Multi-agent based frequency stability control method for ac microgrid

Abstract

The application discloses a kind of Multi-Agent-based alternating current microgrid frequency stability control method, it is related to power grid control technical field.Light storage Agent uses virtual synchronous machine control, collects frequency deviation, frequency change rate, power, SOC, current margin and time stamp.Light storage Agent first generates pre-compensation by frequency consistency error, power sharing error and frequency change rate, then freezes its symbol and calculates SOC weight, available power weight and capacity boundary;When direction is not consistent, positive discharge and negative absorption set are normalized respectively, overdue neighbor does not participate in consistency error.Each Agent modifies virtual inertia, damping and active compensation in parallel, and inputs active reference into VSG active loop;Small interference executes damping enhancement and power balance, and inertia support, energy storage release and load shedding under current limiting and SOC constraint when large interference, reduce frequency deviation and inhibit control drift.

Description

Technical Field

[0001] This invention relates to the fields of new energy grid connection control, AC microgrid frequency stability control, and multi-agent cooperative control technology, and more specifically to a multi-agent-based AC microgrid frequency stability control method. Background Technology

[0002] As the proportion of distributed energy sources such as photovoltaics, wind power, and energy storage in AC microgrids increases, the mechanical inertia and damping provided by traditional synchronous generators decrease. Under conditions such as load fluctuations, sudden changes in photovoltaic output, grid connection point disturbances, islanding switching, and short-circuit faults, microgrids are prone to problems such as increased frequency change rate, increased frequency deviation, power oscillation, and prolonged recovery time.

[0003] Virtual synchronous motor control, by simulating the rotor motion equations of a synchronous generator, enables power electronic grid-connected inverters to possess equivalent inertia support and damping adjustment capabilities. Existing photovoltaic-storage VSG control systems mostly focus on a single unit or adjust virtual inertia and damping only based on local frequency deviations, making it difficult to consider the coordination relationships between multiple photovoltaic-storage units, controllable loads, grid-connected interfaces, and fault support resources.

[0004] Existing multi-agent microgrid control schemes can achieve power sharing through consistency errors. Other schemes incorporate energy storage SOC into capacity allocation, and virtual synchronous motor control schemes adjust inertia and damping based on frequency deviation or rate of change. However, directly using the final collaborative compensation amount in SOC weight calculation can easily lead to an algebraic loop where weights and compensation amounts are interdependent. If weighting is uniformly applied based solely on SOC levels, opposing protection requirements will arise in both energy storage discharge support and power absorption recovery directions, causing low-SOC units to continue discharging or high-SOC units to continue absorbing power. Furthermore, if the state variables of adjacent agents have expired but still participate in consistency calculations, control drift may be introduced after communication delays or packet loss.

[0005] Centralized microgrid controllers require collecting information from the entire network before uniformly calculating and issuing control quantities, making them susceptible to communication delays, single-point failures, and computational bottlenecks. Traditional droop-mounted or single-unit VSG control lacks global coordination, potentially leading to uneven distribution of support capabilities, exceeding of energy storage SOC limits, and inconsistent frequency recovery after faults. Therefore, a frequency stabilization control method is needed that uses the photovoltaic-storage system as an agent and coordinates with other agents for parallel control, while simultaneously handling SOC directionality constraints, expired communication data, and conflicts between small and large disturbance modes. Summary of the Invention

[0006] To address the issues of insufficient inertia, insufficient damping, insufficient distributed resource coordination, and limited real-time performance and robustness of centralized control in high-proportion renewable energy AC microgrids, this invention provides a multi-agent-based frequency stability control method for AC microgrids.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A multi-agent-based frequency stability control method for AC microgrids includes the following steps: Construct a photovoltaic storage agent, a grid connection interface agent, and a controllable load agent; In the Within each control cycle, each Agent collects a local state vector with a timestamp. Each Agent sends its local state vector to its neighboring Agents and calculates the frequency consistency error and power sharing error based on the effective neighbor set. The optical-storage agent generates an unweighted pre-compensation quantity from the frequency consistency error, power sharing error, and frequency change rate estimate. It freezes the symbol of the unweighted pre-compensation quantity in the current control cycle, calculates the SOC directional weight and directional available power weight based on the frozen symbol, and then calculates the capacity weight based on the converter current margin, communication age, and weight normalization activation threshold. according to Generate the collaborative active power compensation amount, where, This is the coordinated active power compensation amount after weighting and saturation limiting. Available active capacity to support discharge. Available capacity for power absorption; A disturbance identifier with hysteresis is used to determine the small or large disturbance mode of the i-th optical storage agent access point. The disturbance identifier determines the small or large disturbance mode of the i-th optical storage agent at its grid connection access point and provides the determination result for collaborative use by the grid connection interface agent and the controllable load agent.

[0008] In virtual synchronous motor control, the photovoltaic-storage agent corrects the virtual moment of inertia in parallel within the same control cycle based on the current physical state. Damping coefficient and and will As the active loop input of the virtual synchronous motor to generate the PWM modulation signal for the grid-connected inverter; When the communication delay between the i-th optical storage agent and its neighboring agents exceeds the preset time limit or is interrupted, the optical storage agent that loses communication maintains local virtual synchronous motor control and decays the most recent collaborative compensation amount to zero in an exponential manner. The agent that does not lose communication renormalizes the weight of the remaining capacity. When the weight normalization denominator of the corresponding direction is not greater than the activation threshold, the collaborative compensation in that direction is set to zero and falls back to local virtual synchronous motor control.

