AC charging pile control system with voltage stabilization function

Abstract

The application discloses an alternating current charging pile control system with voltage stabilization function and relates to the technical field of smart grids.The system comprises a model calibration module, which is used for feeding back the adjusted output voltage to a digital twin virtual model for parameter calibration, starting a new control cycle according to the calibrated digital twin virtual model, and obtaining a preliminary impedance response spectrum of a charging loop; an impedance generation module, which is used for obtaining parameter configuration instructions of a high-frequency test current signal based on the preliminary impedance response spectrum, injecting the high-frequency test current signal, performing band-pass filtering and fast Fourier transform, and obtaining impedance characteristics; and a risk judgment module, which is used for extracting a multi-dimensional feature vector from the impedance characteristics, calculating a power grid risk index, and performing fault judgment to form a closed-loop optimization.The application generates multi-domain response spectra through a digital twin model, realizes advanced prediction of power grid disturbance waveforms, and enables the control process to have predictive compensation capability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart grid, in particular to an alternating current charging pile control system with voltage stabilization function. BACKGROUND

[0002] With the rapid increase of the popularity of electric vehicles, as a key component of infrastructure, the stability and intelligence level of the control system of alternating current charging piles directly affect the charging efficiency and power grid safety. The traditional charging pile control system mainly adopts a voltage stabilization scheme based on PID control, and realizes output voltage stabilization through voltage and current double closed loop control. In recent years, with the development of digital twin technology, some advanced control systems have begun to introduce power grid signal modeling and prediction mechanisms, such as using Fourier transform for harmonic analysis or using machine learning algorithms for short-term load prediction.

[0003] The prediction method according to the fixed parameter model in the prior art cannot effectively deal with nonlinear disturbances of the power grid, and simple feedback control is difficult to cope with complex working conditions of multivariable coupling. For example, when the charging pile faces load mutation and power grid voltage fluctuation at the same time, the traditional system needs to be adjusted through multiple iterations to stabilize the output, resulting in delayed voltage stabilization response and affecting the charging efficiency. In addition, the existing impedance detection method mostly uses offline measurement or single frequency excitation, lacking real-time sensing ability of full-band impedance characteristics, making it difficult for the system to identify potential faults such as increased contact resistance in time. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an alternating current charging pile control system with voltage stabilization function to solve the problems of response lag of traditional charging piles under dynamic disturbance and insufficient potential fault warning capability.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] The application provides an alternating current charging pile control system with voltage stabilization function, which comprises a signal acquisition module, a disturbance prediction module, a difference compensation module, an instruction output module, a model calibration module and an impedance generation module.

[0008] As a preferred scheme of the alternating current charging pile control system with voltage stabilization function, the generation of the multi-domain response spectrum of the power grid comprises the following specific steps:

[0009] The real-time power grid voltage signal and the output current signal are acquired and subjected to time domain denoising and filtering to obtain a time domain signal.

[0010] The time domain signal is converted into a time-frequency domain representation, and the time-frequency characteristics are extracted through fast Fourier transform.

[0011] A digital twin virtual model is constructed according to the time-frequency characteristics, and the structure and parameters of the digital twin virtual model are adjusted.

[0012] The adjusted digital twin virtual model is subjected to time-space feature analysis to obtain the multi-domain response spectrum of the power grid.

[0013] As a preferred scheme of the alternating current charging pile control system with voltage stabilization function, the construction of the dynamic graph structure based on the multi-domain response spectrum and the prediction of the future power grid disturbance waveform through the time-space attention mechanism comprise the following specific steps:

[0014] The dynamic graph structure is constructed based on the time-space feature dimensions in the multi-domain response spectrum.

[0015] The dynamic dependency relationship between nodes is extracted from the dynamic graph structure through a space-time attention mechanism to generate a space-time feature vector;

[0016] The space-time feature vector is subjected to a weighted integral operation to generate a prediction parameter of the future state of the power grid;

[0017] The prediction parameter of the future state of the power grid is subjected to time-domain simulation to predict a future power grid disturbance waveform.

[0018] As a preferred scheme of the alternating current charging pile control system with voltage stabilization function, the specific steps of generating the feedforward compensation amount are as follows,

[0019] The time-frequency features of the grid voltage and the future grid disturbance waveform are extracted from the real-time grid voltage signal and the future grid disturbance waveform;

[0020] The difference information between the grid voltage and the future grid disturbance waveform is obtained by performing difference analysis on the time-frequency features through a convolutional neural network, and the dynamic characteristics of the grid voltage change are captured;

[0021] Based on the dynamic characteristics of the grid, the feedforward compensation amount of the grid voltage is calculated in combination with the real-time grid voltage signal and the future grid disturbance waveform.

[0022] As a preferred scheme of the alternating current charging pile control system with voltage stabilization function, the specific steps of generating the error signal are as follows,

[0023] The actual output voltage and the voltage set value are collected and injected with a high-frequency test current signal, and the error signal frequency domain feature between the actual output voltage and the voltage set value is obtained by using fast Fourier transform;

[0024] The amplitude and phase of the error signal frequency domain feature are optimized and adjusted through frequency domain weighted filtering, and the optimized error signal is generated in combination with the dynamic characteristics of the grid.

[0025] As a preferred scheme of the alternating current charging pile control system with voltage stabilization function, the specific steps of inputting the feedforward compensation amount and the error signal into the multivariable predictive coordinated controller, outputting the composite robust control instruction, and converting it into the execution pulse width modulation signal to adjust the output voltage are as follows,

[0026] The feedforward compensation amount and the optimized error signal are input into the multivariable predictive coordinated controller to generate the original control instruction;

[0027] The original control instruction is dynamically corrected by the nonlinear disturbance observer to output the composite robust control instruction;

[0028] The composite robust control instruction is converted into a pulse width modulation signal with a specific duty cycle through a reverse symmetrical pulse width modulation strategy;

[0029] The switch tube in the power converter is driven by the pulse width modulation signal to adjust the output voltage.

