V2G-based alternating current charging power adjusting method and device

Abstract

The invention relates to a V2G-based alternating current charging power adjusting method and device. The method comprises the following steps: carrying out multi-scale data acquisition and preprocessing on a V2G alternating current charging system to obtain a standardized system state data set; constructing a system dynamic model comprising a power grid voltage characteristic and a charging power characteristic; carrying out decomposition to obtain a power output mapping relation calculation, and obtaining a power modulation control parameter; establishing a fast and slow time scale coordination control framework, and designing a power regulation agent and a voltage regulation agent to obtain a coordination control strategy; checking a charging power limit value constraint, a voltage limit value constraint and a power factor constraint to obtain a constraint violation judgment result, and executing constraint projection calculation to obtain a target control instruction; and controlling a power conversion circuit of the charging pile to execute power regulation operation. According to the invention, a hierarchical control structure is adopted, the control complexity of the system is reduced, and the robustness and the adaptive ability of the control strategy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of AC charging power regulation, and in particular to a V2G-based AC charging power regulation method and device. Background Art

[0002] With the rapid development of the electric vehicle industry, V2G (Vehicle-to-Grid) technology, as a two-way interactive mode of energy flow, plays an important role in improving the peak-shaving capacity of the power grid and promoting the consumption of new energy. However, the traditional V2G AC charging system faces multiple challenges in the power regulation process. Due to the influence of factors such as grid fluctuations, changes in the characteristics of the on-board energy storage system, and load disturbances, the dynamic characteristics of the system show strong coupling, nonlinearity, and time-varying characteristics.

[0003] The current V2G AC charging power regulation method is mainly based on fixed parameter controller design, which is difficult to cope with the dynamic changes of system parameters and working conditions. At the same time, due to the different time scale characteristics of charging power regulation and grid voltage regulation, the traditional single time scale control strategy cannot meet the requirements of fast power response and voltage stability at the same time. In addition, multiple constraints such as voltage fluctuations on the grid side and charging and discharging power limits of the on-board energy storage system further increase the complexity of power regulation control. Summary of the invention

[0004] The main purpose of the present invention is to provide a V2G-based AC charging power regulation method and device. The present invention adopts a hierarchical control structure, which reduces the control complexity of the system and improves the robustness and adaptability of the control strategy.

[0005] To achieve the above object, the present invention provides a V2G-based AC charging power regulation method, comprising the following steps: Perform multi-scale data collection and preprocessing on the V2G AC charging system to obtain a standardized system status data set; Building a state space model including grid voltage characteristics and charging power characteristics based on the standardized system state data set, and verifying the state space model to obtain a system dynamic model; Decomposing the charging power instruction in the system dynamic model to obtain a reference power component and a compensation power component, and calculating the power output mapping relationship to obtain a power modulation control parameter; The power modulation control parameters are divided into fast power regulation parameters and voltage stability parameters, a fast and slow time scale coordinated control framework is established, and a power regulation agent and a voltage regulation agent are designed to obtain a coordinated control strategy; Checking the charging power limit constraint, the voltage limit constraint and the power factor constraint of the collaborative control strategy to obtain a constraint violation judgment result, and performing constraint projection calculation to obtain a target control instruction; The target control instruction is converted into a power regulation signal through a PWM modulator, and the power conversion circuit of the charging pile is controlled to perform a power regulation operation.

[0006] The present invention also provides an AC charging power adjustment device based on V2G, comprising: The acquisition module is used to collect and preprocess multi-scale data of the V2G AC charging system to obtain a standardized system status data set; A construction module, configured to construct a state space model including grid voltage characteristics and charging power characteristics based on the standardized system state data set, and verify the state space model to obtain a system dynamic model; A decomposition module, used to decompose the charging power instruction in the system dynamic model to obtain a reference power component and a compensation power component, and calculate the power output mapping relationship to obtain a power modulation control parameter; A design module is used to divide the power modulation control parameters into fast power regulation parameters and voltage stability parameters, establish a fast and slow time scale coordinated control framework, and design a power regulation agent and a voltage regulation agent to obtain a coordinated control strategy; A checking module, used to check the charging power limit constraint, voltage limit constraint and power factor constraint of the collaborative control strategy, obtain constraint violation judgment results, and perform constraint projection calculation to obtain target control instructions; The control module is used to convert the target control instruction into a power regulation signal through a PWM modulator, and control the power conversion circuit of the charging pile to perform a power regulation operation.

[0007] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0008] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0009] In summary, the technical solution provided by the present invention realizes efficient collection and standardized processing of grid-side and on-board energy storage system data through multi-scale data acquisition and preprocessing technology, and provides accurate system status information; adopts a method combining nonlinear sparse regression and unscented Kalman filtering to establish an accurate dynamic model including grid voltage characteristics and charging power characteristics; based on power command decomposition and mapping relationship construction, precise control of charging power is achieved, and the impact of grid fluctuations on the system is effectively suppressed; through the separation of fast and slow time scales and multi-agent reinforcement learning methods, the coordinated control problem of power regulation and voltage stability is solved; the constraint violation detection and projection calculation mechanism are introduced to ensure the safe and stable operation of the system under multiple constraints; the hierarchical control structure is adopted to reduce the control complexity of the system and improve the robustness and adaptability of the control strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 1 is a schematic diagram of steps of a method for adjusting AC charging power based on V2G in one embodiment of the present invention; Figure 2 It is a structural block diagram of an AC charging power adjustment device based on V2G in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0011] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0013] Reference Figure 1 This embodiment provides a V2G-based AC charging power adjustment method, comprising the following steps: S1, multi-scale data collection and preprocessing of the V2G AC charging system to obtain a standardized system status data set; In this embodiment, the basic principle of the V2G AC charging power conditioning device to transmit vehicle power to the power grid is based on power flow theory and phase control theory. Specifically, when the electric vehicle is used as a distributed power source to feed power to the power grid, the DC power of the on-board energy storage device needs to be converted into AC power with the same frequency and phase as the power grid through a power conversion circuit. The details are as follows: According to the power flow theory of AC circuits, the direction and magnitude of bidirectional power transmission depends on the phase difference and amplitude ratio between the voltage vectors. Assume that the grid voltage is ° (as a phase reference), the output voltage of the power conversion circuit is , the equivalent impedance of the line is , then the transmitted active power P and reactive power Q can be expressed as: ; ; From the above formula, we can see that by controlling the output voltage The magnitude and phase angle of , can realize the bidirectional flow of power. When θ is positive, power flows from the vehicle to the grid; When it is negative, power flows from the grid to the vehicle. This theory provides a mathematical basis for the power control of the V2G system. In actual control, the device adopts a current control strategy based on instantaneous power theory. The three-phase voltage and current are converted to a synchronous rotating coordinate system through Clark transformation and Park transformation, and the following mathematical model is established: ; in, , is the dq axis voltage component, , is the dq axis current component, , is the component of the grid voltage on the dq axis, L is the equivalent inductance, R is the equivalent resistance, is the grid angular frequency. and , active power and reactive power can be adjusted separately. Based on the above theoretical model, the device combines space vector PWM technology to achieve precise power regulation. The duty cycle and phase of the PWM modulation signal are determined by the following relationship: ; ; Where m is the modulation index, is the DC bus voltage. Such a control mechanism ensures the stability and controllability of the process of power transmission to the power grid. The three-phase voltage of the AC power grid in the V2G AC charging system is collected to capture the original voltage data on the power grid side. At the same time, the three-phase current data in the charging circuit is collected in real time to obtain the original current data on the power grid side. By calculating the power of the original voltage data and the original current data, the instantaneous power characteristics of the system are obtained. At the same time, the component distribution of these signals in the frequency domain is analyzed by combining Fourier transform to separate the harmonic components and high-frequency interference in the system operation, and the power factor calculation data is obtained. Zero drift compensation is performed on the original voltage and current data and the power factor calculation data on the power grid side. Since the sampling circuit and sensor will introduce offsets, these offsets will cause the original data to deviate from the true value. These errors can be effectively eliminated through zero drift compensation technology to improve data quality. After completing zero drift compensation, the data is filtered to remove high-frequency noise and other interference signals to obtain filtered power grid side data. At the same time, the CAN bus interacts with the battery management system (BMS) of the electric vehicle in a preset communication cycle to obtain key parameters about the operation of the on-board energy storage system, including the battery state of charge (SOC), maximum allowable charging power, and maximum allowable discharge power. These data directly reflect the current operating capacity of the on-board energy storage system and the health status of the battery. The filtered grid-side data is combined with the on-board energy storage system operating parameters extracted from the BMS and standardized to eliminate the dimensional differences between different physical quantities. Through the standardization operation, all types of data are normalized to the same scale range to obtain normalized state data. Outlier detection is performed on the normalized state data, and abnormal data points are identified by setting reasonable thresholds or using statistical methods (such as Z scores or box plot methods). For the detected outliers, data reconstruction technology is used to repair them to obtain a standardized system state data set.