[0009] Optionally, the photovoltaic-storage agent includes a photovoltaic unit, an energy storage unit, a DC-DC converter unit, a grid-connected inverter unit, a sampling and communication unit, and a local control unit.

[0010] Optionally, when the unweighted pre-compensation values ​​of multiple optical storage agents have inconsistent signs, the positive discharge support set and the negative absorption power set are normalized respectively within the effective communication range.

[0011] Optionally, the virtual synchronous motor active power loop of the photovoltaic-storage agent satisfies: Where the subscript i represents the i-th optical storage Agent, t is the time, and J i Let ω be the virtual rotational inertia of the i-th optical storage agent, ω0 be the rated angular velocity, and ω i Let dω be the virtual angular velocity of the i-th optical storage agent. i / dt is the rate of change of the virtual angular velocity with respect to time, P i ref The active power reference value is ΔP. i MA P represents the coordinated active power compensation amount after weighting and saturation limiting. i e For grid-connected active power, D i θ is the damping coefficient. i For the virtual phase angle, dθ i / dt is the rate of change of the virtual phase angle with respect to time; the reactive voltage loop satisfies: , of which E i The virtual electromotive force is E0, which is the rated virtual electromotive force, and K is the virtual electromotive force. q Q is the reactive power-voltage droop factor. i ref Q is the reactive power reference value. i For grid-connected reactive power, ΔE i MA This refers to the reactive voltage compensation amount for multi-agent collaborative operation.

[0012] Optionally, the effective neighbor set, frequency consistency error, power sharing error, and unweighted pre-compensation amount must satisfy the following:

[0013]

[0014]

[0015] ； Where k is the control cycle number, i is the current Agent number, j is the j-th neighboring Agent number, l is the index of the adjacency weight normalized summation, and N i Let N be the set of neighboring agents of the i-th agent. i v (k) is the set of effective neighbors of the i-th Agent during the k-th control period, t i (k) is the local state timestamp of the i-th Agent in the k-th control cycle, used to determine whether adjacent state variables have expired; t j (k) is the timestamp of the state quantity of the j-th adjacent Agent, T age This is the maximum communication age threshold allowed for adjacent state variables to participate in the calculation. ij Let a be the original adjacency weight between the i-th Agent and the j-th Agent. il Let be the original adjacency weight between the i-th Agent and the l-th valid neighbor. To remove expired neighbors in N i v (k) Effective adjacency weights obtained by internal renormalization, where Σ represents the summation over the Agent within the specified effective neighbor set. It is an empty set. ω,i For frequency consistency error, b i Let ωi be the traction weight of the i-th Agent with respect to the rated angular velocity, where ω0 is the rated angular velocity, and ωi is the traction weight of the i-th Agent with respect to the rated angular velocity. i and ω j Let be the virtual angular velocities of the i-th Agent and the j-th Agent, respectively. For power sharing error, P i and P j S represents the grid-connected active power of the i-th Agent and the j-th Agent, respectively. i and S j The rated active capacity or power base values ​​of the i-th Agent and the j-th Agent are respectively, ΔP i pre K is the unweighted pre-compensation amount. ω,i K P,i and K r,i These are the frequency consistency error gain, power sharing error gain, and frequency change rate suppression gain, respectively. i Let be the estimated rate of change of the frequency of the i-th Agent.

[0016] when When the summation term involving adjacent Agents is 0, if The corresponding adjacency weight is set to zero in this control cycle. It is obtained solely by renormalizing the remaining valid neighbors.

[0017] Optionally, the available active capacity and available power absorption capacity of the discharge support can satisfy:

[0018]

[0019] in and These are the energy storage discharge power and the rechargeable power, respectively. The active power margin can be increased for photovoltaic systems. This is the conversion factor from current to active power. Rated active power margin for grid-connected inverter units ， This represents the DC-side power absorption margin; when any capacity boundary is negative or the current margin is negative... Set the capacity of the corresponding direction to 0.

[0020] Optionally, the following factors can be used to satisfy the following: power factor, SOC factor, converter current margin factor, and communication age factor:

[0021]

[0022] ； Among them, sat [0,1] The parentheses (·) are saturation limiting functions used to restrict the variable within the brackets to the interval [0,1]. The value is 0 when the variable is less than 0, 1 when it is greater than 1, and its original value when it lies between 0 and 1. min The minimum state of charge (SOC) allowed for energy storage low To reduce the weight of the low SOC threshold in the discharge support direction, SOC high To reduce the weight of the high SOC threshold in the direction of absorbed power, SOC max The maximum permissible state of charge (SOC) for energy storage. min <SOC low <SOC high <SOC max This is used to ensure that the denominator of the SOC segmentation function is positive, and to form the lower discharge limit protection range and the upper charge limit protection range. i,max For the maximum allowable current of the i-th converter, I i,limI is the warning or limiting threshold for the i-th converter current margin to begin to decay. i,lim i,max This is used to ensure that the denominator of the current margin factor is positive. T warn T is the communication age warning threshold. age τ is the threshold for determining communication aging. i (k) represents the communication age of the most recent valid communication for the i-th Agent. 0 <T warn <T age This is used to ensure that the denominator of the communication age coefficient is positive, and to gradually transition the communication status from normal and early warning to exiting collaboration due to exceeding the age limit.

