AC / DC power dynamic cooperative regulation and control system for charging pile

Abstract

The invention relates to a charging pile AC / DC power dynamic cooperative regulation and control system, which belongs to the technical field of charging piles and comprises a power grid parameter acquisition module, a charging load characteristic analysis module, a coupling effect calculation module, a power distribution optimization module and a cooperative control execution module. The system can accurately distribute power to vehicles which really need and can be efficiently utilized, mutual interference between modules can be effectively restrained, power oscillation and cascading failures are avoided, the stability and reliability of the whole charging station under high-density operation are guaranteed, the robustness of the system is enhanced, and the service life of the charging station is prolonged. A dynamic and globally optimal balance point is found and maintained among three mutually restricted targets, namely, power deviation, coupling power and power grid constraint, by establishing and solving a multi-target optimization model containing power deviation, coupling power and power grid constraint, the stability of the multi-target optimization model is guaranteed, and the power grid safety is maintained, which cannot be achieved by any single-target or static-weight control strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of charging piles, in particular to a charging pile AC / DC power dynamic collaborative regulation system. BACKGROUND

[0002] With the rapid popularization of electric vehicles, large charging stations face the complex scenario of multiple vehicles charging in parallel. In a high-density charging environment, traditional power distribution methods have technical defects such as low power distribution accuracy leading to reduced charging efficiency, serious mutual interference between charging modules affecting system stability, and significant decrease in overall conversion efficiency with increasing load. The root cause of these problems lies in the lack of effective quantification and dynamic regulation mechanism for the coupling effect between charging modules.

[0003] The traditional charging management system has relatively simple control logic. It will receive a 120kW power request and try to meet the request without exceeding the static limit of the total capacity of the transformer. At this time, a series of chain negative effects will occur: for other vehicles that have been charged to 90% SOC, the system cannot identify that they cannot efficiently receive high power, but can still maintain their original power distribution, causing waste of electric energy and risk of battery overheating; in an already fragile power grid environment, the sudden increase of 120kW of large load will cause the voltage at the entrance of the charging station to drop further, affecting all vehicles being charged, and even causing some sensitive vehicles to interrupt charging due to under-voltage protection; the harmonics and voltage fluctuations generated by the newly connected large power load will be transmitted to adjacent charging modules through shared power distribution lines. Due to the lack of quantification and compensation of this coupling effect, the power control loop of adjacent modules will be severely disturbed, resulting in power oscillation, reducing charging stability, and in severe cases, causing module failure, making the entire system inefficient, unstable and not friendly to the power grid SUMMARY

[0004] The present application aims to provide a charging pile AC / DC power dynamic collaborative regulation system to solve the problems raised in the background.

[0005] The technical solution of the present application is as follows: an electric grid parameter acquisition module is used to acquire the voltage effective value, frequency deviation, harmonic distortion rate and voltage unbalance degree of a three-phase electric grid; the electric grid parameter acquisition module generates an electric grid voltage quality comprehensive evaluation index based on the acquired parameters;

[0006] A charging load characteristic analysis module is used to monitor the real-time power demand, battery state of charge and charging curve characteristics of each charging terminal; the charging load characteristic analysis module generates a dynamic load characteristic function based on the monitored parameters;

[0007] a coupling effect calculation module for quantifying the degree of power coupling among charging pile modules; the coupling effect calculation module uses an information theory method to calculate a power coupling coefficient;

[0008] a power distribution optimization module for receiving the grid voltage quality comprehensive evaluation index, the dynamic load characteristic function, and the power coupling coefficient; wherein the power distribution optimization module establishes an optimization model to generate a power distribution strategy;

[0009] a cooperative control execution module for adjusting the output power of each charging module according to the power distribution strategy; wherein the cooperative control execution module implements adaptive control by introducing decoupling compensation.

[0010] Preferably, the grid parameter acquisition module generates the grid voltage quality comprehensive evaluation index by normalizing and weightedly summing the voltage effective value, the frequency deviation, the harmonic distortion rate, and the voltage unbalance degree.