[0030] As a preferred scheme of the AC charging pile control system with voltage stabilization function, the adjusted output voltage is fed back to the digital twin virtual model for parameter calibration, and a new control cycle is started according to the calibrated digital twin virtual model, and a preliminary impedance response spectrum of the charging loop is obtained, and the specific steps are as follows,

[0031] According to the adjusted output voltage, the predicted output voltage of the digital twin virtual model at the corresponding time is collected;

[0032] According to the difference between the actual output voltage and the predicted output voltage, the error feature is extracted;

[0033] According to the error feature, the internal parameters of the digital twin virtual model are dynamically calibrated through reinforcement learning driven strategy;

[0034] The calibrated digital twin virtual model is used to start a new control cycle, and a small signal disturbance is injected to obtain a preliminary impedance response spectrum of the charging loop.

[0035] As a preferred scheme of the AC charging pile control system with voltage stabilization function, the preliminary impedance response spectrum is used to obtain the parameter configuration instruction of the high-frequency test current signal and inject the high-frequency test current signal, and bandpass filtering and fast Fourier transform are performed to obtain the impedance characteristic, and the specific steps are as follows,

[0036] The parameter configuration instruction of the high-frequency test current signal is obtained from the preliminary impedance response spectrum;

[0037] According to the parameter configuration instruction, the signal generator generates a high-frequency test current signal with corresponding frequency scanning and injects it into the charging loop, and the voltage response signal of the charging loop is collected;

[0038] The voltage response signal and the high-frequency test current signal are bandpass filtered and fast Fourier transformed to obtain the frequency domain voltage spectrum and the current spectrum;

[0039] The measured impedance spectrum is obtained from the frequency domain voltage spectrum and the current spectrum, and the measured impedance spectrum is fused with the predicted impedance spectrum of the digital twin virtual model to obtain the impedance characteristic.

[0040] As a preferred scheme of the AC charging pile control system with voltage stabilization function, the parameter configuration instruction contains the frequency scanning range and the signal amplitude.

[0041] As a preferred scheme of the AC charging pile control system with voltage stabilization function, wherein: the multi-dimensional feature vector is extracted from the impedance characteristic, and the power grid risk index is calculated, and the fault judgment is performed at the same time, forming a closed loop optimization, and the specific steps are as follows,

[0042] The multi-dimensional feature vector is extracted from the impedance characteristic, and the real-time load current and the environmental temperature are collected at the same time;

[0043] According to the multi-dimensional feature vector, the real-time load current and the environmental temperature, the power grid risk index is calculated;

[0044] When the power grid risk index is in the first threshold interval, it is judged as a poor contact risk and a warning is triggered, and when the power grid risk index exceeds the second threshold, it is judged as an insulation fault and switched to a safe operation mode;

[0045] Based on the judgment result, the weight parameter in the dynamic risk index calculation is dynamically adjusted through the reinforcement learning algorithm, and closed loop optimization is performed.

[0046] The beneficial effects of the present application are: the multi-domain response spectrum is generated through the digital twin model, the power grid disturbance waveform is predicted in advance, and the control process has the ability of predictive compensation; through high-frequency test signal injection and impedance characteristic analysis, closed loop iterative optimization of fault judgment and parameter calibration is formed, and precise stable control of output voltage and active safety protection of operation state are realized. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0048] Fig. 1 It is a schematic diagram of the AC charging pile control system with voltage stabilization function.

[0049] Fig. 2 It is a flowchart for generating the multi-domain response spectrum of the power grid.

[0050] Fig. 3 It is a flowchart for predicting the future power grid disturbance waveform.

[0051] Fig. 4 It is a flowchart for generating the feedforward compensation amount and the error signal. DETAILED DESCRIPTION

[0052] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0053] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details set forth in this description. In other instances, well-known methods have not been described in detail in order to avoid unnecessarily obscuring the present application. Therefore, the present application is not intended to be limited by the specific implementations disclosed herein, which can vary in a number of ways.

[0054] Second, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic under discussion. Thus, "one embodiment" does not mean a single embodiment nor is it to be taken individually or selectively from other embodiments.

[0055] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides an alternating current charging pile control system with voltage stabilization function, including the following steps:

[0056] The signal acquisition module is used to acquire real-time power grid voltage signals and output current signals, and to construct a digital twin virtual model to generate a multi-domain response spectrum of the power grid.

[0057] The real-time power grid voltage signals and output current signals are acquired and processed by time domain denoising and filtering to obtain time domain signals.

[0058] The specific process includes: real-time sampling of grid-side voltage through an electrical sensor to obtain real-time power grid voltage signals; real-time sampling of charging pile output current through a current transformer to obtain output current signals; then, time domain denoising and filtering of real-time power grid voltage signals and output current signals, including using a moving average filter to average data points within a local window to achieve signal smoothing, using a median filter to replace the current point with a sliding window median to suppress pulse interference, and suppressing noise while retaining signal characteristics, and finally obtaining time domain signals.

[0059] The time domain signals are converted to time-frequency domain representation, and time-frequency features are extracted through fast Fourier transform.