[0023] Optionally, capacity weights are first determined by unnormalized capacity metrics. Calculate, then normalize within the corresponding direction set according to the freezing direction of the unweighted pre-compensated amount, satisfying:

[0024]

[0025] ; Where, ρ i The unnormalized capability index of the i-th optical-storage agent is determined by the SOC status, directional available power, converter current margin, and communication age. For the first A set of valid communication agents that a single optical storage agent can obtain, including itself. ρ min Enable threshold for weight normalization. This is used to avoid distorted weight calculations caused by excessively small summations of capability indices within the corresponding directional set. When Or the normalized denominator in the corresponding direction is not greater than season . The freeze direction is determined in the current control cycle and is recalculated in the next control cycle. ψ SOC,i Let ψ be the SOC coefficient of the i-th photovoltaic-storage agent, used to characterize the degree to which the energy storage state of charge can participate in the coordinated active power allocation. P,i ψ is the directional available power coefficient for the i-th photovoltaic-storage agent, used to characterize the available active power capacity margin in the discharge support direction or the power absorption direction. I,i ψ is the converter current margin coefficient for the i-th photovoltaic-storage agent, used to characterize the margin of the current distance from the maximum allowable current. H,i Let be the communication age coefficient of the i-th Agent, used to characterize the freshness of the most recent valid communication data. , , and​ All coefficients are dimensionless coefficients in the interval [0,1]. The larger the value, the more suitable the i-th Agent is to participate in collaborative compensation in the current direction. When any coefficient approaches 0, the corresponding Agent's unnormalized capability index ρ i It then decreases.

[0026] Optional, frequency change rate estimate satisfy: , The virtual moment of inertia and damping coefficient are updated as follows:

[0027]

[0028] Where k is the control cycle number, T s To control the sampling period. Δf i (k) represents the frequency deviation of the i-th optical storage agent in the k-th control cycle, Δf i (k-1) represents the frequency deviation of the previous control cycle. i (k) is the estimated frequency change rate of the i-th optical storage agent in the k-th control cycle, r i (k-1) is the estimated rate of change of frequency in the previous control cycle. α r The first-order filter coefficients for the rate of change of frequency, 0 < α r <1 is used to make a trade-off between response speed and noise suppression. and J are the saturation limiting functions for the virtual moment of inertia and the damping coefficient, respectively. i,min J i,max D represents the lower and upper limits of the virtual rotational inertia of the i-th optical storage agent, respectively. i,min D i,max J represents the lower and upper limits of the damping coefficient for the i-th optical storage agent, respectively. i0 and D i0 These are the baseline values ​​for the virtual moment of inertia and the damping coefficient, ΔJ, respectively. i and ΔD i These are the inertia adjustment and damping adjustment amounts obtained from fuzzy inference, respectively, k J k D The local blur adjustment gain, k JM k DM These represent the synergistic gains from capacity weighting in inertia enhancement and damping enhancement, respectively. i Let |r be the capacity normalization weight for the i-th optical storage agent. i | represents the absolute value of the estimated rate of change of frequency, |Δf i| represents the absolute value of the frequency deviation. δ f δ is the frequency deviation dead zone threshold. r The dead zone threshold for the rate of change of frequency, when |Δf i |<δ f And |r i |<δ r At that time, it is assumed that the frequency deviation and rate of change are within the allowable small range, and let ΔJ i =0、ΔD i =0 to avoid parameter jitter, and continue with J i and D i Apply a rate of change limit.

[0029] Optionally, the disturbance identifier and communication disconnection handling satisfy: (1) If , , and Then it enters the small interference mode. Where Δf i Let r be the frequency deviation of the i-th optical storage agent access point. i The estimated rate of change of frequency, ΔU i For voltage deviation, ΔP i For active power deviation, F1, R1, U1, and P1 are the frequency deviation, frequency change rate, voltage deviation, and active power deviation judgment thresholds under small disturbance mode, respectively.

[0030] (2) If , , , If any of the conditions in are met, or If the condition is met, the system enters a high-interference mode. Here, F2, R2, U2, and P2 are the corresponding judgment thresholds under the high-interference mode. , , , And set the hold time. .

[0031] (3) In the large interference mode, when Prioritize maintaining current limiting and voltage support, and limit , and The increment.

[0032] (4) If the communication of this Agent is interrupted, the most recent collaborative active power compensation amount will be calculated as follows: Attenuation, in which .

[0033] (5) When the energy storage unit includes lithium batteries and supercapacitors, the required energy storage compensation power will be... Decomposed into low-frequency components and high frequency components .