[0011] Preferably, the charging load characteristic analysis module generates a charging efficiency correction coefficient η soc,i (t) and a charging curve characteristic coefficient γ curve,i (t); the charging efficiency correction coefficient η soc,i (t) is determined based on the battery state of charge;

[0012] The charging efficiency correction coefficient η soc,i (t) is defined according to the charging characteristics of lithium batteries as follows:

[0013] η soc,i (t) = 0.8 + 0.2·exp(-β·SOC i (t))

[0014] wherein SOC i (t) is the real-time battery state of charge percentage of the vehicle connected to the i-th terminal; β is a coefficient representing the decay rate, which is obtained by nonlinear curve fitting on a large amount of typical lithium battery charging data, and its typical value is 3.0; this formula accurately simulates the physical and chemical characteristics of lithium batteries, i.e., the internal resistance is small and the charging acceptance ability is strong (η soc,i (t) tends to 1.0) at low SOC, while the charging acceptance ability decreases (η soc,i (t) tends to 0.8) due to the intensification of polarization effect at high SOC;

[0015] The dynamic load characteristic function is determined by the power request value, the charging efficiency correction coefficient, and the charging curve characteristic coefficient;

[0016] The mathematical expression of the dynamic load characteristic function is as follows:

[0017] L i (t) = Preq,i (t)·η soc,i (t)·γ curve,i (t)

[0018] In the formula, i is the index of the charging terminal; t represents time; P req,i (t) is the power request value reported by the i-th charging terminal at time t by the vehicle BMS through the charging communication protocol; η soc,i (t) is a dimensionless value representing a correction factor for charging efficiency based on the battery's state of charge; γ curve,i (t) is the characteristic coefficient of the charging curve, which reflects the characteristics of the charging stage and is a dimensionless value.

[0019] Preferably, the coupling effect calculation module calculates the mutual information of the power sequences of the two charging modules; wherein, the coupling effect calculation module calculates the mutual information of the power sequences of the two charging modules; wherein, the coupling effect calculation module normalizes the mutual information; the coupling effect calculation module calculates the electrical distance attenuation factor; the power coupling coefficient is determined by the normalized mutual information and the electrical distance attenuation factor;

[0020] Power coupling coefficient C ij The definition formula is as follows:

[0021]

[0022] In the formula, i and j represent different charging modules; the first term is the normalized mutual information, used to measure the power sequence P of module i. i The power sequence P of module j j Information correlation between them; I(P) i ;P j H(P) represents the mutual information between two power sequences; i ) and H(P j The first term represents the information entropy of the two power sequences, characterizing the uncertainty of their respective power fluctuations; the second term is the electrical distance attenuation factor, used to reflect the coupling attenuation law at the physical level; d ij The equivalent electrical impedance of modules i and j in the charging station's power distribution network is calculated by analyzing the topology of the power distribution network and the resistance and reactance parameters of the lines.

[0023] Preferably, the optimization model established by the power allocation optimization module includes a power deviation term, a coupled power term, and a grid constraint term; wherein, the power allocation optimization module performs a weighted summation of the three terms to construct an objective function.

[0024] Preferably, the synergic control execution module generates a basic control output; wherein the synergic control execution module calculates the sum of the product of the control output of other charging modules and the corresponding power coupling coefficient as a decoupling compensation; the synergic control execution module subtracts the decoupling compensation from the basic control output to generate a final control signal.

[0025] Preferably, the power distribution optimization module dynamically adjusts the weight coefficient by using a reinforcement learning method; wherein the power distribution optimization module uses a Q-learning algorithm to learn and dynamically adjust the weight coefficient (ω1, ω2, ω3) of the power deviation term, the coupling power term and the grid constraint term in the target function online.

[0026] The state space (S) of the Q-learning algorithm is defined as a discretized vector containing the real-time grid quality index (Q grid ) and the number of vehicles in the low, medium and high state of charge intervals, to represent the overall working condition of the system.

[0027] The algorithm updates the strategy according to a comprehensive reward function (R t ), which gives positive rewards for low deviation of allocated power from dynamic load demand, and negative penalties for total power fluctuation and behavior exceeding the dynamic power supply upper limit determined by the grid quality, thereby guiding the system to learn to dynamically balance between meeting user demand, ensuring system stability and maintaining grid safety.

[0028] Preferably, the grid parameter acquisition module analyzes the correlation between grid quality and power fluctuation in historical operation data; wherein the grid parameter acquisition module dynamically optimizes the weight coefficient by using gradient descent method.

[0029] Preferably, it further comprises a communication interface module for obtaining charging load information from each charging terminal; wherein the coupling effect calculation module is implemented by using a digital signal processor; the power distribution optimization module solves the optimization problem by using an interior point method.

[0030] The present application improves the charging pile AC / DC power dynamic synergic control system, which has the following improvements and advantages compared with the prior art.

[0031] 1. By introducing a dynamic load characteristic function, the system can accurately allocate power to vehicles that really need it and can efficiently utilize it.

[0032] 2. It can effectively suppress mutual interference between modules, reduce interference level, avoid power oscillation and chain failure, and ensure the stability and reliability of the entire charging station under high-density operation.

