. Exemplarily, higher weights are given to the features of high-frequency charging time periods, stable charging durations, and high power requirements.

[0023] The input voltage fluctuation signal of the front-stage converter is obtained through a high-speed sampling module and undergoes frequency-domain decomposition processing to extract the harmonic energy ratio in the frequency band of 2 kHz to 10 kHz as the electromagnetic interference feature. The specific method is to perform a fast Fourier transform on the voltage signal and screen the energy ratio of the target frequency band to quantify the electromagnetic interference intensity.

[0024] The user behavior feature vector is further matched with typical scenarios in the historical charging mode library. Exemplarily, the matching degree between the current behavior and historical scenarios such as "fast charging mode" and "balanced mode" is calculated through the cosine similarity algorithm, and the mode with the highest similarity is selected to generate a scenario matching coefficient for identifying the current charging scenario type.

[0025] Finally, the above features are fused by a spatio-temporal alignment engine: in the time dimension, an interpolation algorithm is used to align the sampling timestamps of different sensors; in the space dimension, the distribution difference of environmental parameters is corrected by the inverse distance weighted method. The humidity influence factor, thermal resistance coefficient, electromagnetic interference characteristics, scenario matching coefficient, and real-time voltage and temperature data are integrated in a time series to form a multi-dimensional feature matrix.

[0026] Exemplarily, within a 1-minute time window, the matrix contains 60 groups of data in 6 dimensions, providing a structured input for the subsequent prediction model.

[0027] S02: Based on the multi-dimensional feature matrix, a trained time series prediction model is used to perform multi-scale prediction on the charging energy demand, generating an energy demand curve including a short-term prediction window and a long-term prediction window.

[0028] In step S02, the multi-dimensional feature matrix is input into a pre-trained time series prediction model for multi-scale energy demand prediction.

[0029] First, the model performs standardization processing on the input multi-dimensional feature data, ensures the consistency of the data timestamps of each dimension through time series alignment, and uses the min-max normalization method to scale parameters with different dimensions to a unified interval, eliminating the interference of magnitude differences on model training. Exemplarily, the voltage parameter may be normalized to the range of 0-1, while the temperature parameter is scaled proportionally according to the historical maximum and minimum values.

[0030] The time series prediction model adopts a two-layer recurrent neural network structure, where the first layer is a long short-term memory network (LSTM) for capturing dynamic changes in short-term time series. The LSTM network receives the standardized input sequence, takes the input voltage, current instantaneous fluctuation, and power change rate of the previous-stage converter as the main inputs, stores and updates the short-term state information through memory units, and outputs the predicted value of the power change rate within the next 30 seconds and the load fluctuation period characteristics. Exemplarily, in the scenario of a sudden increase in charging load, the LSTM can identify a short-term trend of a 20% increase in power within 5 seconds and mark the fluctuation period as a high-frequency mode.

[0031] The second layer is a gated recurrent unit (GRU) network, which focuses on analyzing long-term trend data. The GRU network takes the user behavior feature vector, the scene matching coefficient, and the historical charging pattern as inputs, combines the load data of the past 24 hours, and predicts the steady-state power interval range and the time period when the load peak appears in the next 10 minutes to 1 hour. Exemplarily, the GRU can predict that within the next 30 minutes, the system will enter a steady-state power interval of 50kW - 55kW, and a load peak lasting 5 minutes will occur at the 15th minute.

[0032] The model dynamically fuses the influence of environmental parameters and real-time device states through an attention mechanism. Specifically, the attention layer automatically assigns weights to different parameters according to the current humidity influence factor, the thermal resistance coefficient, and the device temperature data. Exemplarily, when the device temperature exceeds 60°C, the attention mechanism will increase the weight of the thermal resistance coefficient to strengthen the correction effect of the heat dissipation efficiency on the long-term power trend, while reducing the weight of the humidity parameter. The finally generated environment-device coupling factor is used to adjust the prediction result, making the energy demand curve more conform to the actual working conditions.

[0033] In the feature fusion stage, the short-term fluctuation features (such as the power change rate gradient) output by the LSTM, the long-term trend features (such as the steady-state power interval) output by the GRU, and the environment-device coupling factor are input into the fully connected layer for integration. The fully connected layer generates a multi-scale energy demand curve through weighted superposition. The curve in the short-term prediction window (such as 30 seconds) reflects the details of the instantaneous load fluctuation, and the curve in the long-term prediction window (such as 1 hour) shows the overall power change trend. Exemplarily, the model can output a short-term curve with dense fluctuations and a long-term curve with a smooth rise, respectively guiding the front-end and back-end control strategies.