[0060] The specific process includes: converting the time domain signals of real-time power grid voltage and output current to time-frequency domain representation through short-time Fourier transform, which obtains the frequency component distribution of time domain signals at different time points by segmenting and windowing the time domain signals and performing fast Fourier transform on each windowed signal segment, and then extracting time-frequency features from the time-frequency domain representation through fast Fourier transform, including the amplitude and phase information of each frequency component, and finally completing the extraction of time-frequency features.

[0061] A digital twin virtual model is constructed according to the time-frequency features, and the structure and parameters of the digital twin virtual model are adjusted.

[0062] The specific process includes: performing time-frequency transformation on the real-time grid voltage signal and the output current signal to obtain time-frequency features containing time and frequency dimensions, and inputting the time-frequency features into an initial architecture of a digital twin virtual model, which is composed of a state space equation or an equivalent circuit topology, wherein the structural parameters correspond to the electrical dynamic characteristics of the grid; simulating and deducing the current grid state using the initial architecture of the digital twin virtual model to generate a predicted output, comparing the actual collected adjusted output voltage with the predicted output of the initial architecture of the digital twin virtual model to obtain a state error, and iteratively updating the adjustable parameters in the initial architecture of the digital twin virtual model according to the state error using a gradient descent method or a recursive least squares method, and determining whether the number of state variables or the topology connection relationship of the initial architecture of the digital twin virtual model needs to be increased or decreased according to the error convergence characteristics; after the structure and parameters are adjusted, the updated initial architecture of the digital twin virtual model is determined as the digital twin virtual model used in the current control period, realizing the construction and calibration of the digital twin virtual model.

[0063] The adjusted digital twin virtual model is analyzed for time-space characteristics to obtain a multi-domain response spectrum of the grid.

[0064] The specific process includes: using the adjusted digital twin virtual model to analyze the dynamic changes of the real-time grid voltage signal and the output current signal at different time points and frequency dimensions for time-space characteristics, extracting feature information of the real-time grid voltage signal and the output current signal in the time domain, the frequency domain and the coupling relationship between the two (for example: instantaneous amplitude fluctuation, phase shift, total harmonic distortion (THD), inter-harmonic component, voltage sag / rise duration, current zero-crossing offset, power factor dynamic change, and voltage-current cross-spectrum density, coherence function, time-frequency joint distribution, and other time-frequency coupling characteristics of the voltage / current), and mapping these feature information to a unified representation space to generate a multi-domain response spectrum of the grid reflecting the multi-dimensional dynamic response characteristics of the grid.

[0065] A disturbance prediction module is configured to construct a dynamic graph structure based on the multi-domain response spectrum, and predict a future grid disturbance waveform through a space-time attention mechanism.

[0066] A dynamic graph structure is constructed based on the time-space characteristic dimensions in the multi-domain response spectrum.

[0067] The specific process includes that, based on the time-space feature dimension in the multi-domain response spectrum of the power grid, different time points and response features in the frequency dimension are taken as nodes (each node represents the dynamic state of the power grid under a specific time-frequency combination, including voltage amplitude, current phase, harmonic energy, and coherence, etc. multi-dimensional attributes), and connection edges are established according to the mutual relationship of the nodes in time evolution and frequency correlation (for example, the state transition continuity between adjacent time steps, the energy correlation of the same frequency component at different time points, the harmonic coupling relationship between different frequency components, and the nonlinear intermodulation or phase synchronization relationship across the frequency band), to form a dynamic graph structure that can reflect the time-varying dynamic behavior of the power grid.

[0068] The dynamic dependence relationship between the nodes is extracted from the dynamic graph structure through the space-time attention mechanism, and a space-time feature vector is generated.

[0069] The specific process includes that, the space-time attention mechanism is applied to the dynamic graph structure, the mutual influence of the nodes in the dynamic graph structure at different time steps and frequency dimensions is analyzed, and the attention weights of the connections between the nodes are allocated, the dynamic dependence relationship between the nodes is described according to the attention weights, and the dynamic dependence relationship is integrated into a space-time feature vector that can represent the space-time evolution characteristics of the power grid disturbance.

[0070] It should be noted that the space-time attention mechanism describes the dynamic dependence relationship between the nodes by allocating attention weights to the connections between the nodes at different time steps and frequency dimensions in the dynamic graph structure.

[0071] The space-time feature vector is subjected to a weighted integral operation to generate a prediction parameter of the future state of the power grid.

[0072] The specific process includes that, the space-time feature vector is subjected to a weighted integral operation, the space-time feature vector is integrated according to the preset weighting coefficient in the time dimension, the feature contribution degrees of different time points are fused according to the importance of the space-time feature vector, and a prediction parameter that can represent the future disturbance trend of the power grid is generated.

[0073] It should be noted that the weighting coefficient is determined in the training stage of the digital twin virtual model according to the influence degree of the features at different time points on the disturbance evolution in the historical disturbance data of the power grid.

[0074] The feature contribution degree refers to the relative importance or influence of the space-time feature vector at different time points on the prediction of the future disturbance trend of the power grid.

[0075] The prediction parameter of the future state of the power grid is subjected to time domain simulation to predict the future disturbance waveform of the power grid.

[0076] The specific process includes time domain simulation of the prediction parameters of the future state of the power grid, inputting the prediction parameters of the future state of the power grid as initial conditions and boundary conditions into the time domain simulation process, obtaining differential equations describing the dynamic behavior of the power grid by using a numerical integration method, and performing step-by-step advancement on the time axis to generate continuous time series waveforms reflecting the voltage or current disturbance patterns that may occur in the power grid in the future time period, and finally obtaining the future power grid disturbance waveform.

[0077] The difference compensation module is configured to perform difference analysis on the future power grid disturbance waveform and the real-time power grid voltage signal through a convolutional neural network to generate a feedforward compensation amount, and simultaneously collect an actual output voltage and a voltage set value, and generate an error signal through a fast Fourier transform.