[0033] 3. With the ability to perceive the health of the power grid, it can actively reduce the load when the large power grid is fragile, and fully utilize the resources when the power grid is strong, realizing the transformation from passive power consumption to active adaptation, and enhancing the robustness of the system;

[0034] 4. By establishing and solving a multi-objective optimization model containing power deviation, coupled power and grid constraints, especially by dynamically adjusting the weights of each objective using reinforcement learning, the system can find and maintain a dynamic, globally optimal balance point between meeting user demand, ensuring its own stability and maintaining grid safety, which is beyond the reach of any single objective or static weight control strategy. BRIEF DESCRIPTION OF DRAWINGS

[0035] The present application will be further explained in conjunction with the accompanying drawings and examples:

[0036] Figure 1 is a flowchart of the charging pile AC / DC power dynamic collaborative control system of the present application. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application will be further explained in detail below in conjunction with specific examples.

[0038] Example 1:

[0039] Please refer to Figure 1 The present application provides a technical scheme of a charging pile AC / DC power dynamic collaborative control system, which comprises: a power grid parameter acquisition module for acquiring the voltage effective value, frequency deviation, harmonic distortion rate and voltage unbalance degree of a three-phase power grid; the power grid parameter acquisition module generates a power grid voltage quality comprehensive evaluation index based on the acquired parameters;

[0040] A charging load characteristic analysis module is used to monitor the real-time power demand, battery state of charge and charging curve characteristics of each charging terminal; the charging load characteristic analysis module generates a dynamic load characteristic function based on the monitored parameters;

[0041] A coupling effect calculation module is used to quantify the degree of power coupling between charging pile modules; the coupling effect calculation module calculates the power coupling coefficient using information theory method;

[0042] A power distribution optimization module is used to receive the power grid voltage quality comprehensive evaluation index, dynamic load characteristic function and power coupling coefficient; wherein the power distribution optimization module establishes an optimization model to generate a power distribution strategy;

[0043] A collaborative control execution module is used to adjust the output power of each charging module according to the power distribution strategy; wherein the collaborative control execution module implements adaptive control by introducing decoupling compensation.

[0044] The embodiment discloses a charging pile AC / DC power dynamic collaborative regulation system, realizes a closed-loop and globally collaborative dynamic power regulation architecture through precise cooperation of a power grid parameter acquisition module, a charging load characteristic analysis module, a coupling effect calculation module, a power distribution optimization module and a collaborative control execution module; the system regards the entire charging station as a whole which dynamically interacts with the power grid and all charging vehicles, quantifies power grid health, vehicle power receiving capacity and pile interference in real time, and performs globally optimal power distribution and decoupling control based on this; this fundamentally solves technical problems such as low efficiency and system instability caused by power distribution blindness in a high-density charging scene, and realizes a fundamental paradigm shift from traditional passive and isolated control to active prediction and adaptive collaborative control.

[0045] Embodiment 2

[0046] The power grid parameter acquisition module generates a power grid voltage quality comprehensive evaluation index through normalization processing and weighted summation of the voltage effective value, the frequency deviation, the harmonic distortion rate and the voltage unbalance degree.

[0047] The power grid parameter acquisition module analyzes the correlation between power grid quality and power fluctuation in historical operation data; wherein, the power grid parameter acquisition module adopts a gradient descent method to dynamically optimize the weight coefficient.

[0048] The power grid parameter acquisition module in the embodiment accurately quantifies the real-time bearing capacity of the power grid through a comprehensive evaluation index; the core of this function is the power grid voltage quality comprehensive evaluation index Q grid ; the design of the index is based on the deep application of the national standard GB / T12325-2008 power quality evaluation system, aiming to overcome the one-sidedness of traditional control strategies which only focus on a single parameter, and condenses multi-dimensional power grid quality parameters into a single, sensitive and comprehensive scalar index, providing an accurate and reliable power grid constraint boundary for subsequent power distribution optimization;

[0049] The mathematical expression of the index is:

[0050]

[0051] In the formula, V rms represents the average value of the three-phase voltage effective value obtained in real time through a voltage transformer and digital sampling; V rated represents the rated voltage value of the charging station connected to the power grid, and the standard value is 380V or 220V; Δf represents the absolute value of the power grid frequency deviation measured by using a phase-locked loop technology, and the unit is hertz; f rated represents the rated frequency of the power grid, which is 50Hz in the Chinese standard system; THD vVUF is the voltage unbalance degree calculated according to the positive and negative sequence components, which is a dimensionless percentage; a1, a2, a3, a4 are weight coefficients of each parameter, and the sum satisfies the normalization condition ∑a i = 1.