[0034] During the model training process, historical charging data is used as the training set, including the working parameters of the front-end converter, the user charging behavior records, and the corresponding actual energy demand data under different environmental conditions. During training, the short-term prediction part uses second-level data as the supervision signal, and the long-term prediction part uses minute-level aggregated data as the supervision signal. The prediction accuracy of both is balanced through a joint loss function. Exemplarily, when training the model, it is necessary to simultaneously optimize the mean square error of the 30-second window and the mean absolute percentage error of the 1-hour window to ensure the coordination of multi-scale prediction.

[0035] S03: Dynamically adjust the modulation parameters of the front-end converter according to the energy demand curve of the short-term prediction window, and adaptively switch the hybrid modulation mode of the back-end converter based on the energy demand curve of the long-term prediction window.

[0036] In step S03, fine-grained collaborative control of the front-end converter and the back-end converter is performed based on the multi-scale prediction results.

[0037] For the dynamic regulation of the pre-stage converter, first extract the power values at second intervals from the energy demand curve of the short-term prediction window, and calculate the power difference between adjacent time points as the change rate gradient.

[0038] Exemplarily, when it is detected that the power linearly increases from 50 kW to 70 kW within the next 5 seconds, the gradient is calculated as 4 kW / s. This gradient value is input into the fuzzy controller, and the membership degree is determined according to the preset fuzzy level division (such as 0 - 2 kW / s being "zero", 2 - 5 kW / s being "medium", and exceeding 5 kW / s being "large positive"). Combining with the rule base (such as "a large positive gradient corresponds to a 10% increase in the duty cycle"), the adjustment amount is calculated in real time. If the gradient is 4 kW / s and the membership degrees are 70% "medium" and 30% "large positive", then the duty cycle is increased from 55% to 61.5%. At the same time, perform real-time FFT analysis on the load waveform to identify the main fluctuation frequency: if 80% of the energy is concentrated in the 1.5 kHz frequency band, then according to the preset negative correlation mapping table (1.5 kHz corresponds to a switching frequency of 90 kHz), the pre-stage switching frequency is reduced from 100 kHz to 90 kHz to reduce losses. The adjusted duty cycle and switching frequency parameters are input into the PWM module to generate a drive signal. Exemplarily, a pulse waveform with a duty cycle of 61.5% and a period of 11.1 μs is generated at a frequency of 90 kHz, and the output ripple is monitored in real time through a current sensor. When the ripple exceeds 5%, the duty cycle is automatically compensated by ±2%.

[0039] For the mode switching of the post-stage converter, analyze the steady-state power interval of the long-term prediction curve: if the average power in the next 10 minutes continuously exceeds 90% of the rated value (such as 90 kW) for 3 minutes, then switch to the PWM mode, fix the switching frequency at 50 kHz, and dynamically adjust the duty cycle according to the output voltage error (such as increasing by 2% when the target voltage is 800 V and the actual voltage is 780 V). When it is detected that there is a load peak (such as 110 kW) lasting for 2 minutes in the next 5 minutes, then switch to the PFM mode and adjust the frequency according to the load current (200 A corresponds to 30 kHz, 100 A corresponds to 20 kHz). During mode switching, adjust the hysteresis width according to the real-time thermal resistance coefficient and device temperature: when the temperature exceeds 65 °C and the thermal resistance coefficient increases, the hysteresis expands from ±2% to ±5% to avoid frequent switching caused by temperature fluctuations. The switching process adopts a ramp transition strategy. Exemplarily, the PWM duty cycle linearly decreases from 60% to 50% within 10 ms, and at the same time, the PFM frequency gradually increases from 30 kHz to 40 kHz to ensure a smooth transition of the output voltage. If the bus voltage fluctuation exceeds ±5% or the temperature suddenly rises by 10 °C / s within 2 seconds after switching, then immediately revert to the original mode and start fault diagnosis.

[0040] In coordinated control, the switching frequency change of the front stage is synchronized to the rear stage controller in real time (for example, when the front stage drops to 80 kHz, the PWM frequency of the rear stage is synchronized and reduced), and the current ripple data output by the rear stage is fed back to the front stage to trigger duty cycle compensation. The system is built with multiple protection mechanisms: when overvoltage (>850 V) or overcurrent (>120% of the rated value) occurs, the drive is immediately cut off and the discharge resistor is started; when the temperature ≥ 85 °C, it is forced to switch to the PFM mode and the maximum power is limited to 80%. After the temperature drops to 70 °C, the limit is gradually lifted. The parameter self-learning module optimizes the threshold setting monthly. Exemplarily, the PWM switching threshold is corrected from 90 kW to 88 kW based on historical data, and the fuzzy rule base is updated using reinforcement learning, adding "gradient change acceleration" as a control variable.

