A Method and System for Cooperative Stability Control of Grid-Based Photovoltaic-Storage Microgrids Based on Impedance Adaptive Tuning
By establishing a quantitative mapping between the impedance model and virtual admittance parameters of a grid-type photovoltaic inverter, and combining this with an intelligent optimization algorithm to tune the virtual admittance parameters online, the problems of complex impedance modeling and poor subsynchronous oscillation suppression in grid-type photovoltaic-storage microgrids are solved, and adaptive stability control under multiple operating conditions is achieved.
Patent Information
- Application Number
- CN202611431767.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-15
- Publication Date
- 2026-11-17
AI Technical Summary
The impedance modeling of existing grid-type photovoltaic-storage microgrids is complex and difficult to apply online. Fixed-parameter virtual admittance cannot adapt to changes in grid operating conditions, and the subsynchronous oscillation suppression effect is poor. Furthermore, the existing methods lack a quantitative mapping relationship between control parameters and real-time oscillation characteristics of the system, making it difficult to predict the effectiveness of suppression strategies under complex operating conditions.
An impedance model for a grid-connected photovoltaic inverter with virtual admittance is established. Based on the generalized Nyquist stability criterion, the subsynchronous oscillation mode and damping ratio are identified online. An intelligent optimization algorithm is introduced through a quantized mapping relationship, and the virtual admittance parameters are tuned online using the soft actor-critic SAC algorithm to achieve adaptive adjustment of the virtual admittance parameters.
It achieves adaptive suppression of subsynchronous oscillations in multi-subgrid photovoltaic-storage microgrids, improves the stability margin and adaptability of the system under strong and weak grid switching and multiple operating conditions, overcomes the limitations of traditional methods, and enhances engineering practicality.
Smart Images

Figure CN123121804A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of coordinated control of grid-connected photovoltaic-storage microgrids in power systems, and specifically to a coordinated stability control method and system for grid-connected photovoltaic-storage microgrids based on impedance adaptive tuning. Background Technology
[0002] Currently, renewable energy sources, represented by photovoltaics, are being integrated into the power system on a large scale, and the "dual high" characteristics of high proportion of new energy and high proportion of power electronic equipment are becoming increasingly prominent. Unlike grid-following control, which relies on grid voltage for synchronization, grid-forming control can actively construct grid voltage and frequency, providing inertia and damping support for the system, and has become an important development direction in the grid connection of new energy sources such as photovoltaic power generation. Grid-forming photovoltaic-storage microgrids, composed of photovoltaics, energy storage, and local loads, are also gradually becoming an important organizational form for the consumption of distributed new energy.
[0003] However, grid-connected converters exhibit multi-timescale control interactions among their internal power synchronization loop, voltage loop, current loop, and phase-locked / synchronization components. When coupled with grid impedance, these interactions can easily induce broadband oscillations in the frequency range of several hertz to hundreds of hertz, with subsynchronous oscillations in the range of several hertz to tens of hertz being particularly prominent. These oscillations severely threaten the safe and stable operation of both the converter and the grid. Impedance analysis, due to its ability to decouple the converter from the grid and its clear physical meaning, has become an important tool for oscillation stability analysis. While existing research has established impedance models for grid-connected converters containing virtual admittance elements, it primarily focuses on the output impedance characteristics of a single inverter, failing to fully consider the impact of power interactions between subgrids on oscillation characteristics in scenarios involving multiple subgrids interconnected via flexible DC. Furthermore, the modeling process relies on detailed system control parameters, resulting in complex models that are difficult to identify and apply online.
[0004] In terms of oscillation suppression, existing subsynchronous oscillation suppression strategies are mainly divided into two categories. The first category is the virtual admittance (impedance reshaping) parameter optimization method, which reshapes the inverter output impedance by introducing virtual resistance and virtual inductance in the control loop, thereby improving system damping. For example, Chinese patent document CN117713085A discloses a power oscillation suppression method and device for grid-connected converters based on virtual admittance, which achieves oscillation suppression through the coordinated action of voltage and current controllers, virtual admittance links, and drooping active power controllers; however, the virtual admittance parameters of this type of method are fixed once tuned offline, making it difficult to adaptively adjust with changes in grid strength and operating conditions. When the system switches from a weak grid to a strong grid or when photovoltaic output fluctuates drastically, the fixed parameters cannot guarantee stability margin under all operating conditions. The second category is supplementary damping control methods. For example, Chinese patent document CN115483689A discloses a method for suppressing subsynchronous oscillations in a photovoltaic system transmitted via LCC-HVDC, which suppresses subsynchronous oscillations through sliding mode supplementary control; Chinese patent document CN116613749A discloses a method for suppressing subsynchronous oscillations in a photovoltaic grid-connected system based on data-driven rolling identification and model predictive control. However, supplementary damping control generally suffers from technical limitations such as difficulty in tuning phase compensation parameters and the inability of a single suppression channel to cope with time-varying oscillation frequencies. Furthermore, its suppression effect depends on the accurate model or online identification model of the controlled object, limiting its engineering adaptability.