[0052] In the system operation, the power grid parameter acquisition module continuously calculates Q grid value; when the power grid quality is good, Q grid value tends to 1, indicating that the power grid has strong power support capability, and the power distribution module can distribute higher total power accordingly; on the contrary, when the power grid quality decreases, Q grid value decreases, and the system actively and intelligently limits the total charging power as a strong constraint; this avoids the risk of aggravating the power grid deterioration or even leading to system off-grid due to overload charging, greatly enhancing the friendly interaction between the charging station and the power grid;

[0053] The determination of weight coefficients a1, a2, a3, a4 is not static setting; the initial value can be set to a1 = 0.4, a2 = 0.2, a3 = 0.2, a4 = 0.2 according to engineering experience to highlight the basic influence of voltage stability; a key refinement is that the system has self-learning ability, and by analyzing the long-term accumulated historical operation data, a correlation model between each quality component of the power grid and the total power fluctuation of the charging system is established; the system uses gradient descent algorithm to minimize the error between the power fluctuation predicted based on the quality components of the power grid and the actual power fluctuation as the target, and periodically optimizes the a weight combination online;

[0054] The system establishes a linear prediction model to correlate the quality components of the power grid and the total power fluctuation:

[0055]

[0056] In the formula, is the predicted total power variance, w i is the model parameter. The system minimizes the mean square error between the predicted variance and the actual total power variance calculated in a sliding window of T length to periodically update the weight coefficient a i by gradient descent method; the update rule here aims to make the change trend of Q grid index and the actual power fluctuation that the power grid can withstand have stronger negative correlation, thereby guiding the adjustment direction of the weight;

[0057] This dynamic optimization makes Q gridThe index can evolve from a general model to a customized evaluation standard for specific station grid characteristics, and the evaluation results are more targeted and accurate.

[0058] α i is proportional to the gradient change of w i , or the optimized w i is mapped to a new α i through a functional relationship, so that those skilled in the art can clearly implement it.

[0059] Embodiment 3

[0060] The charging load characteristic analysis module generates a charging efficiency correction coefficient η soc,i (t) and a charging curve characteristic coefficient γ curve,i (t); the charging efficiency correction coefficient η soc,i (t) is determined based on the battery state of charge; and the dynamic load characteristic function is determined by the power request value, the charging efficiency correction coefficient and the charging curve characteristic coefficient.

[0061] The charging load characteristic analysis module in this embodiment can accurately depict the real power acceptance capability of each charging terminal at any time, and its core is to construct a dynamic load characteristic function L i (t); the module generates a charging efficiency correction coefficient η soc,i (t) and a charging curve characteristic coefficient γ curve,i (t) by deeply analyzing the data reported by the vehicle BMS, and multiplies them with the BMS request power P req,i (t) to jointly constitute the function;

[0062] The mathematical expression of the dynamic load characteristic function is:

[0063] L i (t)=P req,i (t)·η soc,i (t)·γ curve,i (t)

[0064] In the formula, i is the index of the charging terminal; t represents time; P req,i (t) is the power request value of the i-th charging terminal at t reported by the vehicle BMS through the charging communication protocol; η soc,i (t) is the charging efficiency correction coefficient based on the battery state of charge, which is a dimensionless value; γ curve,i (t) is the charging curve characteristic coefficient reflecting the charging stage characteristics, which is a dimensionless value;

[0065] The charging efficiency correction coefficient η soc,i (t) is defined according to the charging characteristics of lithium batteries as follows:

[0066] η soc,i (t) = 0.8 + 0.2 · exp(-β · SOC i (t))

[0067] wherein SOC i (t) is the real-time battery state of charge percentage of the vehicle connected to the ith terminal; β is a coefficient representing the decay rate, which is obtained by nonlinear curve fitting on a large number of typical lithium battery charging data, and its typical value is 3.0; this formula accurately simulates the physical and chemical properties of lithium batteries, i.e., small internal resistance and strong charging acceptance ability (η soc,i (t) tends to 1.0) at low SOC, and the charging acceptance ability decreases (η soc,i (t) tends to 0.8) due to the intensification of polarization effect at high SOC;

[0068] The value logic of the charging curve characteristic coefficient γ curve,i (t) is clear: in the constant current charging phase, the power is basically constant, and at this time γ curve,i (t) takes the value of 1.0; when the charging process enters the constant voltage phase, the coefficient starts to linearly decrease from 1.0 to 0.7 according to the ratio of the real-time battery terminal voltage to the constant voltage charging starting point voltage;