[0078] The time-frequency features of the power grid voltage and the future power grid disturbance waveform are extracted from the real-time power grid voltage signal and the future power grid disturbance waveform.

[0079] The specific process includes performing short-time Fourier transform processing on the real-time power grid voltage signal and the future power grid disturbance waveform through time-frequency domain decomposition, expanding the real-time power grid voltage signal and the future power grid disturbance waveform in the joint time-frequency domain to obtain the frequency components and amplitude-phase information corresponding to different time points, thereby extracting the time-frequency features of the power grid voltage and the time-frequency features of the future power grid disturbance waveform.

[0080] The time-frequency features are analyzed through a convolutional neural network to obtain difference information between the power grid voltage and the future power grid disturbance waveform, and to capture the dynamic characteristics of the power grid in the power grid voltage change.

[0081] The specific process includes performing difference analysis on the time-frequency features of the real-time power grid voltage signal and the time-frequency features of the future power grid disturbance waveform through a convolutional neural network, using the multi-layer convolution and pooling operations of the convolutional neural network to extract the local and global patterns of the power grid voltage and the future power grid disturbance waveform in the time and frequency dimensions layer by layer, and comparing the feature response differences of the power grid voltage and the future power grid disturbance waveform at the same space-time position, thereby obtaining the difference information between the power grid voltage and the future power grid disturbance waveform, and identifying the nonlinear change law of the power grid voltage evolution over time and capturing the dynamic characteristics of the power grid in the power grid voltage change in the process.

[0082] Further, the training process of the convolutional neural network takes the time-frequency features of the grid voltage and the time-frequency features of the future grid disturbance waveform as inputs, takes the difference information (such as voltage amplitude deviation, phase mismatch, harmonic energy difference, or time-frequency response residual) between the two as a supervised signal, updates the weight parameters of the convolutional neural network using the back propagation algorithm, so that the convolutional neural network can accurately extract the difference information between the grid voltage and the future grid disturbance waveform and effectively capture the dynamic characteristics of the grid voltage change (such as inertia response characteristics, damping decay behavior, or excitation state of the resonance mode).

[0083] It should be noted that the local mode refers to the transient characteristics appearing in a short time window or a narrow frequency band, such as the initial edge of voltage sag, the burst energy of a specific harmonic, and the phase jump point; the global mode refers to the structural characteristics across a longer time period or a wide frequency range, such as the base frequency drift trend, the overall distortion form of the harmonic spectrum, and the time-frequency distribution center migration of the disturbance energy.

[0084] Based on the grid dynamic characteristics, the feedforward compensation amount of the grid voltage is calculated by combining the real-time grid voltage signal and the future grid disturbance waveform, and the expression is:

[0085] ;

[0086] Among them, represents the feedforward compensation amount at time , represents the time variable, represents the integration time range, represents the weighting coefficient at time , represents the error between the real-time grid voltage signal and the future grid disturbance waveform at time , represents the grid dynamic characteristics at time .

[0087] The specific process includes, based on the grid dynamic characteristics, combining the real-time grid voltage signal and the future grid disturbance waveform, first determining the deviation between the real-time grid voltage signal and the future grid disturbance waveform at each time point, and according to the grid voltage change law reflected by the grid dynamic characteristics, weighting the deviation at each time point in the time dimension, and integrating the weighted deviation at each time point to generate the feedforward compensation amount for compensating the impact of the grid disturbance.

[0088] The actual output voltage and the voltage set value are collected and a high-frequency test current signal is injected, and the error signal frequency domain features between the actual output voltage and the voltage set value are obtained by using fast Fourier transform.

[0089] The specific process includes: collecting the actual output voltage through an electrical sensor, obtaining the voltage set value through a setting unit, subtracting the two to obtain a time-domain error signal, and synchronously injecting a high-frequency test current signal to excite the high-frequency response of the charging loop, then performing fast Fourier transform on the time-domain error signal containing the high-frequency disturbance component to convert the time-domain error signal into a frequency-domain representation, thereby obtaining the amplitude and phase distribution of the time-domain error signal at different frequencies, forming the error signal frequency-domain feature.

[0090] It should be noted that the setting unit refers to a module for providing or configuring the charging pile output voltage target value (i.e. the voltage set value), which can be a user input interface, an upper computer instruction receiver or a built-in control strategy generator.

[0091] The error signal frequency-domain feature refers to the amplitude, phase and distribution characteristics of each frequency component obtained by spectral analysis of the time-domain error signal, including fundamental deviation, harmonic content and high-frequency disturbance response, etc.

[0092] The amplitude and phase of the error signal frequency-domain feature are optimized and adjusted through frequency-domain weighted filtering, and the optimized error signal is generated in combination with the dynamic characteristics of the power grid.

[0093] The specific process includes: optimizing and adjusting the amplitude and phase of the error signal frequency-domain feature through frequency-domain weighted filtering, specifically referring to setting the corresponding weighting coefficients in the frequency-domain representation obtained by fast Fourier transform according to the influence degree of each frequency component on the voltage stabilization control performance, respectively weighting and correcting the amplitude spectrum and phase spectrum in the error signal frequency-domain feature, weakening the frequency components affected by noise or non-critical disturbances, and strengthening the frequency components closely related to the voltage stability of the power grid (such as the steady-state deviation component near the fundamental frequency, the low-frequency voltage drift component and the sensitive frequency band coupled with the charging load resonance frequency), combining the weighted amplitude and phase information with the voltage change trend embodied by the dynamic characteristics of the power grid, and generating the optimized error signal in the time domain through the inverse frequency-domain reconstruction method.