[0014] S2, construct a state space model including grid voltage characteristics and charging power characteristics based on the standardized system state data set, and verify the state space model to obtain the system dynamic model; Specifically, the standardized system state data set is divided into a training set and a validation set according to the time series. The training set contains grid voltage data, current data, power factor data and on-board energy storage system data for building the model; the validation set is used to test the accuracy and robustness of the model. The time series division ratio is adjusted according to the characteristics of the system data and the model requirements to ensure that the data can cover both long-term dynamic characteristics and short-term response characteristics. Based on the various types of data in the training set, the grid voltage data, current data, power factor data and on-board energy storage system data are processed using polynomial fitting technology. Through polynomial fitting, discrete data points are converted into a continuous state vector sequence to describe the dynamic change law of each variable in time. Sparse feature extraction is performed on the continuous state vector sequence, and the core dynamic characteristics of the system are captured by constructing a sparse coefficient matrix. Sparse feature extraction extracts the most representative features from high-dimensional data while ignoring redundant information, thereby simplifying the complexity of the model. The sparse coefficient matrix is ​​truncated to filter out high-frequency or irrelevant components that have little impact on the dynamic behavior of the system, and the nonlinear dynamic characteristics of the system are obtained. Based on the nonlinear dynamic characteristics, the state space is constructed and the dimension reduction is performed. Through the dimension reduction operation, the complex high-dimensional state features are compressed into the low-dimensional space, making the state space expression more concise and effective. In order to improve the accuracy of the state space model, the unscented Kalman filter is used to predict the state of the reduced state space features. The unscented Kalman filter is a nonlinear filtering method that can achieve accurate state estimation in nonlinear dynamic systems and generate time-related state prediction results. The state transition probability matrix of the system is calculated based on the state prediction results, which describes the probability distribution of the system from one state to another. After the initial estimation of the state transition probability, the maximum likelihood estimation method is used to optimize the transition probability parameters to ensure that the state transition probability matrix can accurately reflect the dynamic laws and behavioral characteristics of the system. The optimized transition probability parameters are combined with the reduced state space features to form a complete state space model. In order to verify the effectiveness and accuracy of the state space model, the model is predicted using the validation set, and the predicted state is compared with the actual state to calculate the root mean square error and relative error between them. The root mean square error measures the degree of deviation between the predicted value and the actual value, while the relative error is used to evaluate the proportion of the prediction error within the actual value range. These error indicators comprehensively reflect the prediction ability and applicability of the model. The model verification index is compared with the preset error threshold. When the root mean square error is less than the first target value and the relative error is less than the second target value, the model is considered to meet the accuracy requirements. At this time, the output state space model is the system dynamic model.

[0015] S3, decomposing the charging power command in the system dynamic model to obtain a reference power component and a compensation power component, and calculating the power output mapping relationship to obtain a power modulation control parameter; It should be noted that the grid voltage state data in the system dynamic model is subjected to discrete Fourier transform, and the fundamental component is separated from the complex time domain signal, and its amplitude and phase angle are extracted. The amplitude of the fundamental component represents the voltage amplitude characteristics of the grid, while the phase angle reflects the phase dynamic behavior of the grid. By calculating the real-time changes of these fundamental components, the real-time grid state characteristics are obtained. The real-time state characteristics of the grid are input into the power command decomposer for processing. The reference power correction coefficient is calculated based on the deviation rate between the voltage amplitude and the rated voltage. The deviation rate reflects the degree of deviation between the current state of the grid voltage and the standard rated value. The charging power command is corrected by this coefficient to obtain the reference power component. The reference power component is used to meet the basic charging needs of electric vehicles, and the basic balance between the charging system and the grid is maintained by adjusting this component. The power dynamic response data in the system dynamic model is subjected to wavelet decomposition. Wavelet decomposition is a multi-scale analysis method that can decompose the power dynamic response data into high-frequency disturbance components and low-frequency steady-state components. The high-frequency disturbance component is related to the short-term fluctuations in the grid, while the low-frequency steady-state component reflects the long-term trend of the system. After these dynamic characteristic components are input into the compensator, the dynamic compensation value is calculated by the proportional-integral-differential control algorithm to generate the compensation power component. The role of the compensation power component is to quickly respond to short-term fluctuations in the power grid, thereby maintaining the dynamic stability of the system. The reference power component and the compensation power component are linearly superimposed to construct the initial mapping relationship of the power output. Linear superposition is an effective method that takes into account the basic charging demand and the dynamic response demand. The initial power mapping function is further optimized by the piecewise linear interpolation method to generate a more accurate piecewise power mapping relationship. During the piecewise optimization process, the optimal mapping coefficient is calculated for each segment to ensure the accurate match between the power output and the system demand. The final piecewise power mapping relationship can dynamically adapt to different grid states and load conditions. The piecewise power mapping relationship is input into the modulator parameter calculation module to generate the modulation control parameters for power conversion. Based on the power command value, the corresponding pulse width modulation (PWM) depth and phase angle are calculated. The PWM modulation depth determines the amplitude of the power output, while the phase angle affects the power factor between the charging system and the power grid. In order to improve the accuracy of the modulation parameters, the modulation control reference parameters are compensated for dead time and switching loss. Dead time compensation is used to eliminate the delay problem caused by the non-ideal characteristics of switching devices in power conversion circuits, while switching loss compensation can minimize energy loss, improve the energy utilization efficiency of the system, and ultimately obtain power modulation control parameters.