[0005] Furthermore, some existing technologies attempt to introduce intelligent algorithms to optimize oscillation suppression parameters. For example, Chinese patent document CN118889391A discloses a subsynchronous oscillation suppression method based on deep Q-networks, which uses deep reinforcement learning to adjust the parameters of the additional damping controller. However, its action space is discretized, making it difficult to achieve fine tuning of continuous control parameters such as virtual admittance. Moreover, its optimization object is still the additional damping channel, failing to address the oscillation suppression problem from the root cause of inverter impedance reshaping. Overall, existing methods lack a quantitative mapping relationship between control parameters and the real-time oscillation characteristics of the system. Parameter tuning and online oscillation identification are disconnected, making it difficult to predict the suppression effect of the suppression strategy under complex operating conditions such as strong and weak grid switching and multi-subgrid collaborative operation.
[0006] In summary, the key technical challenge in this field is how to establish a broadband impedance model for grid-connected photovoltaic inverters with virtual admittance, achieve online identification of subsynchronous oscillation modes and damping ratios, establish a quantitative mapping relationship between virtual admittance parameters and oscillation damping characteristics, and introduce intelligent optimization algorithms to achieve adaptive online tuning of virtual admittance parameters so that grid-connected photovoltaic-storage microgrids can effectively suppress subsynchronous oscillations under multiple operating conditions. Summary of the Invention
[0007] To address the technical problems of complex impedance modeling and difficulty in online application in existing grid-type photovoltaic-storage microgrids, as well as the inability of fixed-parameter virtual admittance to adapt to changes in grid operating conditions and poor subsynchronous oscillation suppression, this invention provides a collaborative stability control method and system for grid-type photovoltaic-storage microgrids based on impedance adaptive tuning. This method establishes an impedance model of the grid-type photovoltaic inverter containing a virtual admittance element, identifies subsynchronous oscillation modes and damping ratios online based on the generalized Nyquist stability criterion, and establishes a quantitative mapping relationship between virtual admittance parameters and oscillation damping characteristics. Then, using the real-time identified oscillation frequency and damping ratio as input, an improved SAC algorithm is used to tune the optimal virtual admittance parameters online, achieving adaptive suppression of broadband oscillations. This aims to improve the stability margin and adaptability of multi-subgrid grid-type photovoltaic-storage microgrids under strong and weak grid switching and multi-condition operation.
[0008] The technical solution adopted in this invention is as follows: A method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning includes the following steps: S1: Establish a multi-subgrid photovoltaic-storage microgrid architecture, with inertial synchronous control adopted for the grid-type photovoltaic inverters in each subgrid; S2: Establish an impedance model for a grid-type photovoltaic inverter containing a virtual admittance element, and analyze the oscillation stability of the subsynchronous frequency band based on the impedance model; S3: Based on the impedance model, establish a quantitative mapping relationship between the virtual admittance parameters of the virtual admittance element and the subsynchronous oscillation damping characteristics of the system; S4: Using the real-time identified oscillation frequency and damping ratio as input, and based on the quantization mapping relationship, the virtual admittance parameters are tuned online using the soft actor-critic SAC algorithm to achieve adaptive suppression of broadband oscillations.