[0094] It should be noted that the inverse frequency-domain reconstruction method refers to converting the weighted and corrected frequency-domain amplitude and phase information back to the time domain through inverse fast Fourier transform or minimum phase reconstruction, etc., to generate an optimized error signal that retains key dynamic characteristics and suppresses interference components.

[0095] It should be noted that the dynamic characteristics of the power grid refer to the nonlinear change law of the power grid voltage evolving over time under disturbance or load change, which is captured through deep learning processing of the time-frequency characteristics of the real-time power grid voltage signal and future power grid disturbance waveform by a convolutional neural network.

[0096] The instruction output module is configured to input the feedforward compensation and the error signal into a multivariable predictive coordinated controller, output a composite robust control instruction, and convert the composite robust control instruction into a pulse width modulation signal for adjusting an output voltage.

[0097] The feedforward compensation and the optimized error signal are input into a multivariable predictive coordinated controller, and an original control instruction is output.

[0098] The specific process includes simultaneously inputting the feedforward compensation and the optimized error signal into the multivariable predictive coordinated controller. The multivariable predictive coordinated controller performs collaborative prediction under multivariable coupling constraints based on future disturbance pre-compensation information represented by the feedforward compensation and current output deviation reflected by the optimized error signal, and outputs an original control instruction by comprehensively considering the advance nature and accuracy of voltage regulation.

[0099] It should be noted that the multivariable predictive coordinated controller is a controller that performs collaborative prediction and generates a control instruction based on multivariable coupling relationships.

[0100] The original control instruction is dynamically corrected by a nonlinear disturbance observer to generate a composite robust control instruction.

[0101] The specific process includes dynamically correcting the original control instruction by a nonlinear disturbance observer. The nonlinear disturbance observer estimates nonlinear disturbance effects existing in the process of power grid operation in real time, and superimposes disturbance estimation results of the nonlinear disturbance effects on the original control instruction to compensate for deviations caused by external disturbances and uncertainties of the digital twin virtual model, thereby generating a composite robust control instruction with stronger anti-interference capability.

[0102] It should be noted that the nonlinear disturbance observer is an observer constructed based on nonlinear dynamic equations, which is used to estimate nonlinear disturbance effects caused by factors such as load mutation, voltage fluctuation, or parameter time variation in the process of power grid operation in real time, and output disturbance estimation values to assist correction of the control instruction.

[0103] The composite robust control instruction is converted into a pulse width modulation signal with a specific duty cycle through a reverse symmetric pulse width modulation strategy.

[0104] The specific process comprises: converting the composite robust control instruction into a pulse width modulation signal through a reverse symmetrical pulse width modulation strategy, the reverse symmetrical pulse width modulation strategy generating two complementary pulse signals in each switching cycle, the on time of one pulse being equal to the off time of the other pulse, the two pulses being reversely symmetrically distributed on the time axis, and the duty cycle being determined by the amplitude of the composite robust control instruction; when the composite robust control instruction increases, the on time of the forward pulse is prolonged and the on time of the reverse pulse is shortened accordingly, so as to ensure the regulation accuracy and dynamic response performance of the output voltage, and finally generate a pulse width modulation signal with a specific duty cycle.

[0105] It should be noted that the reverse symmetrical pulse width modulation strategy is a modulation method for generating a pulse width modulation signal with a specific duty cycle by symmetrically distributing high-level and low-level pulses about the midpoint of the cycle in each switching cycle and dynamically adjusting the on and off time ratio according to the composite robust control instruction.

[0106] The switch tube in the power converter is driven by the pulse width modulation signal to adjust the output voltage.

[0107] The specific process comprises: driving the switch tube in the power converter by the pulse width modulation signal, controlling the on and off time ratio of the switch tube by adjusting the duty cycle of the pulse width modulation signal, changing the energy transfer characteristics of the power converter, thereby adjusting the amplitude of the output voltage, and making the output voltage approach the voltage set value.

[0108] The model calibration module is used to feed back the adjusted output voltage to the digital twin virtual model for parameter calibration, and simultaneously start a new control cycle according to the calibrated digital twin virtual model and obtain the preliminary impedance response spectrum of the charging loop.

[0109] According to the adjusted output voltage, the predicted output voltage of the digital twin virtual model at the corresponding time is collected.

[0110] It should be noted that according to the adjusted output voltage, the predicted output voltage generated by the digital twin virtual model at the same time point is obtained through the running state of the digital twin virtual model at the corresponding time, and the corresponding relationship between the adjusted output voltage and the predicted output voltage of the digital twin virtual model is established, which provides a basis for the extraction of error features in the subsequent process.

[0111] According to the difference between the actual output voltage and the predicted output voltage, the error feature is extracted.

[0112] The specific process includes comparing the values of the actual output voltage and the predicted output voltage at the same time point according to the difference between the actual output voltage and the predicted output voltage, obtaining a voltage deviation sequence, and performing time domain and frequency domain analysis on the voltage deviation sequence to extract characteristic quantities reflecting the prediction accuracy of the digital twin virtual model and the degree of deviation from the actual running state, and forming error characteristics.

[0113] According to the error characteristics, the internal parameters of the digital twin virtual model are dynamically calibrated through a reinforcement learning driven strategy.

[0114] The specific process includes quantifying the error characteristics into reward or punishment signals, inputting the reinforcement learning agent, evaluating the pros and cons of the current prediction performance of the digital twin virtual model according to the cumulative return, and generating parameter adjustment actions to update the internal parameters of the digital twin virtual model in the control cycle iteration process, thereby improving the prediction accuracy of the digital twin virtual model for future power grid behavior.