[0016] S4, divide the power modulation control parameters into fast power regulation parameters and voltage stability parameters, establish a fast and slow time scale coordinated control framework, and design the power regulation agent and voltage regulation agent to obtain a coordinated control strategy; Specifically, the time response characteristics of the power modulation control parameters are analyzed. By setting a preset time threshold, the parameters with response time less than the threshold are classified as fast power regulation parameters, which are related to the short-term dynamic response of the power grid; and the parameters with response time greater than the threshold are classified as voltage stability parameters, which reflect the long-term stability requirements of the system. Through this step, dual-time scale control variables are formed. The dual-time scale control variables are input into the time scale separation module, and the fast and slow dynamics of the system are decoupled using the singular perturbation method. The singular perturbation method introduces fast and slow time variables to extract fast dynamic and slow dynamic characteristics respectively, thereby decomposing the complex system into fast subsystems and slow subsystems. The hierarchical control structure can independently handle control tasks at different time scales and construct a fast and slow time scale coordinated control framework. In this framework, the fast subsystem is responsible for real-time response to the power grid dynamics, and the slow subsystem is responsible for maintaining the long-term voltage stability of the power grid. A reinforcement learning environment is designed for the fast power regulation parameters and voltage stability parameters in the fast and slow time scale coordinated control framework. In the fast subsystem, the state-action space of the power regulation agent is constructed with charging power and power factor as state variables and PWM modulation depth as action variable. This design enables the agent to autonomously adjust the modulation depth under different power demand scenarios, thereby optimizing the dynamic response of power output. In the slow subsystem, the state-action space of the voltage regulation agent is constructed with grid voltage and voltage deviation as state variables and compensation coefficient as action variable. This design enables the agent to dynamically adjust the compensation parameters to ensure that the grid voltage remains stable in the long term. The reinforcement learning environment of the power regulation agent and the voltage regulation agent is input into the reward function construction module, and the multi-agent reward function is designed based on the control objective of the system. In the fast subsystem, the design of the reward signal is based on the power tracking error, and the size of the reward signal directly reflects the degree of deviation between the output power and the target power; in the slow subsystem, the design of the reward signal mainly focuses on voltage stability, aiming to encourage the agent to minimize the voltage deviation through appropriate control actions. By combining the reward signals of the two subsystems, a joint multi-agent reward function is obtained. After the reward function is constructed, the deep deterministic policy gradient algorithm is used to train the power regulation agent and the voltage regulation agent in parallel. The deep deterministic policy gradient algorithm is a reinforcement learning algorithm that can handle continuous action space. It generates control strategies by using the policy network and evaluates the strategy by combining the value network to gradually optimize the policy network parameters of the agent. During the training process, the agent learns the ability to quickly respond to grid fluctuations and maintain long-term voltage stability through continuous interaction and self-optimization, forming an optimized control strategy. The optimized control strategy is input into the coordination mechanism design module, and a multi-agent coordination mechanism is constructed based on the Nash equilibrium.The introduction of Nash equilibrium ensures that the strategy selection between the power regulation agent and the voltage regulation agent is compatible with each other, which can meet the needs of rapid response and ensure long-term stability. Through this mechanism, a distributed collaborative decision-making network is formed, which can coordinate the control framework according to the fast and slow time scales and perform power regulation and voltage regulation tasks in a hierarchical control structure. The collaborative control strategy dynamically adapts to the complex power grid operation environment through multi-agent collaboration, while meeting the charging needs of electric vehicles and the multiple goals of power grid operation, and obtains a collaborative control strategy.

[0017] The dynamic characteristics of the fast power regulation parameters and voltage stability parameters in the dual-time scale control variables are analyzed. By analyzing the dynamic behavior of the charging power change rate and power factor, they are set as fast state variables. These variables reflect the fast-response dynamic characteristics of the system. At the same time, the voltage change rate and voltage deviation reflect the slow dynamic characteristics of the system and are set as slow state variables. The fast and slow variables are combined to construct a state matrix containing mixed time scales. The state variables and control variables in the mixed time scale state matrix are introduced into the time scale parameter ε to construct the state space equation in the standard singular perturbation form. The singular perturbation form divides the system dynamics into fast variables and slow variables by clarifying the role of the time scale parameter ε, forming a dynamic model of the singular perturbation system, and describing the coupling relationship between charging power regulation and voltage stability control at different time scales. The fast variable boundary layer analysis is performed on the dynamic model of the singular perturbation system. By setting the time scale parameter ε to zero, the quasi-steady-state equation of the fast variable is solved to obtain the boundary layer system equation. The boundary layer system equation describes the dynamic behavior of the fast variable, and its solution results reveal the stability characteristics of the system on the fast time scale. In order to verify the stability of fast variables, the boundary layer system equation is input into the stability analysis module, and a quadratic Lyapunov function with charging power and power factor as variables is designed. By calculating the derivative of the Lyapunov function and judging its sign, the stability condition of the boundary layer system is obtained, and the stability constraint of the fast subsystem is formed. At the same time, the slow variables in the dynamic model of the singular perturbation system are simplified and analyzed. By ignoring the influence of fast variables, the simplified system equation of the slow variable is obtained. This equation only contains slow dynamic characteristics and can effectively reflect the behavioral characteristics of the system on a longer time scale. Similarly, the simplified system equation is input into the stability analysis module, and a quadratic Lyapunov function with voltage change rate and voltage deviation as variables is designed. By analyzing the derivative of the Lyapunov function and judging its sign, the stability condition of the simplified system is obtained, and the stability constraint of the slow subsystem is formed. Based on the stability constraint of the fast subsystem and the stability constraint of the slow subsystem, a Lyapunov function of the fast and slow variables is constructed. Through this composite function, the dynamic characteristics of the fast variables and the slow variables are considered at the same time, and the stability of the system is analyzed as a whole. By performing negative qualitative analysis on the derivative of the composite Lyapunov function, the overall stability criterion of the hierarchical control system is derived, and the conditions under which the system can maintain stable operation after fast and slow dynamic decoupling are clarified. The stability criterion of the hierarchical control system is input into the control framework design module, and the fast power regulation controller and the slow voltage regulation controller are hierarchically designed based on the fast and slow time scales. In the fast subsystem, the controller needs to respond quickly to changes in charging power demand and dynamically adjust the PWM modulation depth to achieve power regulation; while in the slow subsystem, the controller stabilizes the grid voltage by optimizing compensation parameters.Through hierarchical design, the two types of controllers are ensured to work independently on different time scales, and at the same time, overall synergy is achieved through a coordination mechanism, resulting in a coordinated control framework for fast and slow time scales.

[0018] S5, checking the charging power limit constraint, voltage limit constraint and power factor constraint of the coordinated control strategy, obtaining the constraint violation judgment result, and performing constraint projection calculation to obtain the target control instruction; Among them, the charging power value output by the collaborative control strategy is compared and judged in real time with the maximum allowable charging and discharging power of the on-board energy storage system. The charging and discharging capacity of the on-board energy storage system is affected by multiple factors such as the capacity, health state (SOH) and temperature of the battery, and its maximum allowable charging and discharging power data changes dynamically. During the comparison process, if the output charging power value of the collaborative control strategy exceeds the maximum charging and discharging power range allowed by the on-board energy storage system, the power amount exceeding the range is recorded as the charging power limit constraint violation data, which directly reflects the overload degree of power output during system operation. At the same time, the grid voltage value output by the collaborative control strategy is calculated by ratio with the nominal voltage of the grid to detect whether the voltage exceeds the specified range. The nominal voltage is the reference voltage for normal operation of the grid, such as 220V or 110V. The stability of the grid voltage is evaluated by calculating the deviation ratio of the actual output voltage to the nominal voltage. When the ratio exceeds the preset range, the corresponding voltage limit constraint violation data will be recorded. These data are used to determine whether the charging pile has caused voltage fluctuations or other adverse effects on the grid. At the same time, a threshold judgment is made on the power factor value output by the collaborative control strategy. The power factor reflects the ratio between the active and reactive components of the system power, which is crucial to the efficiency and stability of the power system. If the power factor is lower than the set lower limit, it means that the reactive power component of the system is too large, resulting in increased grid losses or decreased stability. In this case, the power factor constraint violation data is recorded to evaluate the impact of system operation on the grid. After completing the above constraint checks, the charging power limit constraint violation data, the voltage limit constraint violation data, and the power factor constraint violation data are input into the constraint determiner. The constraint violation degree is calculated by the weighted summation method, in which the weight of each constraint type is dynamically adjusted according to its importance to the system operation. The result of the weighted calculation reflects the total amount of constraint violations, and the violation level is divided by normalizing the total amount. When the normalized constraint violation total amount exceeds the preset threshold, the constraint processing process is triggered to obtain the constraint violation judgment result. This result is used as a reference for subsequent projection correction to ensure that the correction process is more targeted. The constraint violation judgment result is input into the constraint projection calculation module to correct the output control quantity of the collaborative control strategy. Based on the charging power limit constraint violation data, voltage limit constraint violation data and power factor constraint violation data, a constrained convex set is constructed as the projection space. The constrained convex set geometrically describes the range of all control outputs that meet the constraints, and its shape is determined by the boundaries of the constraints. Through this projection space, the control output is effectively restricted to operate within a safe range. In the projection correction stage, the output control quantity of the collaborative control strategy is orthogonally projected in the constrained projection space. The orthogonal projection projects the original control quantity onto the surface or interior of the constrained convex set to ensure that the corrected control quantity fully meets all constraints. The result generated by the projection operation is the projection corrected control quantity, which represents the optimal control output under the constraints.The projection correction control quantity is converted into the PWM modulation depth and phase angle parameters required for charging pile power regulation through coordinate transformation. These parameters are command signals that directly act on the charging pile hardware and can achieve precise control of the power converter.