[0041] Exemplarily, when the short-term prediction shows that the power gradient reaches +4 kW / s within 5 seconds, the front stage duty cycle is increased from 55% to 62%, and the switching frequency is decreased from 100 kHz to 85 kHz, so that the bus voltage is stabilized at 805 V ± 5 V; the long-term predicted steady-state power of 95 kW triggers the rear stage PWM mode, and the duty cycle is maintained at 58% at a frequency of 50 kHz; when the load peak of 105 kW lasts for 3 minutes and the device temperature reaches 68 °C, it switches to the PFM mode, and the frequency of 25 kHz makes the temperature drop back to 63 °C. In an abnormal scenario, if the PFM mode causes the voltage to fluctuate to 830 V, it switches back to the PWM mode within 10 ms and limits the front stage duty cycle to 50% until the voltage recovers to 810 V ± 10 V.

[0042] Embodiment 2 The present invention also provides a front and rear stage coordinated control system for a DC fast charging module, which is used to execute a front and rear stage coordinated control method for a DC fast charging module. Refer to Figure 2 As shown, the system includes: A multi-dimensional matrix construction module 100, which is used to monitor the working parameters and environmental parameters of the front stage converter and the rear stage converter in real time, and construct a multi-dimensional feature matrix by combining the previously obtained historical charging modes and user behavior data.

[0043] A multi-scale prediction module 200, which is used to perform multi-scale prediction of the charging energy demand based on the multi-dimensional feature matrix, and generate an energy demand curve including a short-term prediction window and a long-term prediction window.

[0044] A dynamic adjustment module 300, which is used to dynamically adjust the modulation parameters of the front stage converter according to the energy demand curve of the short-term prediction window, and adaptively switch the hybrid modulation mode of the rear stage converter based on the energy demand curve of the long-term prediction window.

[0045] In this embodiment, the dynamic adjustment of the modulation parameters of the front stage converter includes: Analyze the energy demand curve of the short-term prediction window, and extract the power change rate gradient and the load fluctuation frequency; According to the power change rate gradient, calculate the duty cycle adjustment amount of the pre-stage converter through a fuzzy control algorithm; Based on the load fluctuation frequency, dynamically adjust the switching frequency of the pre-stage converter so that it has a negative correlation with the fluctuation frequency; Input the adjusted duty cycle and switching frequency into the pulse width modulation module of the pre-stage converter to generate a real-time drive signal.

[0046] The hybrid modulation mode for adaptively switching the post-stage converter includes: Analyze the energy demand curve of the long-term prediction window, and obtain the upper and lower limits of the steady-state power interval and the load peak period; When the predicted steady-state power interval exceeds a preset second threshold, switch to the pulse width modulation mode; When the duration of the load peak period exceeds a set value, switch to the pulse frequency modulation mode; Based on the dynamic relationship between the thermal resistance coefficient and the real-time device temperature, adjust the switching hysteresis width between the pulse width modulation mode and the pulse frequency modulation mode.

[0047] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0048] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and this program can be stored in a computer-readable storage medium. The storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disc memories, tape memories, or any other medium that can be used to carry or store data and is computer-readable.

[0049] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent in such process, method, commodity or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the element.

Claims

1. A method for collaborative control of the front and rear stages of a DC fast charging module, characterized in that: Including: S01. Real-time monitor the working parameters and environmental parameters of the pre-stage converter and the post-stage converter, and construct a multi-dimensional feature matrix by combining the pre-acquired historical charging modes and user behavior data; S02. Based on the multi-dimensional feature matrix, use a trained time series prediction model to perform multi-scale prediction on the charging energy demand, and generate an energy demand curve including a short-term prediction window and a long-term prediction window; S03. Dynamically adjust the modulation parameters of the pre-stage converter according to the energy demand curve of the short-term prediction window, and adaptively switch the hybrid modulation mode of the post-stage converter based on the energy demand curve of the long-term prediction window.

2. The method for collaborative control of the front and rear stages of the DC fast charging module according to claim 1, wherein: The construction of the multi-dimensional feature matrix includes: Generate a humidity influence factor based on the humidity parameter in the environmental parameters, and calculate the thermal resistance coefficient in combination with the real-time device temperature; Extract the charging time period preference, single charging duration and demand power in the user behavior data, and generate a weighted user behavior feature vector; Perform frequency domain decomposition on the input voltage fluctuation amount of the pre-stage converter, extract the harmonic energy ratio of the selected frequency band, and use it as the electromagnetic interference feature; Perform similarity matching between the user behavior feature vector and the historical charging mode to generate a scene matching coefficient; Fuse the humidity influence factor, the thermal resistance coefficient, the electromagnetic interference feature and the scene matching coefficient through a preset spatio-temporal alignment engine, and fuse the real-time monitored input voltage and real-time device temperature to generate a multi-dimensional feature matrix.