[0009] S1 includes: 1.1: Construct a photovoltaic-storage microgrid topology containing n grid-type AC subgrids. Each subgrid is connected to the common DC transmission bus through interconnecting converters. Each subgrid consists of photovoltaic power generation devices, energy storage devices, and local loads. The energy storage converters and interconnecting converters in the subgrids are all connected in parallel using virtual synchronous machine voltage source mode. Here, n is the number of subgrids. 1.2: Based on the photovoltaic-storage microgrid topology constructed in 1.1, establish a subnetwork. The virtual rotor motion equation of the energy storage converter: ; ; And establish subnets The virtual rotor motion equation of the interconnected converter: ; ; in, Subnet number; For time; The rated angular frequency; The angular frequency of the AC bus voltage of the subnet; , These are the angular frequencies of the energy storage converter and the interconnect converter, respectively. , These are the phase differences between the energy storage converter, the interconnected converter, and the local load, respectively. , These are the virtual inertia and damping coefficient of the energy storage converter, respectively. , These are the virtual inertia and damping coefficient of the interconnected converter, respectively. , These are the virtual mechanical power reference values for the energy storage converter and the interconnected converter, respectively. , These are the actual electromagnetic power of the energy storage converter and the interconnect converter, respectively. 1.3: Based on the virtual rotor motion equations established in 1.2, and according to the power conservation within the subnet, the subnet is determined. The equivalent load power satisfies: ; Furthermore, based on the power conservation between each sub-network and the common DC transmission bus, the dynamic equation of the DC bus is established: ; in, , These are the output voltages of the energy storage converter and the interconnect converter, respectively. L is the equivalent load voltage of the local load; L is the equivalent line inductance between the converter and the local load. This represents the equivalent load power of the subnet. This refers to the local load power. Photovoltaic power; The DC bus voltage of the common DC transmission bus; An integrated capacitor on a common DC transmission bus.
[0010] In S1, the grid-type photovoltaic inverters in each subgrid are single-stage photovoltaic inverters, and the inertial synchronization control includes: 1.4: Real-time acquisition of measured DC-side voltage values of the single-stage photovoltaic inverter. ; 1.5: Substituting the squared deviation between the measured DC-side voltage and the reference DC-side voltage into the inertial synchronization control equation, the angular frequency of the dq rotating coordinate system is obtained: ; in, Let be the angular frequency of the dq rotating coordinate system under the Park transform; For the Laplace operator; This is the measured value of the DC-side voltage of a single-stage photovoltaic inverter; This is the reference value for the DC-side voltage of a single-stage photovoltaic inverter; The damping coefficient for inertial synchronization; These are the inertial simulation coefficients; This is the DC voltage tracking coefficient; 1.6: Adjust the power output of the single-stage photovoltaic inverter according to the angular frequency obtained in 1.5: When the photovoltaic output power increases, causing the DC side voltage to rise, the inertial synchronization control equation outputs a positive frequency deviation, which makes the inverter accelerate the output of power; conversely, it makes the inverter decelerate the output of power, so as to simulate the inertial response characteristics of the synchronous generator.
[0011] S2 includes: 2.1: Small-signal linearization is performed on the active-frequency control equation and reactive-voltage control equation of the virtual synchronous machine to obtain the small-signal model of the power outer loop: ; ; Where s is the Laplace operator; The small-signal disturbance component of the output phase angle of the virtual synchronizer; 、 These are the transfer function matrices of the frequency loop for current disturbance and voltage disturbance, respectively; 、 These are the transfer function matrices of the voltage loop for current disturbances and voltage disturbances, respectively. 、 These are the small-signal disturbance components of the grid-connected point current and grid-connected point voltage on the dq axis in the control system coordinate system, respectively. This represents the small-signal disturbance component of the inverter output voltage reference value along the dq axis in the control system coordinate system; the superscript c indicates the control system coordinate system, and the superscript indicates the reference value. 2.2: Based on the small-signal disturbance component of the inverter output voltage reference value obtained in 2.1, the virtual admittance element is linearized using small-signal method to obtain the small-signal expression for the inner current loop reference value: ; ; in, The small-signal disturbance component of the inner current loop reference value of the dq axis in the control system coordinate system; The transfer function matrix of the virtual admittance element; The transfer function matrix of the voltage feedback low-pass filter in the virtual admittance stage; , These are the virtual resistance and virtual inductance of the virtual admittance, respectively; 2.3: Based on the current inner loop reference value obtained in 2.2, combined with the small-signal equation of the current inner loop PI control: ; Based on the steady-state phase angle difference between the control system coordinate system and the electrical system coordinate system, a small-signal transformation relationship between the two coordinate systems is established: ; ; By combining the equations in 2.1 to 2.3, the dq impedance model of the grid-type photovoltaic inverter is obtained; in, This refers to the small-signal disturbance component of the modulated voltage reference value at the inverter output on the dq axis in the system coordinate system. This is the transfer function matrix of the current inner-loop PI controller; The rated angular frequency; For inverter-side filter inductance; This refers to the steady-state phase angle difference between the control system coordinate system and the electrical system coordinate system. , These are the small-signal disturbance components of the grid-connected point voltage and grid-connected point current along the dq axis in the electrical system coordinate system, respectively; the superscript 's' indicates the electrical system coordinate system. , , This is the coordinate transformation matrix.