[0019] S6, converting the target control instruction into a power regulation signal through the PWM modulator, and controlling the power conversion circuit of the charging pile to perform the power regulation operation.

[0020] Specifically, the target control instruction is converted into a power regulation signal through a PWM modulator. PWM modulation (pulse width modulation) is a technology that controls the average output voltage by adjusting the pulse width, which is used for power management and energy conversion. The target control instruction is input into the PWM modulator for signal preprocessing. This includes normalization of the modulation depth and phase angle to ensure that the input signal is within the allowable range of the PWM modulator. A carrier signal is generated according to the target frequency. The carrier signal is a triangular wave or a sawtooth wave, and its frequency determines the switching frequency of the PWM. A PWM signal sequence is generated by comparing the carrier signal with the modulation signal in the target control instruction. In each PWM signal cycle, the length of the pulse width is determined by the modulation depth, and the time distribution of the pulse is related to the target phase angle. In this way, the PWM signal can accurately express the power regulation requirements in the target control instruction. The generated PWM signal is used as a driving signal for the switching device in the power conversion circuit of the charging pile. The power conversion circuit of the charging pile is composed of switching devices such as IGBT (insulated gate bipolar transistor) or MOSFET (metal oxide semiconductor field effect transistor). The PWM signal controls the on-off state of these switching devices, thereby adjusting the magnitude and frequency of the DC voltage or AC voltage in the circuit. Taking the DC / DC converter as an example, the PWM signal controls the conduction time of the IGBT, thereby adjusting the average value of the output voltage and realizing precise control of the charging power of the electric vehicle battery; in the DC / AC inverter, the PWM signal can also adjust the frequency and waveform of the output voltage, so that the charging pile can adapt to different grid and load requirements. In the control process, in order to ensure the safe and reliable operation of the power conversion circuit of the charging pile, the PWM modulator introduces a protection mechanism. For example, the current and voltage of the switching device are monitored in real time through the detection circuit. If an abnormal situation such as overcurrent or overvoltage is detected, the PWM modulator will immediately adjust the output signal, shorten the pulse width or temporarily turn off the switching device to avoid damage to the power conversion circuit. At the same time, the dead time compensation technology is used to prevent the upper and lower bridge arm switching devices from being turned on at the same time to cause a short circuit. As the PWM signal drives the power conversion circuit to perform power regulation operations, the output voltage and current will directly act on the battery system of the electric vehicle, thereby completing the charging process. Through the current sensor and voltage sensor, the actual value of the output power is monitored in real time and compared with the target control instruction. If a deviation is found, the PWM modulator will dynamically adjust the pulse width or phase angle according to the feedback signal to compensate for the error and ensure that the output power is consistent with the target.

[0021] In one example, multi-scale data collection and preprocessing are performed on the V2G AC charging system to obtain a standardized system status data set, including: The three-phase voltage data of the AC power grid in the V2G AC charging system is collected to obtain the original voltage data on the grid side, and the three-phase current data of the charging circuit is collected to obtain the original current data on the grid side; Perform power calculation and Fourier transformation on the grid side voltage raw data and grid side current raw data to obtain power factor calculation data; Perform zero drift compensation on the grid side voltage raw data, grid side current raw data and power factor calculation data to obtain filtered grid side data; Obtain battery status data from the electric vehicle battery management system through the CAN bus at a preset communication cycle, extract the battery state of charge, maximum allowable charging power and maximum allowable discharging power, and obtain the operating parameters of the on-board energy storage system; The filtered grid-side data and the operating parameters of the on-board energy storage system are standardized to obtain normalized state data, and outlier detection and data reconstruction are performed on the normalized state data to obtain a standardized system state data set.

[0022] In this example, voltage transformers (VT) and current transformers are installed in the charging system to collect voltage and current data on the grid side. , convert the three-phase voltage and current signals of the power grid into signals suitable for processing by the data acquisition system. Assume that the collected three-phase voltage data of the power grid is The collected three-phase current data of the power grid is These data change with time in the time domain, reflecting the operating status of the power grid. The original data is converted into digital form and enters the data acquisition module to obtain discrete voltage and current data. The collected voltage and current data are used to calculate power. Instantaneous active power in a three-phase system and reactive power Calculated by the following formula: ; ; in, Indicates the instantaneous active power, Represents instantaneous reactive power. The calculation results are used to further analyze the power factor of the system. In order to analyze the harmonic components and power factor in the power grid, the voltage and current data are Fourier transformed. Assume that the fundamental component of the three-phase voltage is , the fundamental component of the three-phase current is , the expression of Fourier transform is: ; in, is the fundamental period, is the angular frequency. By extracting the fundamental component, the power factor of the system is calculated : ; This result is used to evaluate the energy efficiency of the power grid. After obtaining data such as voltage and power factor, the original data is compensated for zero drift. Zero drift is a systematic error caused by sensor offset or sampling error, which will affect the accuracy of the data. The formula for zero drift compensation is as follows: ; in, and are the average offsets of voltage and current respectively. After compensation, the filtered grid-side data is obtained. At the same time, in order to obtain operating parameters from the on-board energy storage system, the CAN bus is used to read the data of the battery management system (BMS) at a preset communication cycle. Through communication instructions, the battery state of charge (SOC), maximum allowable charging power and the maximum allowable discharge power For example, SOC represents the ratio of the remaining capacity of the battery to the total capacity and is defined as: ; in, and Represent the current remaining capacity and total capacity of the battery respectively. After integrating the filtered grid-side data and the on-board energy storage system operating parameters, they are standardized to eliminate the dimensional differences of different physical quantities. The standardization formula is as follows: ; in, Represents the original data, and are the minimum and maximum values ​​of the data, respectively. The standardized data is normalized to the range of [0, 1] to facilitate subsequent calculations and analysis. Outlier detection and data reconstruction are performed on the normalized state data. Outlier detection uses statistical methods, such as setting 3 times the standard deviation as the threshold, and the detection formula is: ; in, is the data mean, is the standard deviation. If an outlier is detected, it is repaired by interpolation or a prediction model based on historical data to obtain a standardized system status data set.

[0023] In one example, a state space model including grid voltage characteristics and charging power characteristics is constructed based on a standardized system state data set, and the state space model is verified to obtain a system dynamic model, including: The standardized system status data set is divided proportionally according to the time series to obtain a training set and a validation set containing grid voltage data, current data, power factor data, and vehicle energy storage system data; Based on the training set, polynomial fitting is performed on the grid voltage data, current data, power factor data and on-board energy storage system data to obtain a continuous state vector sequence; Perform sparse feature extraction on the continuous state vector sequence to obtain a sparse coefficient matrix, and then perform truncation on the sparse coefficient matrix to obtain the nonlinear dynamic characteristics of the system; Based on the nonlinear dynamic characteristics of the system, the state space is constructed and dimensionally reduced to obtain the state space characteristics after dimension reduction, and the state is predicted by the unscented Kalman filter to obtain the state estimation result; The state transition probability matrix is ​​calculated according to the state estimation results, and the transition probability parameters are optimized by maximum likelihood estimation to obtain the state space model; The state prediction is performed on the verification set, the root mean square error and relative error between the predicted state and the actual state are calculated, the model verification index is obtained, and the model verification index is compared with the preset error threshold. When the root mean square error is less than the first target value and the relative error is less than the second target value, the state space model is output as the system dynamic model.