[0069] The system determines whether the charging process enters the constant voltage phase by analyzing the charging state message reported by the vehicle-mounted BMS through the charging communication protocol; when the system receives the state flag bit sent by the BMS indicating the start of the constant voltage charging mode, it is determined that the constant voltage phase starts, and the γ_curve, i(t) coefficient starts to linearly decrease; or, in the case where the flag bit cannot be obtained, the backup judgment logic is: when the system monitors that the battery terminal voltage of the terminal is stable at the preset constant voltage charging voltage value for 3 consecutive sampling periods, it is also determined that the constant voltage phase starts; to accurately reflect the trend of natural attenuation of the charging current and power during the constant voltage period;

[0070] Through this mechanism, the L i (t) generated by the charging load characteristic analysis module is no longer a static request value reported by the vehicle alone, but a dynamic and accurate effective power demand that comprehensively considers the instantaneous demand of the vehicle, the intrinsic state of the battery, and the charging process phase. This enables the subsequent power allocation decision to be based on the most real acceptance ability of each load, avoiding energy waste and battery damage caused by allocating redundant power to high SOC vehicles, and preventing power allocation mismatch due to ignoring the constant voltage conversion point, thereby significantly improving the effectiveness of allocation and the energy conversion efficiency of the entire charging process.

[0071] Embodiment 4

[0072] The coupling effect calculation module calculates mutual information of the power sequences of the two charging modules; the coupling effect calculation module normalizes the mutual information; the coupling effect calculation module calculates an electrical distance attenuation factor; and the power coupling coefficient is determined by the normalized mutual information and the electrical distance attenuation factor.

[0073] The coupling effect calculation module in the embodiment constitutes a key technical innovation for the system to realize collaborative control, which introduces an information theory tool to quantitatively evaluate the mutual interference between the charging modules; the core output of the module is the power coupling coefficient C ij ; the technical motivation is to provide an accurate mathematical description for the module coupling effect which is traditionally difficult to quantify; this is intended to overcome the decoupling problem caused by the inability to perceive and quantify the coupling relationship in the traditional control strategy;

[0074] The definition formula of the power coupling coefficient C ij is as follows:

[0075]

[0076] In the formula, i and j represent different charging modules; the first term is the normalized mutual information, which is used to measure the information correlation between the power sequence P i of the module i and the power sequence P j of the module j; I(P i ; P j ) is the mutual information between the two power sequences; H(P i ) and H(P j ) are the information entropy of the two power sequences, respectively, representing the uncertainty of the power fluctuation of each module; the second term is the electrical distance attenuation factor, which is used to reflect the coupling attenuation law in the physical layer; d ij represents the equivalent electrical impedance of the modules i and j in the power distribution network of the charging station, which is calculated by analyzing the topological structure of the power distribution network and the resistance and reactance parameters of the lines, and the unit is ohm;

[0077] The calculation process first models the alternating current power distribution network of the charging station as a circuit diagram, in which the secondary side of the transformer is the main node, the access points of each charging pile are the sub-nodes, and the resistance and reactance of the cable are the branch impedance; by constructing the node admittance matrix of the network and using the matrix inversion method, the transfer impedance between any two charging pile nodes i and j is calculated, and the modulus of the transfer impedance is the equivalent electrical impedance d ij ;

[0078] λ is an attenuation constant, which is calibrated by measuring the voltage disturbance transmission level between modules at different electrical distances in an actual or simulated environment and fitting the model, and its typical value is 10 ohm, which ensures that the attenuation factor can truly reflect the physical attenuation effect;

[0079] To ensure the field implementability of the model, the calculation procedure of mutual information I(P i ; P j ) is defined as follows: discretize the continuous power value into 20 equal-width intervals; in a sliding time window with a length of 60 seconds, count the frequency of occurrence of each power interval and take it as the probability estimate; according to the joint probability distribution and the marginal probability distribution obtained by statistics, calculate the mutual information and the information entropy according to the standard definition of information theory. This parameter setting is the result of optimization between guaranteeing statistical significance and dynamic response speed;

[0080] The coupling effect calculation module periodically calculates and updates the coupling coefficients between all module pairs (i, j), forming a dynamically updated coupling coefficient matrix quantifying the interference strength between modules; this matrix provides indispensable quantitative basis for subsequent power allocation optimization and decoupling control and is the fundamental prerequisite for reducing the crosstalk between modules and improving the stability of the system.