[0115] A new control cycle is started using the calibrated digital twin virtual model, and a small signal disturbance is injected to obtain the preliminary impedance response spectrum of the charging loop.

[0116] The specific process includes starting a new control cycle using the calibrated digital twin virtual model, injecting a small signal disturbance into the charging loop at the initial stage of the new control cycle, wherein the small signal disturbance is a test signal with small amplitude and controllable frequency, then collecting the voltage response signal and current response signal of the charging loop to the small signal disturbance, and converting the voltage response signal and current response signal into frequency domain voltage spectrum and frequency domain current spectrum through frequency selective filtering and spectrum analysis, and then obtaining the impedance value of the charging loop at the corresponding frequency according to the ratio of the frequency domain voltage spectrum and the frequency domain current spectrum at each frequency point, and combining the impedance values at each frequency point to form the preliminary impedance response spectrum of the charging loop.

[0117] It should be noted that frequency selective filtering is used to extract response signals in a specific frequency band to suppress noise interference, and spectrum analysis converts time domain response signals into frequency domain representation through methods such as fast Fourier transform, thereby obtaining the amplitude and phase information of voltage and current at each frequency component.

[0118] The impedance generation module is configured to obtain the parameter configuration instruction of the high-frequency test current signal based on the preliminary impedance response spectrum, inject the high-frequency test current signal, and perform bandpass filtering and fast Fourier transform to obtain the impedance characteristics.

[0119] The parameter configuration instruction of the high-frequency test current signal is obtained from the preliminary impedance response spectrum.

[0120] The specific process includes identifying the frequency region representing the key role by analyzing the impedance amplitude and phase change of each frequency point in the preliminary impedance response spectrum, setting the minimum frequency and the maximum frequency of the frequency region as the frequency scanning range of the high-frequency test current signal, and determining the signal amplitude of the high-frequency test current signal at the corresponding frequency according to the sensitivity of the impedance response of each frequency point. Finally, the frequency scanning range and signal amplitude are combined into the parameter configuration instruction of the high-frequency test current signal.

[0121] According to the parameter configuration instruction, the signal generator generates a high-frequency test current signal with corresponding frequency scanning and injects it into the charging circuit, while collecting the voltage response signal of the charging circuit.

[0122] The specific process includes driving the signal generator to gradually sweep the frequency from the starting frequency to the ending frequency according to the frequency scanning range, and outputting the corresponding amplitude of the sine or sweep current waveform according to the signal amplitude at each frequency point, thereby generating a complete high-frequency test current signal; after the high-frequency test current signal is injected into the charging circuit, it excites the electrical response of the charging circuit in a wide frequency band, and at the same time, the voltage response signal generated by the high-frequency test current signal is synchronously collected through the electrical sensor, providing basic data for subsequent impedance characteristic analysis.

[0123] The voltage response signal and the high-frequency test current signal are bandpass filtered and fast Fourier transformed to obtain the frequency domain voltage spectrum and the current spectrum.

[0124] The specific process includes bandpass filtering the voltage response signal and the high-frequency test current signal respectively, filtering out the out-of-band noise and irrelevant frequency band interference, retaining the effective components consistent with the frequency range of the high-frequency test current signal, and then performing fast Fourier transform on the filtered voltage response signal and high-frequency test current signal respectively, converting the time domain signal to frequency domain representation, thereby obtaining the frequency domain voltage spectrum composed of voltage amplitude and phase information corresponding to each frequency point and the frequency domain current spectrum composed of current amplitude and phase information.

[0125] The measured impedance spectrum is obtained from the frequency domain voltage spectrum and the current spectrum, and the measured impedance spectrum is fused with the predicted impedance spectrum of the digital twin virtual model to obtain the impedance characteristic.

[0126] The specific process includes extracting the complex voltage and complex current components corresponding to each frequency point from the frequency domain voltage spectrum and frequency domain current spectrum. Based on the ratio of the complex voltage component to the complex current component, the complex impedance value at that frequency point is determined. After traversing all frequency points in sequence, the complex impedance values ​​are arranged in frequency order to form a measured impedance spectrum. The measured impedance spectrum is compared with the predicted impedance spectrum of the digital twin virtual model at the same frequency point. A weighted average fusion strategy is used to combine the measured information of the measured impedance spectrum with the prior information of the predicted impedance spectrum of the digital twin virtual model to generate an impedance characteristic that includes both measured data and model predictions.

[0127] The risk assessment module is used to extract multi-dimensional feature vectors from impedance characteristics and calculate the power grid risk index, while simultaneously determining faults to form a closed-loop optimization.

[0128] Multi-dimensional feature vectors are extracted from impedance characteristics, while real-time load current and ambient temperature are collected.

[0129] The specific process includes extracting multi-dimensional feature vectors reflecting the electrical state of the charging circuit from the impedance characteristics. These multi-dimensional feature vectors include characteristics such as impedance amplitude, impedance phase, impedance frequency response slope, and resonant point distribution. At the same time, real-time load current is collected through a current sensor, and ambient temperature is collected through a temperature sensor, providing input basis for the subsequent calculation of the power grid risk index.

[0130] Based on multi-dimensional feature vectors, real-time load current, and ambient temperature, the power grid risk index is calculated as follows:

[0131] ;

[0132] in, Indicates time The power grid risk index Indicates the ambient temperature regulation factor. Indicates time Ambient temperature, Indicates the reference ambient temperature. Indicates time The multidimensional feature vectors, Indicates time Real-time load current of the power grid.