[0024] In this example, the standardized system status dataset is divided into a training set and a validation set according to the time series. Assume that the dataset contains Time series samples, data points include grid voltage data , current data , power factor data and on-board energy storage system data (such as battery state of charge , Maximum allowable charge and discharge power ). To ensure that the model training and verification have good generalization performance, the proportion and Divide the data set. The training set data is recorded as , the validation set data is the remaining . Perform polynomial fitting on the multivariate time series data in the training set, and transform the discrete standardized data points into a continuous state vector sequence. The fitting process uses the least squares method to solve the polynomial coefficients. For each variable (like or ), the fitting formula is: ; in, is the order of the polynomial, is the fitting coefficient, is the time index. By solving , get the fitted polynomial function , for example, the fitted grid voltage state vector is expressed as In this way, each variable data in the training set is converted into a continuous state vector sequence. The continuous state vector sequence is processed using the sparse feature extraction method to extract the most representative features. Suppose the state vector sequence is , projecting it into a sparse coefficient matrix By optimizing the objective function: ; in, is the basis matrix, is the regularization coefficient, which is used to control the sparsity. By solving this problem, we can obtain the sparse coefficient matrix , and truncate the matrix to retain only the sparse coefficients with large feature contributions, and obtain the nonlinear dynamic characteristics of the system. Based on the nonlinear dynamic characteristics, the state space is constructed and the dimension is reduced. Assume that the state characteristics after dimension reduction are , whose dimension is much smaller than that of the original state vector, is obtained through principal component analysis or other dimensionality reduction algorithms. The dimensionality reduction formula is: ; in, is the dimension reduction transformation matrix, which is determined by maximizing feature variance or minimizing information loss. State features after dimension reduction Represents the main dynamic behavior of the system. In order to predict the future changes of state characteristics, the unscented Kalman filter is used for state estimation. Assume that the state space model after dimensionality reduction is: ; ; in, is the state transfer matrix, is the control input, is the process noise, is the observed quantity, is the observation noise. The state estimate is recursively updated by the unscented Kalman filter algorithm , obtain accurate state prediction. According to the state estimation results, calculate the state transition probability matrix .set up Elements Indicates that the system is in state Transfer to state The probability of Values: ; in, and represents the partition interval of the state space, is the indicator function. The constructed state space model is validated using the validation set. By comparing the predicted state The actual status , calculate the root mean square error (RMSE): ; And the relative error: ; When the RMSE is less than the preset target value and the relative error is less than the second target value, the state space model is considered to have achieved the expected accuracy and the output is the system dynamic model.

[0025] In one example, the charging power instruction in the system dynamic model is decomposed to obtain a reference power component and a compensation power component, and a power output mapping relationship is calculated to obtain a power modulation control parameter, including: Perform discrete Fourier transform on the grid voltage state data in the system dynamic model, and obtain the real-time grid state characteristics by calculating the amplitude and phase angle of the fundamental component; The real-time grid state characteristics are input into the power command decomposer, the reference power correction coefficient is calculated according to the deviation rate between the voltage amplitude and the rated voltage, the charging power command is corrected, and the reference power component is obtained; Perform wavelet decomposition on the power dynamic response data in the system dynamic model, extract high-frequency disturbance components and low-frequency steady-state components, obtain dynamic characteristic components, and input the dynamic characteristic components into the compensator, calculate the dynamic compensation value through the proportional-integral-differential control algorithm, and obtain the compensated power component; The reference power component and the compensation power component are linearly superimposed, and the initial power mapping function is constructed by piecewise linear interpolation. The initial power mapping function is segmented optimized, and the optimal mapping coefficient of each segment is calculated to obtain the segmented power mapping relationship. The segmented power mapping relationship is input into the modulator parameter calculation module, and the corresponding PWM modulation depth and phase angle are calculated according to the power command value to obtain the modulation control reference parameters. The modulation control reference parameters are then compensated for dead time and switching loss to obtain the power modulation control parameters.

[0026] In this example, grid voltage status data is obtained from the system dynamics model , perform discrete Fourier transform on it to extract the amplitude and phase angle of the fundamental component and obtain the real-time grid status characteristics. The formula for discrete Fourier transform is as follows: ; ; ; in, It is the complex form of the fundamental component of the three-phase voltage, including amplitude and phase angle; is the number of sampling points, is the spectral index corresponding to the fundamental frequency. Extract the magnitude from the complex form and phase angle , these values ​​reflect the real-time voltage status of the power grid. For example, if at a certain moment V, , which means that the amplitude of the fundamental wave of the grid A phase voltage is 220V and the phase angle is 0 degrees. The real-time grid state characteristics are input into the power command decomposer, and the reference power correction coefficient is calculated according to the deviation rate between the voltage amplitude and the rated voltage. , to correct the charging power instruction , get the reference power component The correction coefficient is calculated as follows: ; in, is the rated voltage. The corrected reference power component is: ; Power dynamic response data in the system dynamic model Perform wavelet decomposition to extract high-frequency disturbance components and low-frequency steady-state components to obtain dynamic characteristic components The wavelet decomposition formula is as follows: ; in, is the high frequency wavelet basis function, is the low frequency scaling function, and is the wavelet coefficient. High frequency disturbance , low frequency steady-state component . The dynamic characteristic component Input compensator, which calculates dynamic compensation value through proportional-integral-derivative (PID) control algorithm : ; in, is the error, are the proportional, integral and differential gains of the PID controller respectively. and compensation power components Perform linear superposition to obtain the total power instruction: ; To further optimize the output power, the initial power mapping function is constructed by piecewise linear interpolation. Assume that the output power range is divided into intervals, each interval corresponds to a linear mapping function , optimize the coefficients by the least squares method , and calculate the optimal mapping coefficient of each segment to obtain the segmented power mapping relationship. The segmented power mapping relationship is input into the modulator parameter calculation module, and the PWM modulation depth is calculated according to the power command value. and phase angle : ; The PWM signal is corrected by dead time compensation. The compensation formula is: ; in, is the dead time, It is the function of the influence of dead time on modulation depth. Similarly, the switching loss compensation reduces energy loss by modifying the modulation parameters and finally generates the power modulation control parameters.

[0027] In one example, the power modulation control parameters are divided into fast power regulation parameters and voltage stability parameters, a fast and slow time scale coordinated control framework is established, and a power regulation agent and a voltage regulation agent are designed to obtain a coordinated control strategy, including: The time response characteristics of the power modulation control parameters are analyzed, and the parameters whose response time is less than the preset time threshold are classified as fast power regulation parameters, and the parameters whose response time is greater than the preset time threshold are classified as voltage stability parameters, so as to obtain dual-time scale control variables; The dual-time-scale control variables are input into the time-scale separation module, and the fast and slow dynamics are decoupled by the singular perturbation method. A hierarchical control structure including fast and slow subsystems is established to obtain a fast and slow time-scale coordinated control framework. A state-action space is established for the fast power regulation parameters in the fast and slow time scale coordinated control framework, charging power and power factor are set as state variables, PWM modulation depth is set as action variable, and a reinforcement learning environment for the power regulation agent is obtained; A state-action space is established for the voltage stability parameters in the fast and slow time scale coordinated control framework, and the grid voltage and voltage deviation are set as state variables and the compensation coefficient is set as the action variable to obtain the reinforcement learning environment of the voltage regulation agent. Input the reinforcement learning environment of the power regulation agent and the reinforcement learning environment of the voltage regulation agent into the reward function building module, design the reward signal based on the power tracking error and voltage stability, and obtain the multi-agent reward function; The power regulation agent and the voltage regulation agent are trained in parallel based on the multi-agent reward function, and the policy network parameters are updated using a deep deterministic policy gradient algorithm to obtain the optimized control strategy. The optimized control strategy is input into the coordination mechanism design module, and a multi-agent coordination mechanism based on Nash equilibrium is constructed to obtain a distributed collaborative decision-making network. Based on the distributed collaborative decision-making network, power regulation and voltage regulation tasks are performed through a hierarchical control structure to obtain a collaborative control strategy.