[0081] Embodiment 5

[0082] The optimization model established by the power allocation optimization module includes a power deviation term, a coupling power term and a power grid constraint term; wherein the power allocation optimization module performs weighted summation on the three terms to construct the objective function;

[0083] The power allocation optimization module dynamically adjusts the weight coefficients using the reinforcement learning method; wherein the power allocation optimization module defines a comprehensive reward function containing the energy conversion efficiency, the power fluctuation degree and the user satisfaction degree; the power allocation optimization module uses the Q-learning algorithm to learn the optimal weight combination online;

[0084] The power allocation optimization module in this embodiment serves as the central decision unit of the system and generates the optimal power allocation scheme P i by solving a carefully designed multi-objective optimization model; the objective function J(P i ) of the model aims to minimize three core indicators that are mutually restrictive:

[0085]

[0086] In the formula, the first term is the power deviation term, which aims to minimize the squared error between the allocated power P i and the dynamic load demand L i to ensure that the user charging demand is maximally satisfied; the second term is the coupling power term, which uses the coupling coefficient C ij to minimize the weighted coupling power between all modules to reduce the interference between modules and improve the stability of the system; and the third term is the power grid constraint term, which is used to punish the deviation of the total charging power P total from the upper limit of the dynamic power supply, which is determined by the rated capacity P grid,maxwith real-time grid quality indicator Q grid co-decide; in the objective function, P i is the power allocation variable of each module to be optimized; L i is the dynamic load demand calculated by the charging load characteristic analysis module; C ij is the coupling coefficient provided by the coupling effect calculation module; N is the total number of charging modules; P total =∑P i is the total charging power; ω1,ω2,ω3 are the weight coefficients of each term;

[0087] One key refinement is that the weight coefficients ω1,ω2,ω3 are not statically set, but are adaptively adjusted online by an integrated Q-learning reinforcement learning agent; the system defines a comprehensive reward function, the design principle of which is: give positive rewards to high energy conversion efficiency and high user satisfaction (i.e. low deviation of P i and L i ), while imposing negative penalties on severe total power fluctuations and behaviors that exceed grid constraints; during operation, the Q-learning algorithm constantly explores different weight combinations and updates its Q-value table according to the immediate rewards obtained from the environment, gradually learning how to balance the best strategies for meeting demand, reducing interference, and "protecting the grid" under different system states;

[0088] To implement this Q-learning agent, the system's state space, action space, and reward function

[0089] are specifically defined as follows:

[0090] State space S: state s t is defined as a discretized vector representing the overall operating condition of the system at time t: s t =[round(10·Q grid ),N low_soc ,N mid_soc ,N high_soc ], where the first term is the grid quality indicator rounded to one decimal place, and the last three terms are the number of charging vehicles in the low, medium, and high SOC intervals, respectively;

[0091] Action space A: action a t corresponds to a set of pre-defined weight combination; for example, five typical strategy tendencies can be defined: A={a1,a2,a3,a4,a5}, where a1=(0.8,0.1,0.1) tends to prioritize meeting user demand, while a2=(0.1,0.1,0.8) tends to strictly adhere to grid constraints. Executing an action means adopting the corresponding (ω1,ω2,ω3) combination for the next period of power allocation optimization;

[0092] reward function R(s t ,a t ): after performing action a t and getting a new power allocation result, the agent gets an immediate reward R t , the mathematical expression is as follows:

[0093]

[0094] In the formula, the first term is the user satisfaction reward, the second term is the punishment for total power fluctuation, and the third term is the severe punishment for exceeding the upper limit of the grid dynamics. k1, k2, k3 are reward coefficients, which can be set to 1.0, 0.5, and 10.0, for example;

[0095] This mechanism of combining multi-objective optimization and reinforcement learning adaptive power adjustment gives the power allocation strategy high intelligence and situational awareness; the system can dynamically adjust its optimization focus according to real-time working conditions, ensuring that the system performance is always in the global optimal or near-optimal range under various complex and variable operating conditions.

[0096] Embodiment 6

[0097] The cooperative control execution module calculates the sum of the control outputs of other charging modules and the corresponding power coupling coefficients as the decoupling compensation; the cooperative control execution module subtracts the decoupling compensation from the basic control output to generate the final control signal.

[0098] The cooperative control execution module in this embodiment is responsible for accurately and quickly converting the ideal power target P i given by the power allocation optimization module into actual control actions. The key technology is an adaptive decoupling control strategy based on coupling coefficients; this strategy innovatively introduces a feedforward decoupling compensation term on the basis of the traditional PID control framework; for the i-th charging module, the generation logic of the final control output U i (t) is as follows:

[0099] Final control signal = basic PID control output - decoupling compensation

[0100] Wherein, the basic PID control output is calculated according to the error between the target power P target,i and the actual feedback power P actual,i , through the proportional, integral, and differential links, and constitutes a feedback control loop;

[0101] The calculation of the decoupling compensation is the core of the strategy, and the mathematical expression is as follows:

[0102]