[0034] As can be seen from the above technical solution, compared with the prior art, the present invention provides a multi-agent-based AC microgrid frequency stability control method, which has the following beneficial effects: First, it enables photovoltaics, energy storage, and inverters to collaboratively provide virtual inertia and damping at the Agent level; Second, only low-dimensional timestamped state quantities are exchanged, reducing centralized full computation and suppressing drift caused by expired data; Third, through The direction determines the SOC weight and available power weight direction, and the positive discharge support set and negative absorbed power set are normalized respectively, and then... , and Joint revision , and Fourth, improve frequency support coordination under both small and large interferences; fifth, incorporate SOC, current limiting, communication age, DC-side power margin, and asymmetric charging / discharging power boundaries into the weighting and saturation stages to reduce the risks of overcharging and over-discharging of energy storage, overcurrent in the converter, and the involvement of expired data in control; It directly acts on the VSG active power loop and generates the inverter PWM modulation signal, so that the coordinated control quantity is realized as the physical output of the grid-connected inverter. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the overall architecture of the AC microgrid frequency stability control method based on Multi-Agent of the present invention; Figure 2 This is a schematic diagram of the virtual synchronous motor control structure of the optical storage agent of the present invention; Figure 3 This is a flowchart of the multi-Agent collaborative frequency control process of the present invention; Figure 4 This is a schematic diagram of the parallel control logic for small and large disturbance modes in this invention. Figure 5This is a schematic diagram of the Agent information interaction and consistency compensation topology of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] In this invention, the photovoltaic-storage unit, grid connection interface, controllable load, and monitoring and coordination function are each modeled as an Agent in the control method. The photovoltaic-storage Agent internally employs virtual synchronous motor control, based on frequency deviation. and filtered frequency change rate Using SOC, capacity margin, current limiting status, and agent health as the core input, the virtual moment of inertia is adjusted in real time. Damping coefficient and collaborative active power compensation .

[0039] Unlike schemes that directly allocate collaborative support based on SOC or rated capacity, this invention first addresses frequency consistency error, power sharing error, and... Generate unweighted pre-compensation quantity The system freezes the symbols during the current control cycle, then determines the monotonic direction of the SOC weight and available power weight based on the frozen symbols. Finally, it generates the SOC directional weight, directional available power coefficient, current margin coefficient, and communication age coefficient using a saturated linear function. These coefficients, along with the weight normalization activation threshold, are used to generate... This sequence avoids the algebraic loop between the final collaborative compensation amount and the weights, and distinguishes the SOC protection direction and power capability boundary between the two scenarios of discharge support and power absorption; moreover, the direction set consists only of visible agents within the effective communication range of this agent, avoiding dependence on network-wide synchronization information.

[0040] Each agent exchanges timestamped low-dimensional state variables through the communication topology, calculates the distributed consensus error using the effective neighbor set and renormalized adjacency weights, and generates a collaborative compensation amount by combining capacity weights. Local control is then executed in parallel within the same control cycle. (Exceeding...) Adjacent state variables do not participate in consistency calculations, and when communication is lost, the cooperative quantity decays to zero exponentially. In small-interference mode, damping enhancement, inertia fine-tuning, and capacity-weighted power balancing are implemented. In large-interference mode, within the inverter current limit and energy storage SOC constraints, rapid inertia support, energy storage frequency division power release, grid interface fault support, and controllable load tiered disconnection are implemented. When adjacent agent modes are inconsistent, the large-interference mode takes priority and propagates within the communication topology. The aforementioned unweighted pre-compensation direction freezing, positive and negative direction grouping normalization within the effective communication range, non-negative charge / discharge boundaries, and expired neighbor removal are executed as a combined step within the same control cycle. When the weighted normalization denominator for the corresponding direction does not meet the activation threshold, the cooperative compensation for that direction is set to zero and falls back to local VSG control.

[0041] Specifically, this invention discloses a multi-agent-based frequency stability control method for AC microgrids, applied to AC microgrids including AC buses, photovoltaic-storage systems, grid-connected interfaces, controllable loads, and measurement and control communication networks, comprising the following steps: S1, each photovoltaic-storage system is constructed as a photovoltaic-storage agent, the grid connection interface is constructed as a grid connection interface agent, and the controllable load is constructed as a controllable load agent. The photovoltaic-storage agent includes a photovoltaic unit, an energy storage unit, a DC-DC converter unit, a grid-connected inverter unit, a sampling communication unit, and a local control unit. S2, in the Within each control cycle, each agent collects data with timestamps. Local state vector ,in For frequency deviation, This is the filtered rate of change estimate. Active power Reactive power Bus voltage In the state of energy storage charge, For converter capacity margin, This indicates a health status or a fault. Used to determine whether adjacent state variables have expired; S3, each Agent sends its local state vector to its neighboring Agents and calculates the frequency consistency error based on the effective neighbor set. and power sharing error ; S4, the optical storage agent is... , and Generate unweighted pre-compensation quantity And freeze during the current control cycle. The symbol is used to calculate the SOC directional weight and directional available power weight based on the frozen symbol. Then, the capacity weight is calculated based on the converter current margin, communication age, and weight normalization activation threshold. Among them, multiple optical storage agents When the symbols are inconsistent, the positive discharge support set and the negative absorption power set are normalized respectively within the effective communication range; S5, press Generate collaborative active power compensation amount, the Available active capacity to support discharge. Available capacity for power absorption; S6 uses a disturbance identifier with hysteresis to determine whether it is a small disturbance mode or a large disturbance mode. S7, the photovoltaic storage agent, in the virtual synchronous motor control, according to the disturbance mode, , , In the current limiting state, the virtual moment of inertia is corrected in parallel within the same control cycle. Damping coefficient and and will As the active loop input of the virtual synchronous motor to generate the PWM modulation signal for the grid-connected inverter; S8, when the Agent communication delay exceeds the preset time limit or is interrupted, the disconnected Agent maintains local virtual synchronous motor control and decays the most recent collaborative compensation amount to zero in an exponential decay manner, while the non-disconnected Agent renormalizes the remaining capacity weight; when the weight normalization denominator of the corresponding direction is not greater than the activation threshold, the collaborative compensation in that direction is set to zero and falls back to local virtual synchronous motor control.