[0028] In this example, the time response characteristics of the power modulation control parameters are analyzed. Assume that the response time of the control parameters in the system is , estimate the time constant of each parameter through experiments or modeling. The parameters of are defined as fast power regulation parameters, such as PWM modulation depth and instantaneous power factor compensation coefficient. These parameters are used to quickly respond to grid frequency fluctuations or short-term load changes. Parameters with response time greater than the threshold are classified as voltage stability parameters, such as slow grid voltage compensation coefficient. These parameters mainly regulate the steady-state behavior of the system to ensure long-term stability. By dividing, the set of fast and slow time scale control variables is obtained and The dual time scale control variables are input into the time scale separation module, and the fast and slow dynamics are decoupled by the singular perturbation method. (usually represents the time scale ratio between fast variables and slow variables) decomposes the system into fast subsystems and slow subsystems. Assume that the state equation of the system is: ; in By introducing the time scale transformation , we get the boundary layer equation of the fast variable: ; And the simplified equation for the slow variable: ; By decoupling, a hierarchical control structure of the fast subsystem and the slow subsystem is established to form a fast and slow time scale coordinated control framework. In the fast subsystem, a state action space is established. Assume that the charging power and power factor is the state variable, PWM modulation depth is the action variable. The state space can be expressed as: ; The action space is: ; By defining the state transfer function and reward function, a reinforcement learning environment for the power regulation agent is constructed. In the slow subsystem, a state action space is established. and voltage deviation is the state variable, the voltage compensation coefficient is the action variable. The state space is: ; The action space is: ; Similarly, the reinforcement learning environment of the voltage regulation agent is constructed by defining appropriate state transfer functions and reward functions. The two reinforcement learning environments are input into the reward function building module to design the multi-agent reward function based on power tracking error and voltage stability. The reward signal of the fast subsystem is defined as: ; in is the target charging power, is the target power factor, is the weight coefficient. The reward signal of the slow subsystem is defined as: ; in is the target voltage value, is the weight coefficient. Based on the multi-agent reward function, the deep deterministic policy gradient algorithm (DDPG) is used to train the power regulation agent and the voltage regulation agent in parallel. DDPG uses a policy network and a value network to generate actions and evaluate the value of actions respectively. By updating the policy network parameters and value network parameters To optimize the strategy. The update formula is: ; ; in and is the learning rate, is the value function, It's a reward. is the discount factor. After training, the optimized control strategy is input into the coordination mechanism design module, and the distributed collaborative decision network is obtained by constructing a multi-agent coordination mechanism through Nash equilibrium. Under Nash equilibrium, the strategy of each agent reaches the optimal value given the strategies of other agents, thereby ensuring the global coordination of power regulation and voltage regulation tasks. Through the distributed collaborative decision network, the hierarchical control structure based on fast and slow time scales executes power regulation and voltage regulation tasks, and generates the final collaborative control strategy.

[0029] In one example, the dual-time-scale control variables are input into the time-scale separation module, the fast and slow dynamics are decoupled by the singular perturbation method, and a hierarchical control structure including fast and slow subsystems is established to obtain a fast and slow time-scale coordinated control framework, including: The dynamic characteristics of the fast power regulation parameters and voltage stability parameters in the dual-time-scale control variables are analyzed, the charging power change rate and power factor are set as fast state variables, the voltage change rate and voltage deviation are set as slow state variables, and the mixed time-scale state matrix is ​​obtained; The state variables and control variables in the mixed time scale state matrix are introduced into the time scale parameter ε, the state space equation of the standard singular perturbation form is constructed, and the dynamic model of the singular perturbation system is obtained; The fast variable boundary layer analysis is performed on the dynamic model of the singular perturbation system. The boundary layer system equation is obtained by solving the quasi-steady-state equation by setting the time scale parameter ε to zero. Input the boundary layer system equation into the stability analysis module, design a quadratic Lyapunov function with charging power and power factor as variables, calculate the function derivative to determine the stability of the boundary layer, and obtain the stability constraint conditions of the fast subsystem; Simplify the slow variables of the dynamic model of the singular perturbation system to obtain the simplified system equation, and input the simplified system equation into the stability analysis module. Design the quadratic Lyapunov function with the voltage change rate and voltage deviation as variables, calculate the function derivative to determine the stability of the simplified system, and obtain the stability constraint conditions of the slow subsystem. Based on the stability constraints of the fast subsystem and the slow subsystem, a composite Lyapunov function of fast and slow variables is constructed, and the negative qualitative analysis of the function derivative is performed to obtain the stability criterion of the hierarchical control system. The stability criterion of the hierarchical control system is input into the control framework design module, and the fast power regulation controller and the slow voltage regulation controller are hierarchically designed to obtain a fast and slow time scale coordinated control framework.

[0030] In this example, the dynamic characteristics of the dual time scale control variables are analyzed. Fast power regulation parameters such as charging power change rate and power factor It mainly reflects the system's ability to respond to short-term dynamic changes and is set as a fast state variable; while the voltage change rate and voltage deviation It is mainly related to the long-term stability of the power grid and is set as a slow state variable. The fast state variables are set to form a vector , the slow state variables form a vector , and combined with the control variables Construct the mixing time-scale state matrix: ; in, are fast control variables (such as PWM modulation depth), is a slow control variable (such as voltage compensation coefficient). The state variables and control variables are introduced into the time scale parameter (representing the time scale ratio between fast and slow dynamics), constructing the state space equation in the form of standard singular perturbation, and obtaining the dynamic model of the singular perturbation system: ; ; in, and Describe the dynamic behavior of fast variables and slow variables respectively. Perform fast variable boundary layer analysis on the dynamic model of singular perturbation system by changing the time scale parameter Set to zero and solve the quasi-steady-state equations for the fast variables: ; This equation represents the condition under which the fast variable reaches equilibrium when the slow variable remains unchanged, forming the boundary layer system equation. The boundary layer system equation is input into the stability analysis module to design a quadratic Lyapunov function with charging power and power factor as variables. : ; in, is a symmetric positive definite matrix used to measure the energy of the fast subsystem. By calculating the derivative of the Lyapunov function : ; Determine whether its sign is negative definite, so as to determine the stability constraint of the fast subsystem. Simplify the dynamic equation of the slow variable part, ignore the high-frequency dynamic influence of the fast variable, and obtain the simplified system equation of the slow variable: ; Input the simplified system equation into the stability analysis module and design a quadratic Lyapunov function with voltage change rate and voltage deviation as variables : ; in, is a symmetric positive definite matrix. Similarly, by calculating And judge its negative definiteness, and get the stability constraint of the slow subsystem. Based on the stability constraint of the fast subsystem and the slow subsystem, construct the Lyapunov function of the fast and slow variables : ; Derivatives of composite Lyapunov functions To perform negative qualitative analysis: ; By judging The overall stability criterion of the hierarchical control system is obtained by taking the stability criterion of the hierarchical control system into consideration. The stability criterion of the hierarchical control system is input into the control framework design module, and a fast power regulation controller and a slow voltage regulation controller are designed respectively to obtain a fast and slow time scale coordinated control framework. In the fast subsystem, the controller responds quickly to the power command by adjusting the PWM modulation depth; in the slow subsystem, the controller maintains the grid voltage stability by adjusting the voltage compensation coefficient.