[0133] The specific process includes calculating the power grid risk index according to the multi-dimensional feature vector, the real-time load current and the environmental temperature, taking the deviation of the environmental temperature from the reference environmental temperature as the basis, depicting the enhancement effect of the deviation of the environmental temperature from the reference environmental temperature on the power grid risk through a nonlinear function, combining the impedance characteristic complexity represented by the multi-dimensional feature vector and the coupling effect of the real-time load current, constructing a comprehensive risk measure, and normalizing and adjusting the comprehensive risk measure, and finally outputting a power grid risk index changing with time, which is used to quantify the safety degree of the current power grid operation state.

[0134] It should be noted that the environmental temperature adjustment factor is a correction coefficient for quantifying the influence degree of temperature on the power grid risk, which is obtained by a temperature-risk mapping function (such as an exponential function, a piecewise linear function or a lookup table method) according to the deviation of the environmental temperature from the reference temperature.

[0135] When the power grid risk index is in the first threshold interval, it is determined that there is a poor contact risk and a warning is triggered, and when the power grid risk index exceeds the second threshold, it is determined that there is an insulation failure and the safe operation mode is switched to.

[0136] The specific process includes that when the power grid risk index is in the first threshold interval, it indicates that the comprehensive state of the impedance characteristic of the charging circuit, the real-time load current and the environmental temperature deviates from the normal range but has not reached a dangerous degree, at this time it is determined that the charging circuit has a poor contact risk, and a warning is triggered to prompt the operation and maintenance personnel to pay attention to the potential connection loosening or contact resistance increase problem; when the power grid risk index exceeds the second threshold, it indicates that the impedance abnormality is intensified, and the load current and the environmental temperature jointly cause the insulation performance to deteriorate, at this time it is determined that the charging circuit has an insulation failure, and the safe operation mode is immediately switched to, the output power is limited or the high-voltage output is cut off, to prevent the fault from expanding and to protect the equipment and personnel safety.

[0137] It should be noted that the first threshold interval is that the power grid risk index is between 1.2 and 1.8, which is set according to the statistical range of the power grid risk index of the charging circuit before the contact is poor in the historical operation data. The second threshold is that the power grid risk index is greater than 2.5, which is set according to the critical value of the power grid risk index when the insulation failure of the charging circuit occurs and combined with the safety margin.

[0138] Based on the determination result, the weight parameters in the dynamic risk index calculation are dynamically adjusted through a reinforcement learning algorithm, and closed-loop optimization is performed.

[0139] The specific process comprises: based on the determination result, dynamically adjusting the weight parameters corresponding to the multi-dimensional feature vectors, the real-time load current and the environmental temperature in the dynamic risk index calculation through a reinforcement learning algorithm, the reinforcement learning algorithm taking the accuracy of the poor contact risk early warning and the timeliness of the insulation fault identification as feedback signals, constantly optimizing the contribution proportion of each input factor in the dynamic risk index calculation, forming a closed-loop optimization mechanism, so that the dynamic risk index can more accurately reflect the actual operation risk state of the charging loop.

[0140] It should be noted that the reinforcement learning algorithm is a machine learning method that obtains feedback signals by interacting with the environment and constantly adjusts the decision-making strategy according to the feedback signals to optimize the long-term target performance.

[0141] To sum up, the present application generates a multi-domain response spectrum through a digital twin model, realizes the early prediction of the power grid disturbance waveform, and makes the control process have the ability of predictive compensation; through high-frequency test signal injection and impedance characteristic analysis, a closed-loop iterative optimization of fault determination and parameter calibration is formed, and the accurate stable control of the output voltage and the active safety protection of the operating state are realized.

[0142] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. An alternating current charging pile control system with voltage stabilization function, characterized in that: The application relates to a power grid real-time control method and device. The signal acquisition module is used for acquiring real-time power grid voltage signals and output current signals, and constructing a digital twin virtual model to generate a multi-domain response spectrum of the power grid, and the specific steps are as follows. The real-time power grid voltage signals and the output current signals are acquired, and time-domain denoising and filtering are performed to obtain time-domain signals. The time-domain signals are converted into time-frequency domain representations, and time-frequency features are extracted through fast Fourier transform. The digital twin virtual model is constructed according to the time-frequency features, and the structure and parameters of the digital twin virtual model are adjusted. The adjusted digital twin virtual model is analyzed in time and space to obtain the multi-domain response spectrum of the power grid. The disturbance prediction module is used for constructing a dynamic graph structure based on the multi-domain response spectrum, and predicting a future power grid disturbance waveform through a time-space attention mechanism. The difference compensation module is used for performing difference analysis on the future power grid disturbance waveform and the real-time power grid voltage signal through a convolutional neural network to generate a feedforward compensation amount. The instruction output module is used for inputting the feedforward compensation amount and the error signal into a multivariable prediction coordination controller to output a composite robust control instruction, and converting the composite robust control instruction into an execution pulse width modulation signal to adjust the output voltage. The model calibration module is used for feeding back the adjusted output voltage to the digital twin virtual model for parameter calibration, starting a new control cycle according to the calibrated digital twin virtual model, and obtaining a preliminary impedance response spectrum of the charging loop. The impedance generation module is used for obtaining parameter configuration instructions of a high-frequency test current signal based on the preliminary impedance response spectrum, injecting the high-frequency test current signal, performing band-pass filtering and fast Fourier transform, and obtaining impedance characteristics. The risk judgment module is used for extracting a multi-dimensional feature vector from the impedance characteristics and calculating a power grid risk index, and performing fault judgment to form a closed-loop optimization.