[0103] where i and j are module indices; C ij is the known power coupling coefficient between module i and module j provided by the coupling effect calculation module; U j (t-τ) is the actual control output of module j in the last control period, τ is the control delay determined by the sampling and calculation time of the digital controller, and the typical value is 10 milliseconds; K dec is an adjustable decoupling gain coefficient, the value range is set to be between 0.1 and 0.5, and the specific value is determined by offline simulation optimization, and the optimization goal is to minimize the power adjustment time and overshoot under the premise of ensuring system stability;

[0104] The technical connotation of this decoupling compensation is that it estimates the total interference amount that all other modules will generate to module i based on the known coupling relationship C ij and the control behavior U j (t-τ) of the adjacent module, and performs a feed-forward hedging from the basic control instruction of module i; this operation actively compensates before the actual interference occurs and affects the output power of module i; by implementing this adaptive decoupling control, the dynamic coupling effect between modules is greatly weakened, so that each module can track its power target nearly independently, and the dynamic response speed, stability and power control accuracy of the system are significantly improved, and experiments show that the interference between modules can be reduced by more than 60%.

[0105] Further comprising a communication interface module for obtaining charging load information from each charging terminal; wherein the coupling effect calculation module is implemented by a digital signal processor; and the power distribution optimization module solves the optimization problem by using an interior point method.

[0106] To ensure efficient and reliable operation of the entire regulation system, the embodiment carefully selects hardware and software and deploys algorithms in specific engineering implementation; the system integrates a communication interface module, which uses CAN bus technology and can obtain key load information including power request and battery SOC from the BMS of each charging terminal with a communication period of 100 milliseconds, providing a timely and accurate data source for the charging load characteristic analysis module;

[0107] For the computationally intensive coupling effect calculation module, the embodiment selects a high-performance digital signal processor, specifically TMS320F28335; the processor has a 150MHz main frequency and a hardware floating point operation unit, can efficiently perform complex operations such as probability statistics, information entropy and mutual information in the sliding window, and ensures that the coupling coefficient matrix can be refreshed at a rate of seconds, meeting the needs of system dynamic response;

[0108] In the power distribution optimization module, for the established multi-objective quadratic programming problem, the embodiment adopts a mature and efficient interior point method for solving; the algorithm is deployed on an Intel Core i7 level central processor, and the time for a single completion of the optimization solution of the power of all charging piles in the whole station is less than 50 milliseconds, ensuring that the power distribution strategy can closely follow the rapid changes of the power grid and the load;

[0109] In addition, the bottom layer PWM switch frequency of the cooperative control execution module is set to 20 kHz, and the control period is 50 microseconds, ensuring the rapid and accurate adjustment of the output power; the whole system communicates with the upper computer monitoring system through the Modbus TCP protocol, supports remote monitoring, parameter adjustment and fault diagnosis; this kind of hardware and software cooperative implementation scheme closely combines the advanced control theory with the feasible engineering practice, and constitutes the solid engineering foundation for the successful implementation and significant technical effect of the present application.

[0110] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A dynamic coordinated control system for AC / DC power of a charging pile, characterized in that, include: The power grid parameter acquisition module is used to acquire the effective voltage value, frequency deviation, harmonic distortion rate and voltage imbalance of the three-phase power grid. The power grid parameter acquisition module generates a comprehensive evaluation index of power grid voltage quality based on the acquired parameters; The charging load characteristic analysis module is used to monitor the real-time power demand, battery state of charge, and charging curve characteristics of each charging terminal; the charging load characteristic analysis module generates a dynamic load characteristic function based on the monitored parameters; The coupling effect calculation module is used to quantify the degree of power coupling between charging pile modules; The coupling effect calculation module uses information theory methods to calculate the power coupling coefficient; The power allocation optimization module is used to receive the comprehensive evaluation index of grid voltage quality, dynamic load characteristic function and power coupling coefficient; wherein, the power allocation optimization module establishes an optimization model to generate a power allocation strategy; The collaborative control execution module is used to adjust the output power of each charging module according to the power allocation strategy; wherein, the collaborative control execution module implements adaptive control by introducing decoupling compensation.

2. The AC / DC power dynamic coordinated control system for charging piles according to claim 1, characterized in that, The power grid parameter acquisition module generates a comprehensive evaluation index for power grid voltage quality by normalizing and weighting the effective voltage value, frequency deviation, harmonic distortion rate, and voltage imbalance.