[0031] In one example, the collaborative control strategy is checked for charging power limit constraints, voltage limit constraints, and power factor constraints, a constraint violation judgment result is obtained, and constraint projection calculation is performed to obtain a target control instruction, including: The charging power value output by the collaborative control strategy is compared and judged in real time with the maximum allowable charging and discharging power data of the on-board energy storage system. When the charging power value exceeds the maximum charging and discharging power limit, the power excess is recorded to obtain the charging power limit constraint violation data; The grid voltage value output by the cooperative control strategy is calculated by ratio with the grid nominal voltage to obtain the voltage limit constraint violation data; Perform threshold judgment on the power factor value output by the collaborative control strategy to obtain power factor constraint violation data, and input the charging power limit constraint violation data, voltage limit constraint violation data and power factor constraint violation data into the constraint determiner, calculate the constraint violation degree through weighted summation, and obtain the total constraint violation amount; Normalize the total amount of constraint violations, and classify the violation levels based on the total amount. When the normalized total amount of constraint violations exceeds the threshold, the constraint processing is triggered to obtain the constraint violation judgment result. The constraint violation judgment result is input into the constraint projection calculation module, and a constraint convex set is constructed based on the charging power limit constraint violation data, the voltage limit constraint violation data, and the power factor constraint violation data to obtain the constraint projection space; The output control quantity of the collaborative control strategy is orthogonally projected in the constrained projection space to obtain the projection correction control quantity, which is then converted into PWM modulation depth and phase angle parameters through coordinate transformation to obtain the target control instruction.

[0032] In this example, the charging power value output by the cooperative control strategy The maximum allowable charging and discharging power data of the on-board energy storage system is compared and judged in real time. Assume that the maximum allowable charging power provided by the on-board energy storage system is and the maximum allowable discharge power is .when or When the recording power exceeds the limit : ; This result is the charging power limit constraint violation data, which is used to evaluate the constraint compliance of the system power output. With grid nominal voltage Perform ratio calculation to obtain voltage limit constraint violation data The formula is as follows: ; This data reflects the degree of deviation of the grid voltage. If it is larger, it means that the grid voltage exceeds the allowable range. Perform threshold judgment and set the target power factor threshold as .if , then record the power factor constraint violation data : ; The power factor violation data quantifies the degree of deviation of the impact of charging pile operation on the reactive power of the grid. Input constraint judgement, calculate the total constraint violation degree by weighted summation : ; in is the weight of each constraint, reflecting the importance of different constraints. It is a quantitative indicator of the total violation degree. Perform normalization and set the normalization factor to , the normalized violation degree is: ; in is the maximum allowed violation level. The levels are divided according to the normalized total violation amount. threshold, the constraint processing flow is triggered and the constraint violation judgment result is output otherwise In the constraint processing stage, the constraint violation judgment results are input into the constraint projection calculation module. Constructing a Constrained Convex Set , defined as the control output range that satisfies all constraints: ; Control quantity of the cooperative control strategy output Orthogonal projection to , the projection formula is: ; The projection correction control amount is obtained by projection operation , and convert it into PWM modulation depth through coordinate transformation and phase angle : ; in and It is a predefined mapping function that generates PWM parameters based on the target values ​​of power and voltage. The resulting PWM depth and phase angle The target control command is input into the modulator to achieve precise control of the power conversion circuit.

[0033] Reference Figure 2 , this embodiment provides an AC charging power adjustment device based on V2G, including: Acquisition module 1, used for multi-scale data acquisition and preprocessing of the V2G AC charging system to obtain a standardized system status data set; Construction module 2 is used to construct a state space model including grid voltage characteristics and charging power characteristics based on a standardized system state data set, and verify the state space model to obtain a system dynamic model; Decomposition module 3, used for decomposing the charging power instruction in the system dynamic model to obtain the reference power component and the compensation power component, and calculating the power output mapping relationship to obtain the power modulation control parameter; Design module 4, used to divide the power modulation control parameters into fast power regulation parameters and voltage stability parameters, establish a fast and slow time scale coordinated control framework, and design the power regulation agent and the voltage regulation agent to obtain a coordinated control strategy; The checking module 5 is used to check the charging power limit constraint, the voltage limit constraint and the power factor constraint of the coordinated control strategy, obtain the constraint violation judgment result, and perform the constraint projection calculation to obtain the target control instruction; The control module 6 is used to convert the target control instruction into a power regulation signal through a PWM modulator, and control the power conversion circuit of the charging pile to perform a power regulation operation.

[0034] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0035] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0036] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0037] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0038] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0039] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0040] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A V2G-based AC charging power regulation method, characterized in that: The following steps are involved: Perform multi-scale data collection and preprocessing on the V2G AC charging system to obtain a standardized system status data set; Building a state space model including grid voltage characteristics and charging power characteristics based on the standardized system state data set, and verifying the state space model to obtain a system dynamic model; Decomposing the charging power instruction in the system dynamic model to obtain a reference power component and a compensation power component, and calculating the power output mapping relationship to obtain a power modulation control parameter; The power modulation control parameters are divided into fast power regulation parameters and voltage stability parameters, a fast and slow time scale coordinated control framework is established, and a power regulation agent and a voltage regulation agent are designed to obtain a coordinated control strategy; Checking the charging power limit constraint, the voltage limit constraint and the power factor constraint of the collaborative control strategy to obtain a constraint violation judgment result, and performing constraint projection calculation to obtain a target control instruction; The target control instruction is converted into a power regulation signal through a PWM modulator, and the power conversion circuit of the charging pile is controlled to perform a power regulation operation.

2. The V2G-based AC charging power regulation method according to claim 1, characterized in that: The multi-scale data collection and preprocessing of the V2G AC charging system is performed to obtain a standardized system status data set, including: The three-phase voltage data of the AC power grid in the V2G AC charging system is collected to obtain the original voltage data on the grid side, and the three-phase current data of the charging circuit is collected to obtain the original current data on the grid side; Performing power calculation and Fourier transformation on the grid-side voltage original data and the grid-side current original data to obtain power factor calculation data; Performing zero drift compensation on the grid-side voltage original data, the grid-side current original data and the power factor calculation data to obtain filtered grid-side data; Obtain battery status data from the electric vehicle battery management system through the CAN bus at a preset communication cycle, extract the battery state of charge, maximum allowable charging power and maximum allowable discharging power, and obtain the operating parameters of the on-board energy storage system; The filtered grid-side data and the operating parameters of the on-board energy storage system are standardized to obtain normalized state data, and outlier detection and data reconstruction are performed on the normalized state data to obtain a standardized system state data set.

3. The V2G-based AC charging power regulation method according to claim 2, characterized in that: The state space model including grid voltage characteristics and charging power characteristics is constructed based on the standardized system state data set, and the state space model is verified to obtain a system dynamic model, including: Proportionally dividing the standardized system status data set according to the time series to obtain a training set and a verification set including grid voltage data, current data, power factor data and vehicle-mounted energy storage system data; Based on the training set, polynomial fitting is performed on the grid voltage data, current data, power factor data and on-board energy storage system data to obtain a continuous state vector sequence; Performing sparse feature extraction on the continuous state vector sequence to obtain a sparse coefficient matrix, and performing truncation processing on the sparse coefficient matrix to obtain nonlinear dynamic characteristics of the system; Based on the nonlinear dynamic characteristics of the system, a state space is constructed and dimensionally reduced to obtain a state space characteristic after dimension reduction, and a state estimation result is obtained by performing state prediction on the state space characteristic after dimension reduction through an unscented Kalman filter; Calculating a state transition probability matrix according to the state estimation result, and optimizing the transition probability parameters by maximum likelihood estimation to obtain a state space model; Perform state prediction on the verification set, calculate the root mean square error and relative error between the predicted state and the actual state, obtain a model verification index, and compare the model verification index with a preset error threshold. When the root mean square error is less than a first target value and the relative error is less than a second target value, output the state space model as a system dynamic model.