[0042] The present invention will be further described below with reference to the accompanying drawings: like Figure 1 As shown, in this method, the AC bus of the AC microgrid, one or more photovoltaic (PV) and energy storage (ESS) units, controllable loads, grid connection interfaces, and monitoring and coordination functions participate in coordinated control as controlled objects or agents, respectively. Each PV-ESS agent is connected to the AC bus and internally includes a photovoltaic array, energy storage units, a common DC bus, a grid-connected inverter, and a local controller. The monitoring and coordination agent is used to maintain the communication topology, publish reference frequencies, and disseminate disturbance identification results; in a distributed implementation, the monitoring and coordination function can also be shared by multiple agents. Figure 2As shown, the photovoltaic-storage agent employs VSG control. The photovoltaic array is connected to the common DC bus via an MPPT-controlled DC / DC converter, the energy storage unit is connected to the common DC bus via a bidirectional DC / DC converter, and the DC / AC grid-connected inverter is connected to the AC bus via an LCL filter. The VSG controller calculates the virtual phase angle based on active power, reactive power, bus voltage, and frequency. Virtual angular velocity and virtual electromotive force Then, a PWM modulation signal is generated through a dual closed-loop voltage and current circuit.

[0043] No. The virtual synchronous motor active power loop of each photovoltaic energy storage agent is: The reactive voltage loop is: in, For virtual rotational inertia, The damping coefficient is... This refers to the amount of active power compensation for multi-agent collaboration.

[0044] like Figure 5 As shown, each agent in the control cycle Collect local state vectors with timestamps: The Agent only sends this low-dimensional state vector to neighboring Agents. The communication notation is as follows: , This is the adjacency weight matrix. For the first The set of neighboring agents of each Agent. The effective neighbor set and the renormalized adjacency weights are: when When the summation term involving adjacent Agents is taken as 0.

[0045] like Figure 3 As shown, the frequency consistency error and power sharing error are: Each error is normalized to the rated angular frequency and rated power before compensation. The estimated rate of frequency change satisfies... ,in First, calculate the unweighted pre-compensation amount: Freeze during the current control cycle The sign of the symbol is used to calculate the directional weight, and then the coordinated active power compensation is generated: The agreement When low frequency drops and At that time, Xiang It generates positive active power support; Available active capacity to support discharge. Available capacity for absorbing power.

[0046] In one embodiment, the available active capacity of the discharge support and the available power absorption capacity are determined as follows: in, and These are the energy storage discharge power and the rechargeable power, respectively. This refers to the active power margin that photovoltaic systems can adjust upwards between MPPT power and the power curtailment directive. This is the current-to-active-power conversion factor determined based on AC bus voltage and power factor or current vector margin. The rated active power margin of the grid-connected inverter unit, This represents the DC-side power absorption margin.

[0047] Capacity weights are calculated using the following formula: in Enable threshold for weight normalization. For the first Each optical storage agent can obtain a set of valid communication agents, including itself. If the corresponding direction set contains... If the activation conditions are not met, the cooperative active power compensation in that direction is set to zero, and only local VSG and SOC protection are maintained; when season The aforementioned The freeze direction is determined in the current control cycle and is recalculated in the next control cycle.

[0048] The usable power factor for directivity can be taken as: in, For the first The rated active capacity of each photovoltaic storage agent; and These represent the available active capacity in the two directions of discharge support and power absorption, respectively.

[0049] In one embodiment, The direction of the unweighted pre-compensation quantity is determined by the saturated linear function: in, .when When indicating energy storage discharge support, the weight of low SOC cells is reduced; when When representing the energy storage absorption power, the weight of high SOC cells is reduced.

[0050] The converter current margin factor and communication age factor can be taken as follows: in, , , For the first The communication age of the most recent valid communication for each agent. hour, The agent exits the coordinated active power increment allocation and retains the local current limiting; when hour, .

[0051] The virtual moment of inertia and damping coefficient are determined jointly by local fuzzy control and cooperative weights: Fuzzy input is and The output is and fuzzy subset is The centroid method is used to resolve fuzziness. and Set a dead zone at the time, so that and and to and Apply and The rate of change limit. Corrected , and The signal enters the VSG active power loop and, through a voltage-current dual closed loop, generates the PWM modulation signal for the grid-connected inverter. The fuzzy rule satisfies the following direction: when... Increase when it is large To reduce the rate of change of frequency; when Increase when larger To reduce frequency deviation; when the frequency enters the recovery phase and Maintain or improve To suppress power oscillation and make Fall back at the preset slope .