3. The AC / DC power dynamic coordinated control system for charging piles according to claim 1, characterized in that, The charging load characteristic analysis module generates a charging efficiency correction coefficient η. soc,i (t) and characteristic coefficient γ of the charging curve curve,i (t); the charging efficiency correction coefficient η soc,i (t) is determined based on the battery state of charge; Charging efficiency correction factor η soc,i (t) is defined according to the charging characteristics of lithium batteries: the soc,i (t)=0.8+0.2·exp(-β·SOC i (t)) In the formula, SOC i (t) represents the real-time battery state-of-charge percentage of the vehicle connected to the i-th terminal; β is a coefficient characterizing the degradation rate, obtained by nonlinear curve fitting of a large amount of typical lithium battery charging data, with a typical value of 3.0; this formula accurately simulates the low internal resistance and strong charge acceptance capability (η) of lithium batteries at low SOC. soc,i (t) approaches 1.0), while at high SOC, the increased polarization effect leads to a decrease in charge acceptance (η). soc,i (t) approaches 0.8) physicochemical properties; The dynamic load characteristic function is determined by the power request value, the charging efficiency correction coefficient, and the charging curve characteristic coefficient. The mathematical expression of the dynamic load characteristic function is: L i (t)=P req,i (t)·η soc,i (t)·γ curve,i (t) In the formula, i is the index of the charging terminal; t represents time; P req,i (t) is the power request value reported by the i-th charging terminal at time t by the vehicle BMS through the charging communication protocol; η soc,i (t) is a dimensionless value representing a correction factor for charging efficiency based on the battery's state of charge; γ curve,i (t) is the characteristic coefficient of the charging curve, which reflects the characteristics of the charging stage and is a dimensionless value.

4. The AC / DC power dynamic coordinated control system for charging piles according to claim 1, characterized in that, The coupling effect calculation module calculates the mutual information of the power sequences of the two charging modules; wherein, the coupling effect calculation module normalizes the mutual information; the coupling effect calculation module calculates the electrical distance attenuation factor; the power coupling coefficient is determined by the normalized mutual information and the electrical distance attenuation factor; Power coupling coefficient C ij The definition formula is as follows: In the formula, i and j represent different charging modules; the first term is the normalized mutual information, used to measure the power sequence P of module i. i The power sequence P of module j j Information correlation between them; I(P) i ;P j H(P) represents the mutual information between two power sequences; i ) and H(P j The first term represents the information entropy of the two power sequences, characterizing the uncertainty of their respective power fluctuations; the second term is the electrical distance attenuation factor, used to reflect the coupling attenuation law at the physical level; d ij The equivalent electrical impedance of modules i and j in the charging station's power distribution network is calculated by analyzing the topology of the power distribution network and the resistance and reactance parameters of the lines.

5. The AC / DC power dynamic coordinated control system for charging piles according to claim 1, characterized in that, The optimization model established by the power allocation optimization module includes a power deviation term, a coupled power term, and a grid constraint term; wherein, the power allocation optimization module constructs an objective function by weighted summation of the three terms.

6. The AC / DC power dynamic coordinated control system for charging piles according to claim 1, characterized in that, The collaborative control execution module generates a basic control output; wherein, the collaborative control execution module calculates the sum of the products of the control outputs of other charging modules and their corresponding power coupling coefficients as a decoupling compensation amount; the collaborative control execution module subtracts the decoupling compensation amount from the basic control output to generate the final control signal.

7. The AC / DC power dynamic coordinated control system for charging piles according to claim 5, characterized in that, The power allocation optimization module uses the Q-learning algorithm to learn online and dynamically adjust the weight coefficients (ω1, ω2, ω3) of the power deviation term, coupled power term, and grid constraint term in the objective function; The state space (S) of the Q-learning algorithm is defined as containing real-time power grid quality indicators (Q). grid Discretized vectors of the number of vehicles in low, medium, and high charge states are used to characterize the overall operating conditions of the system. The algorithm is based on a comprehensive reward function (R). t The update strategy rewards low deviations between allocated power and dynamic load demand, and penalizes total power fluctuations and behaviors that exceed the dynamic power supply limit determined by grid quality. This guides the system to learn to dynamically balance meeting user needs, ensuring its own stability, and maintaining grid security.

8. The AC / DC power dynamic coordinated control system for charging piles according to claim 2, characterized in that, The power grid parameter acquisition module analyzes the correlation between power grid quality and power fluctuations in historical operating data; wherein, the power grid parameter acquisition module uses the gradient descent method to dynamically optimize the weight coefficients.

9. The AC / DC power dynamic coordinated control system for charging piles according to claim 1, characterized in that, It also includes a communication interface module for obtaining charging load information from each charging terminal; wherein, the coupling effect calculation module is implemented using a digital signal processor; and the power allocation optimization module uses the interior point method to solve the optimization problem.

Images