4. The V2G-based AC charging power regulation method according to claim 3, characterized in that: The step of decomposing the charging power instruction in the system dynamic model to obtain a reference power component and a compensation power component, and calculating a power output mapping relationship to obtain a power modulation control parameter includes: Performing discrete Fourier transform on the grid voltage state data in the system dynamic model, and obtaining real-time grid state characteristics by calculating the amplitude and phase angle of the fundamental component; The real-time grid state characteristics are input into a power command decomposer, a reference power correction coefficient is calculated according to a deviation rate between the voltage amplitude and the rated voltage, and the charging power command is corrected to obtain a reference power component; Performing wavelet decomposition on the power dynamic response data in the system dynamic model, extracting high-frequency disturbance components and low-frequency steady-state components, obtaining dynamic characteristic components, and inputting the dynamic characteristic components into the compensator, calculating the dynamic compensation value through the proportional-integral-differential control algorithm, and obtaining the compensated power component; The reference power component and the compensation power component are linearly superimposed, an initial power mapping function is constructed by piecewise linear interpolation, and the initial power mapping function is segmentedly optimized, and the optimal mapping coefficient of each segment is calculated to obtain a segmented power mapping relationship; The segmented power mapping relationship is input into the modulator parameter calculation module, and the corresponding PWM modulation depth and phase angle are calculated according to the power command value to obtain the modulation control reference parameters, and the dead time compensation and switching loss compensation are performed on the modulation control reference parameters to obtain the power modulation control parameters.

5. The V2G-based AC charging power regulation method according to claim 4, characterized in that: The power modulation control parameters are divided into fast power regulation parameters and voltage stability parameters, a fast and slow time scale coordinated control framework is established, and a power regulation agent and a voltage regulation agent are designed to obtain a coordinated control strategy, including: Performing time response characteristic analysis on the power modulation control parameters, classifying the parameters whose response time is less than a preset time threshold as fast power regulation parameters, and classifying the parameters whose response time is greater than the preset time threshold as voltage stability parameters, to obtain dual-time scale control variables; The dual-time-scale control variables are input into the time-scale separation module, the fast and slow dynamics are decoupled by the singular perturbation method, a hierarchical control structure including a fast subsystem and a slow subsystem is established, and a fast and slow time-scale coordinated control framework is obtained; Establishing a state-action space for the fast power regulation parameters in the fast-slow time scale coordinated control framework, setting the charging power and power factor as state variables, and the PWM modulation depth as an action variable, to obtain a reinforcement learning environment for the power regulation agent; Establishing a state-action space for the voltage stability parameters in the fast-slow time scale coordinated control framework, setting the grid voltage and voltage deviation as state variables, and the compensation coefficient as an action variable, to obtain a reinforcement learning environment for the voltage regulation agent; Inputting the reinforcement learning environment of the power regulation agent and the reinforcement learning environment of the voltage regulation agent into a reward function construction module, designing a reward signal based on power tracking error and voltage stability, and obtaining a multi-agent reward function; Based on the multi-agent reward function, the power regulation agent and the voltage regulation agent are trained in parallel, and the policy network parameters are updated using a deep deterministic policy gradient algorithm to obtain an optimized control strategy; The optimized control strategy is input into the coordination mechanism design module, a multi-agent coordination mechanism based on Nash equilibrium is constructed, a distributed collaborative decision-making network is obtained, and based on the distributed collaborative decision-making network, power regulation and voltage regulation tasks are performed through a hierarchical control structure to obtain a collaborative control strategy.

6. The V2G-based AC charging power regulation method according to claim 5, characterized in that: The dual-time-scale control variables are input into the time-scale separation module, the fast and slow dynamics are decoupled by the singular perturbation method, a hierarchical control structure including a fast subsystem and a slow subsystem is established, and a fast and slow time-scale coordinated control framework is obtained, including: Performing dynamic characteristic analysis on the fast power regulation parameter and the voltage stability parameter in the dual-time-scale control variable, setting the charging power change rate and the power factor as fast state variables, and setting the voltage change rate and the voltage deviation as slow state variables, to obtain a mixed time-scale state matrix; Introducing the state variables and control variables in the mixed time scale state matrix into the time scale parameter ε, constructing the state space equation in the standard singular perturbation form, and obtaining the dynamic model of the singular perturbation system; Performing fast variable boundary layer analysis on the dynamic model of the singular perturbation system, solving the quasi-steady-state equation by setting the time scale parameter ε to zero, and obtaining the boundary layer system equation; The boundary layer system equation is input into the stability analysis module, a quadratic Lyapunov function with charging power and power factor as variables is designed, the function derivative is calculated to determine the stability of the boundary layer, and the stability constraint condition of the fast subsystem is obtained; Performing a slow variable simplification analysis on the dynamic model of the singular perturbation system to obtain a simplified system equation, and inputting the simplified system equation into a stability analysis module, designing a quadratic Lyapunov function with voltage change rate and voltage deviation as variables, calculating the function derivative to determine the stability of the simplified system, and obtaining the stability constraint condition of the slow subsystem; Based on the stability constraint conditions of the fast subsystem and the stability constraint conditions of the slow subsystem, a Lyapunov function of a fast and slow variable composite is constructed, and a negative qualitative analysis is performed on the function derivative to obtain a stability criterion of the hierarchical control system; The stability criterion of the hierarchical control system is input into a control framework design module, and a fast power regulation controller and a slow voltage regulation controller are hierarchically designed to obtain a fast and slow time scale coordinated control framework.

7. The V2G-based AC charging power regulation method according to claim 6, characterized in that: The checking of charging power limit constraint, voltage limit constraint and power factor constraint of the collaborative control strategy to obtain constraint violation judgment result, and performing constraint projection calculation to obtain target control instruction includes: Compare and judge the charging power value output by the collaborative control strategy with the maximum allowable charging and discharging power data of the on-board energy storage system in real time, and record the power excess when the charging power value exceeds the maximum charging and discharging power limit to obtain charging power limit constraint violation data; Calculate the ratio of the grid voltage value output by the collaborative control strategy to the grid nominal voltage to obtain voltage limit constraint violation data; Performing a threshold judgment on the power factor value output by the collaborative control strategy to obtain power factor constraint violation data, and inputting the charging power limit constraint violation data, the voltage limit constraint violation data and the power factor constraint violation data into a constraint determiner, calculating the constraint violation degree by weighted summation, and obtaining the total constraint violation amount; Normalizing the total amount of constraint violations, dividing the violation levels based on the total amount, and triggering constraint processing when the normalized total amount of constraint violations exceeds a threshold to obtain a constraint violation determination result; Inputting the constraint violation determination result into a constraint projection calculation module, constructing a constraint convex set based on the charging power limit constraint violation data, the voltage limit constraint violation data and the power factor constraint violation data, and obtaining a constraint projection space; An orthogonal projection operation is performed on the output control quantity of the collaborative control strategy in the constrained projection space to obtain a projection correction control quantity, and the projection correction control quantity is converted into PWM modulation depth and phase angle parameters through coordinate transformation to obtain a target control instruction.

8. A V2G-based AC charging power regulation device, characterized in that: For implementing the steps of the method according to any one of claims 1 to 7, the device comprises: The acquisition module is used to collect and preprocess multi-scale data of the V2G AC charging system to obtain a standardized system status data set; A construction module, configured to construct a state space model including grid voltage characteristics and charging power characteristics based on the standardized system state data set, and verify the state space model to obtain a system dynamic model; A decomposition module, used to decompose the charging power instruction in the system dynamic model to obtain a reference power component and a compensation power component, and calculate the power output mapping relationship to obtain a power modulation control parameter; A design module is used to divide the power modulation control parameters into fast power regulation parameters and voltage stability parameters, establish a fast and slow time scale coordinated control framework, and design a power regulation agent and a voltage regulation agent to obtain a coordinated control strategy; A checking module, used to check the charging power limit constraint, voltage limit constraint and power factor constraint of the collaborative control strategy, obtain constraint violation judgment results, and perform constraint projection calculation to obtain target control instructions; The control module is used to convert the target control instruction into a power regulation signal through a PWM modulator, and control the power conversion circuit of the charging pile to perform a power regulation operation.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images