3. The front and rear stage collaborative control method of the DC fast charging module according to claim 2, characterized in that: The time series prediction model uses a two-layer recurrent neural network, where: The first layer is an LSTM network, which is used to process short-term time series data and output the power change rate and the load fluctuation period; The second layer is a GRU network, which is used to process long-term trend data and predict the steady-state power interval and the load peak period; Dynamically fuse the environmental temperature, humidity parameter and real-time device temperature through an attention mechanism to generate a multi-scale energy demand curve.

4. The method for coordinated control of the front and rear stages of the DC fast charging module according to claim 3, wherein: The multi-scale prediction process of the time series prediction model includes: Perform time series alignment and normalization processing on the multi-dimensional feature matrix to generate a standardized input sequence; Input the standardized input sequence into the two-layer recurrent neural network, extract short-term fluctuation features through the LSTM network, and fit long-term trend features through the GRU network; Use the attention mechanism to dynamically allocate the weights of the environmental temperature, humidity parameter and real-time device temperature to generate an environment-device coupling factor; Fuse the short-term fluctuation features, long-term trend features and environment-device coupling factor, and output a multi-scale energy demand curve.

5. The method for collaborative control of the front and rear stages of the DC fast charging module according to claim 1, wherein: The dynamic adjustment of the modulation parameters of the pre-stage converter includes: Analyze the energy demand curve of the short-term prediction window, and extract the power change rate gradient and the load fluctuation frequency; According to the power change rate gradient, calculate the duty cycle adjustment amount of the pre-stage converter through a fuzzy control algorithm; Based on the load fluctuation frequency, dynamically adjust the switching frequency of the pre-stage converter so that it has a negative correlation with the fluctuation frequency; Input the adjusted duty cycle and switching frequency into the pulse width modulation module of the pre-stage converter to generate a real-time drive signal.

6. The method for collaborative control of the front and rear stages of the DC fast charging module according to claim 5, characterized in that: The rule base of the fuzzy control algorithm includes: When the power change rate gradient exceeds a preset first threshold, increase the duty cycle; When the load fluctuation frequency is in the high-frequency band, reduce the switching frequency.

7. The method for coordinated control of the front and rear stages of the DC fast charging module according to claim 1, characterized in that: The hybrid modulation mode of the adaptive switching post-stage converter includes: Analyze the energy demand curve of the long-term prediction window to obtain the upper and lower limits of the steady-state power interval and the load peak period; When the predicted steady-state power interval exceeds a preset second threshold, switch to the pulse width modulation mode; When the duration of the load peak period exceeds a set value, switch to the pulse frequency modulation mode; Based on the dynamic relationship between the thermal resistance coefficient and the real-time device temperature, adjust the switching hysteresis width between the pulse width modulation mode and the pulse frequency modulation mode.

8. A front-stage and rear-stage collaborative control system for a DC fast charging module, characterized in that: It includes: A multi-dimensional matrix construction module for real-time monitoring of the operating parameters and environmental parameters of the pre-stage converter and the post-stage converter, and constructing a multi-dimensional feature matrix in combination with the pre-acquired historical charging mode and user behavior data; A multi-scale prediction module for multi-scale prediction of the charging energy demand based on the multi-dimensional feature matrix using a trained time series prediction model, and generating an energy demand curve including a short-term prediction window and a long-term prediction window; A dynamic adjustment module for dynamically adjusting the modulation parameters of the pre-stage converter according to the energy demand curve of the short-term prediction window, and adaptively switching the hybrid modulation mode of the post-stage converter based on the energy demand curve of the long-term prediction window.

9. The pre-stage and post-stage collaborative control system of the DC fast charging module according to claim 8, characterized in that: The dynamic adjustment of the modulation parameters of the pre-stage converter includes: Analyze the energy demand curve of the short-term prediction window, and extract the power change rate gradient and the load fluctuation frequency; According to the power change rate gradient, calculate the duty cycle adjustment amount of the pre-stage converter through a fuzzy control algorithm; Based on the load fluctuation frequency, dynamically adjust the switching frequency of the pre-stage converter so that it has a negative correlation with the fluctuation frequency; Input the adjusted duty cycle and switching frequency into the pulse width modulation module of the pre-stage converter to generate a real-time drive signal.

10. The pre-stage and post-stage collaborative control system of the DC fast charging module according to claim 8, characterized in that: The hybrid modulation mode of the adaptive switching post-stage converter includes: Analyze the energy demand curve of the long-term prediction window to obtain the upper and lower limits of the steady-state power interval and the load peak period; When the predicted steady-state power interval exceeds a preset second threshold, switch to the pulse width modulation mode; When the duration of the load peak period exceeds a set value, switch to the pulse frequency modulation mode; Based on the dynamic relationship between the thermal resistance coefficient and the real-time device temperature, adjust the switching hysteresis width between the pulse width modulation mode and the pulse frequency modulation mode.

Images