2. The AC charging pile control system with voltage stabilization function according to claim 1, characterized in that: The dynamic graph structure is constructed based on the time-space feature dimensions in the multi-domain response spectrum. The dynamic dependence relationship between nodes in the dynamic graph structure is extracted through the time-space attention mechanism to generate a time-space feature vector. The time-space feature vector is subjected to weighted integral operation to generate a prediction parameter of a future state of the power grid. The prediction parameter of the future state of the power grid is subjected to time-domain simulation to predict a future power grid disturbance waveform. The time-frequency features of the power grid voltage and the future power grid disturbance waveform are extracted from the real-time power grid voltage signals and the future power grid disturbance waveform.

3. The AC charging pile control system with voltage stabilization function according to claim 2, characterized in that: ​ ​ The difference information between the grid voltage and the future grid disturbance waveform is obtained by convolutional neural network difference analysis on the time-frequency characteristics, and the dynamic characteristics of the grid in the grid voltage change are captured. Based on the dynamic characteristics of the grid, the real-time grid voltage signal and the future grid disturbance waveform are combined to calculate the feedforward compensation of the grid voltage.

4. The AC charging pile control system with voltage stabilization function according to claim 3, characterized in that: The specific steps of generating the error signal are as follows, The actual output voltage and the voltage set value are collected and a high-frequency test current signal is injected, and the error signal frequency domain characteristics between the actual output voltage and the voltage set value are obtained by using fast Fourier transform; The amplitude and phase of the error signal frequency domain characteristics are optimized and adjusted by frequency domain weighted filtering, and the optimized error signal is generated in combination with the dynamic characteristics of the grid.

5. The AC charging pile control system with voltage stabilization function according to claim 4, characterized in that: The specific steps of inputting the feedforward compensation and the error signal into the multivariable predictive coordinated controller to output the composite robust control instruction and converting it into the execution pulse width modulation signal to adjust the output voltage are as follows, The feedforward compensation and the optimized error signal are input into the multivariable predictive coordinated controller to generate the original control instruction. The original control instruction is dynamically corrected by the nonlinear disturbance observer to output the composite robust control instruction. The composite robust control instruction is converted into a pulse width modulation signal with a specific duty cycle by the reverse symmetric pulse width modulation strategy, which means that the composite robust control instruction is converted into a pulse width modulation signal by the reverse symmetric pulse width modulation strategy. The reverse symmetric pulse width modulation strategy generates two complementary pulse signals in each switching period, one of which has an on-time equal to the off-time of the other, and the two are distributed in reverse symmetry on the time axis. The duty cycle is determined by the amplitude of the composite robust control instruction. When the composite robust control instruction increases, the on-time of the positive pulse is extended and the on-time of the reverse pulse is shortened accordingly, thereby ensuring the adjustment accuracy and dynamic response performance of the output voltage. Finally, a pulse width modulation signal with a specific duty cycle is generated. The switching tube in the power converter is driven by the pulse width modulation signal to adjust the output voltage.

6. The AC charging pile control system with voltage stabilization function according to claim 5, characterized in that: The specific steps of feeding back the adjusted output voltage to the digital twin virtual model for parameter calibration, starting a new control cycle according to the calibrated digital twin virtual model, and obtaining the preliminary impedance response spectrum of the charging loop are as follows, According to the adjusted output voltage, the predicted output voltage of the digital twin virtual model at the corresponding time is collected; According to the difference between the actual output voltage and the predicted output voltage, the error characteristics are extracted; According to the error characteristics, the internal parameters of the digital twin virtual model are dynamically calibrated by the reinforcement learning driven strategy. A new control cycle is started using the calibrated digital twin virtual model, and a small signal disturbance is injected to obtain the preliminary impedance response spectrum of the charging loop.

7. The AC charging pile control system with voltage stabilization function according to claim 6, characterized in that: The specific steps of obtaining the parameter configuration instruction of the high-frequency test current signal based on the preliminary impedance response spectrum, injecting the high-frequency test current signal, and performing bandpass filtering and fast Fourier transform to obtain the impedance characteristics are as follows, The parameter configuration instruction of the high-frequency test current signal is obtained from the preliminary impedance response spectrum; According to the parameter configuration instruction, the drive signal generator generates a high-frequency test current signal of a corresponding frequency scan and injects it into the charging loop, while collecting the voltage response signal of the charging loop; The voltage response signal and the high-frequency test current signal are band-pass filtered and fast Fourier transformed to obtain the frequency domain voltage spectrum and the current spectrum; The measured impedance spectrum is obtained from the frequency domain voltage spectrum and the current spectrum, and the measured impedance spectrum is fused with the predicted impedance spectrum of the digital twin virtual model to obtain the impedance characteristic.

8. The AC charging pile control system with voltage stabilization function according to claim 7, characterized in that: The parameter configuration instruction includes a frequency scan range and a signal amplitude. 9.The AC charging pile control system with voltage stabilization function of claim 8, wherein: The multi-dimensional feature vector is extracted from the impedance characteristic, and the power grid risk index is calculated, and the fault is determined to form a closed-loop optimization, and the specific steps are as follows, The multi-dimensional feature vector is extracted from the impedance characteristic, and the real-time load current and the environmental temperature are collected; According to the multi-dimensional feature vector, the real-time load current and the environmental temperature, the power grid risk index is calculated; When the power grid risk index is in the first threshold interval, it is determined as a poor contact risk and triggers a warning, and when the power grid risk index exceeds the second threshold, it is determined as an insulation fault and switches to a safe operation mode; Based on the determination result, the weight parameters in the dynamic risk index calculation are dynamically adjusted through the reinforcement learning algorithm, and closed-loop optimization is performed.

Images