[0052] like Figure 4 As shown, the disturbance identifier employs a dual-mode judgment with a hysteresis threshold. If , , and If so, it is determined to be a small interference mode; if , , , If any of the conditions in are met, or If so, it is determined to be a large interference mode; among which , , , And set the hold time. When neighboring agents make inconsistent mode judgments, the mode with the most interference is given priority and broadcast within the communication topology.

[0053] In low-interference mode, each optical storage agent maintains parallel VSG control and performs real-time fine-tuning. and And through consistency compensation, the frequency support capability is made up according to... Distribution. The controllable load agent does not perform load shedding, but only performs flexible load fine-tuning. This mode is suitable for operating conditions such as small load fluctuations, slow changes in photovoltaic output, and step changes in active / reactive power commands.

[0054] In high interference mode, if and Within the allowable range, the photovoltaic-storage agent releases the active power of the stored energy and improves... and ;like In this case, current limiting and voltage support will be implemented first to limit the voltage. , and The incremental load. The controllable load agent performs tiered shelving according to the load importance level, and restores the load in tiers after the frequency recovers to the allowable range and the required holding time is met.

[0055] When the energy storage unit includes lithium batteries and supercapacitors, the required energy storage compensation power is... With cutoff frequency The low-pass filter is decomposed into low-frequency components. and high frequency components : Lithium batteries bear the responsibility Supercapacitors bear the responsibility .

[0056] If a communication by an agent times out, its local controller uses the most recent valid coordination quantity and proceeds accordingly. Attenuation, in which Simultaneously, local VSG control and SOC protection are maintained; other agents continue operation with their capacity weights renormalized within the positive discharge support set and negative absorption power set, respectively. If the timestamps of adjacent state variables exceed... If the data is outdated, it will no longer be used for consistency error calculation to avoid control drift caused by data expiration. The above calculations are performed in each control cycle according to the frequency change rate filtering, error calculation, unweighted pre-compensation calculation, directional weight calculation, saturation compensation calculation, and... , The updates are executed in the order they occur.

[0057] In one specific embodiment, the rated frequency is The VSG control sampling period is to The Agent communication cycle is to ; acceptable , , , The above values ​​are for illustrative purposes only and can be adjusted according to microgrid capacity, energy storage ratio, grid connection specifications, and protection settings.

[0058] The simulation implementation can be configured with four operating conditions: load step, sudden drop in photovoltaic output, frequency drop at grid connection point, and three-phase short circuit clearing. Evaluation metrics include the lowest frequency point and the maximum frequency change rate. Adjustment time and active power overshoot. Compared to fixed-parameter VSG, local fuzzy VSG only, and multi-agent VSG directly weighted by SOC or rated capacity, this method... The direction determines the SOC weight direction, and the positive discharge support and negative absorption power are grouped and normalized. , and Joint revision , and This allows frequency support to be distributed among multiple photovoltaic-storage agents according to available capacity, thereby reducing the risk of individual energy storage overload and uneven support.

[0059] exist , , , Under the condition that the communication graph includes components connected to the reference frequency pinning node and the communication delay is less than the preset upper limit, consistency compensation reduces frequency error and power sharing error; if the communication delay exceeds the upper limit, the local VSG is switched through the coordination quantity attenuation mechanism to avoid control drift caused by expired coordination quantities.

[0060] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0061] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A frequency stability control method for AC microgrids based on Multi-Agent, characterized in that, Includes the following steps: Construct a photovoltaic storage agent, a grid connection interface agent, and a controllable load agent; During each control cycle, each Agent collects a local state vector with a timestamp. Each Agent sends its local state vector to its neighboring Agents and calculates the frequency consistency error and power sharing error based on the set of valid neighbors. The photovoltaic-storage agent generates an unweighted pre-compensation quantity based on the power consistency error, frequency consistency error, and frequency change rate estimate. It freezes the sign of the unweighted pre-compensation quantity in the current control cycle. Based on the frozen sign, it determines the agent's participation capability in the discharge support direction or power absorption direction. Then, it generates capacity weights based on the comprehensive calculation results of converter current margin, communication status, and capability indicators. The unweighted pre-compensation amount is weighted according to the capacity weight, and the weighted result is saturated and limited based on the available active power capacity to generate a coordinated active power compensation amount. A disturbance identifier is used to determine the operating mode at each optical storage agent access point, and the determination results are used collaboratively by the grid connection interface agent and the controllable load agent. In the virtual synchronous motor control, the photovoltaic-storage agent corrects the virtual moment of inertia, damping coefficient, and cooperative active power compensation in parallel within the same control cycle according to the current physical state, and uses the corrected power reference value as the active power loop input of the virtual synchronous motor to generate the grid-connected inverter modulation signal. When the communication between the optical storage agent and its neighboring agents experiences a delay exceeding the limit or is interrupted, the optical storage agent that loses communication maintains local virtual synchronous motor control and gradually reduces its most recent collaborative compensation amount to zero according to a preset method, while the agents that have not lost communication redistribute the capacity weights according to the remaining effective communication set; when the sum of the comprehensive capability indicators in the corresponding direction is lower than the preset activation threshold, the collaborative compensation in that direction is set to zero and falls back to local virtual synchronous motor control.

2. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, The photovoltaic-storage agent includes a photovoltaic unit, an energy storage unit, a DC-DC converter unit, a grid-connected inverter unit, a sampling and communication unit, and a local control unit.

3. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, When the unweighted pre-compensation values ​​of multiple photovoltaic storage agents have inconsistent signs, the positive discharge support set and the negative absorption power set are normalized respectively within the effective communication range.

4. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, The virtual synchronous motor active power loop of the photovoltaic-storage agent satisfies: Where the subscript i represents the i-th optical storage Agent, t is the time, and J i Let ω be the virtual rotational inertia of the i-th optical storage agent, ω0 be the rated angular velocity, and ω i Let dω be the virtual angular velocity of the i-th optical storage agent. i / dt is the rate of change of the virtual angular velocity with respect to time, P i ref The active power reference value is ΔP. i MA P represents the coordinated active power compensation amount after weighting and saturation limiting. i e For grid-connected active power, D i θ is the damping coefficient. i For the virtual phase angle, dθ i / dt is the rate of change of the virtual phase angle with respect to time; the reactive voltage loop satisfies: , of which E i The virtual electromotive force is E0, which is the rated virtual electromotive force, and K is the virtual electromotive force. q Q is the reactive power-voltage droop factor. i ref Q is the reactive power reference value. i For grid-connected reactive power, ΔE i MA This refers to the reactive voltage compensation amount for multi-agent collaborative operation.

5. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, The effective neighbor set, frequency consistency error, power sharing error, and unweighted pre-compensation amount satisfy the following: ； Where k is the control cycle number, i is the current Agent number, j is the j-th neighboring Agent number, l is the index of the adjacency weight normalized summation, and N i Let N be the set of neighboring agents of the i-th agent. i v (k) is the set of effective neighbors of the i-th Agent during the k-th control period, t i (k) is the local state timestamp of the i-th Agent in the k-th control cycle, used to determine whether adjacent state variables have expired; t j (k) is the timestamp of the state quantity of the j-th adjacent Agent, T age The maximum communication age threshold allowed for adjacent state variables to participate in the calculation; a ij Let a be the original adjacency weight between the i-th Agent and the j-th Agent. il Let be the original adjacency weight between the i-th Agent and the l-th valid neighbor. To remove expired neighbors in N i v (k) Effective adjacency weights obtained by internal renormalization, where Σ represents the summation over the Agent within the specified effective neighbor set. It is an empty set; e ω,i For frequency consistency error, b i Let ωi be the traction weight of the i-th Agent with respect to the rated angular velocity, where ω0 is the rated angular velocity, and ωi is the traction weight of the i-th Agent with respect to the rated angular velocity. i and ω j Let be the virtual angular velocities of the i-th Agent and the j-th Agent, respectively. For power sharing error, P i and P j S represents the grid-connected active power of the i-th Agent and the j-th Agent, respectively. i and S j The rated active capacity or power base values ​​of the i-th Agent and the j-th Agent are respectively, ΔP i pre K is the unweighted pre-compensation amount. ω,i K P,i and K r,i These are the frequency consistency error gain, power sharing error gain, and frequency change rate suppression gain, respectively. i This is the estimated rate of change of the frequency of the i-th Agent; when When the summation term involving adjacent Agents is 0, if The corresponding adjacency weight is set to zero in this control cycle. It is obtained solely by renormalizing the remaining valid neighbors.

6. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, The available active power capacity and available power absorption capacity of the discharge support meet the following requirements: Among them, I i,max The maximum allowable current for the i-th converter is... For the first The rated active power capacity of each photovoltaic storage agent. and These are the energy storage discharge power and the rechargeable power, respectively. The active power margin can be increased for photovoltaic systems. This is the conversion factor from current to active power. The rated active power margin of the grid-connected inverter unit, This represents the DC-side power absorption margin; when any capacity boundary is negative or the current margin is negative... Set the capacity of the corresponding direction to 0.

7. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 6, characterized in that, Available power factor, SOC factor, converter current margin factor, and communication age factor satisfy the following: ； Among them, sat [0,1] (·) is a saturation limiting function used to restrict the variable within the brackets to the interval [0,1]; when the variable within the brackets is less than 0, it takes the value 0; when it is greater than 1, it takes the value 1; when it is between 0 and 1, it takes the original value. This is the unweighted pre-compensation amount. State of Charge (SOC) for energy storage min The minimum state of charge (SOC) allowed for energy storage low To reduce the weight of the low SOC threshold in the discharge support direction, SOC high To reduce the weight of the high SOC threshold in the direction of absorbed power, SOC max The maximum allowable state of charge (SOC) for energy storage. min <SOC low <SOC high <SOC max Used to ensure that the denominator of the SOC piecewise function is positive, and to form the lower discharge limit protection range and the upper charge limit protection range; I i,max For the maximum allowable current of the i-th converter, I i,lim I is the warning or limiting threshold for the i-th converter current margin to begin to decay. i,lim i,max Used to ensure that the denominator of the current margin factor is positive; T warn T is the warning threshold for communication age. age τ is the threshold for determining communication aging. i (k) represents the communication age of the most recent valid communication of the i-th Agent; 0 <T warn <T age This is used to ensure that the denominator of the communication age coefficient is positive, and to gradually transition the communication status from normal and early warning to exiting collaboration due to exceeding the age limit.​ 8. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, Capacity weights are first determined by unnormalized capacity indicators. Calculate, then normalize within the corresponding direction set according to the freezing direction of the unweighted pre-compensated amount, satisfying: ； Where, ρ i The unnormalized capability index of the i-th optical-storage agent is determined by the SOC status, directional available power, converter current margin, and communication age. For the first A set of valid communication agents that a single optical storage agent can obtain, including itself; ρ min Enable threshold for weight normalization. This is used to avoid distorted weight calculations caused by excessively small summations of capability indices within the corresponding directional set; when Or the normalized denominator in the corresponding direction is not greater than season ; The freezing direction is determined in the current control cycle and recalculated in the next control cycle; ψ SOC,i ψ is the SOC coefficient of the i-th photovoltaic-storage agent, used to characterize the degree to which the energy storage state of charge can participate in the coordinated active power allocation; P,i ψ is the directional available power coefficient for the i-th photovoltaic-storage agent, used to characterize the available active power capacity margin in the discharge support direction or the power absorption direction; I,i ψ is the converter current margin coefficient for the i-th photovoltaic-storage agent, used to characterize the margin of the current distance's maximum allowable current; H,i Let be the communication age coefficient of the i-th Agent, used to characterize the freshness of the most recent valid communication data; , , and All coefficients are dimensionless coefficients in the interval [0,1]. The larger the value, the more suitable the i-th Agent is to participate in collaborative compensation in the current direction. When any coefficient approaches 0, the corresponding Agent's unnormalized capability index ρ i It then decreases.

9. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, Estimated rate of change of frequency satisfy: ， The virtual moment of inertia and damping coefficient are updated as follows: Where k is the control cycle number, T s To control the sampling period; Δf i (k) represents the frequency deviation of the i-th optical storage agent in the k-th control cycle, Δf i (k-1) represents the frequency deviation of the previous control cycle; r i (k) represents the estimated frequency change rate of the i-th optical storage agent in the k-th control cycle, r i (k-1) is the estimated rate of frequency change in the previous control cycle; α r The first-order filter coefficients for the rate of change of frequency, 0 < α r <1 is used as a trade-off between response speed and noise suppression; and J are the saturation limiting functions for the virtual moment of inertia and the damping coefficient, respectively. i,min J i,max D represents the lower and upper limits of the virtual rotational inertia of the i-th optical storage agent, respectively. i,min D i,max J represents the lower and upper limits of the damping coefficient for the i-th optical-storage agent, respectively; i0 and D i0 These are the baseline values ​​for the virtual moment of inertia and the damping coefficient, ΔJ, respectively. i and ΔD i These are the inertia adjustment and damping adjustment amounts obtained from fuzzy inference, respectively, k J k D The local blur adjustment gain, k JM k DM These represent the synergistic gains from capacity weighting in inertia enhancement and damping enhancement, respectively; w i Let |r be the capacity normalization weight for the i-th optical storage agent. i | represents the absolute value of the estimated rate of change of frequency, |Δf i | represents the absolute value of the frequency deviation; δ f δ is the frequency deviation dead zone threshold. r The dead zone threshold for the rate of change of frequency, when |Δf i |<δ f And |r i |<δ r At that time, it is assumed that the frequency deviation and rate of change are within the allowable small range, and let ΔJ i =0、ΔD i =0 to avoid parameter jitter, and continue with J i and D i Apply a rate of change limit.

10. The frequency stability control method for AC microgrids based on Multi-Agent according to claim 1, characterized in that, The disturbance identifier and communication loss handling satisfy the following: (1) If , , and Then it enters the small interference mode; where Δf i Let r be the frequency deviation of the i-th optical storage agent access point. i The estimated rate of change of frequency, ΔU i For voltage deviation, ΔP i For active power deviation, F1, R1, U1 and P1 are the frequency deviation, frequency change rate, voltage deviation and active power deviation judgment threshold under small disturbance mode, respectively. (2) If , , , If any of the conditions in are met, or If the condition is met, the system enters a high-interference mode; where F2, R2, U2, and P2 are the corresponding judgment thresholds under the high-interference mode. , , , And set the hold time. ; (3) In the large interference mode, when Prioritize maintaining current limiting and voltage support, and limit , and The increment, of which I i,max The maximum allowable current for the i-th converter is... For virtual rotational inertia, The damping coefficient is... This refers to the active power compensation amount for multi-agent collaboration. (4) If the communication of this Agent is interrupted, the most recent collaborative active power compensation amount will be calculated as follows: Attenuation, in which ; (5) When the energy storage unit includes lithium batteries and supercapacitors, the required energy storage compensation power will be... Decomposed into low-frequency components and high frequency components .

Images