[0012] S2 further includes: S2.4: Convert the dq impedance model obtained in S2.3 into equivalent single-input single-output positive-sequence impedance and negative-sequence impedance: ; ; in, , These are the equivalent single-input single-output positive-sequence impedance and negative-sequence impedance of a grid-connected photovoltaic inverter, respectively. , These are the Laplace operators after positive and negative frequency shifts, respectively; , , , These are the positive-positive sequence component, positive-negative sequence coupling component, negative-positive sequence coupling component, and negative-negative sequence component of the improved sequence impedance of a grid-connected photovoltaic inverter, respectively. , , , These represent the positive-positive sequence component, positive-negative sequence coupled component, negative-positive sequence coupled component, and negative-negative sequence component of the improved power grid impedance, respectively; the superscripts p and n in the impedance symbol are the identifiers of the positive and negative sequence components, respectively, and do not represent independent parameters; S2.5: Based on the equivalent single-input single-output positive-sequence impedance and negative-sequence impedance obtained in S2.4, construct the sequence impedance matrix and define the system hysteresis matrix: ; in: The system back-ratio matrix; 、 These are the sequence impedance matrix of the grid-connected photovoltaic inverter and the grid sequence impedance matrix, respectively, which are composed of positive sequence impedance and negative sequence impedance; j is the imaginary unit.
[0013] S3 includes: 3.1: Define the system loop impedance as the sum of the equivalent single-input single-output positive-sequence impedance and the equivalent impedance of the power grid: ; in, The system loop impedance; The equivalent impedance of the power grid; 3.2: Based on the system loop impedance defined in 3.1, the damping ratio of the system in the subsynchronous frequency band is determined by the real and imaginary parts of the system loop impedance: ; in, For the system at frequency Damping ratio at the point; For frequency; , These represent the operations of taking the real part and the operations of taking the imaginary part, respectively. The imaginary unit; Pi; 3.3: Based on the damping ratio determined in 3.2, when the imaginary part of the system loop impedance crosses zero at the resonant frequency, it satisfies... And satisfy At that time, the system was determined to exhibit negative damping, indicating a risk of subsynchronous oscillation instability. Conversely, the system is judged to exhibit positive damping; and with As critical boundary conditions for stability margin, the adjustable ranges of the virtual resistance, virtual inductance, and low-pass filter cutoff frequency are determined, and the quantitative mapping relationship between the virtual admittance parameters and the subsynchronous oscillation damping characteristics of the system is obtained. It is the resonant frequency corresponding to the imaginary part of the system loop impedance crossing zero.
[0014] S4 includes: 4.1: Treat each subnet as an agent and construct the state space of agent i: ; in, The state of agent i; The deviation between the AC frequency of agent i and the rated frequency; DC bus voltage Deviation from the rated voltage of the DC bus; For integration time; 4.2: Based on the state space constructed in 4.1, a reward function is constructed with the goal of minimizing the system frequency deviation and voltage deviation, and a penalty term is introduced to constrain the system to operate within the stability boundary: ; in, The cumulative reward of agent i during the reward statistics period; , This is the reward coefficient; The rated voltage of the DC bus is T; T is the duration of reward statistics. The penalty for agent i is imposed when the combination of virtual admittance parameters causes the system to exceed the stability boundary; 4.3: Based on the reward function constructed in 4.2, the SAC algorithm is used to solve the weighted objective of maximizing the cumulative expected reward and the action distribution entropy, and the optimal virtual admittance parameters are output online. The objective expression of the SAC algorithm is: ; in, Here, E represents the objective function of the SAC algorithm; E is the expected value operation. , These represent the state and action at time t, respectively. Let be the immediate reward function after the action is performed at time t; Discount factor; H is the temperature coefficient; H is the entropy function of the strategy. For the strategy in the state The probability distribution of the next action selection.
[0015] In section 4.3, the SAC algorithm is used to tune the optimal parameters of the virtual admittance online, including: The meta-learning mechanism is integrated into the SAC algorithm to obtain the improved SAC algorithm. The improved SAC algorithm treats each microgrid operation scenario as an independent task, updates network parameters under a specific operation scenario through an inner loop, and optimizes the initial network parameters by aggregating the losses of multiple operation scenarios through an outer loop, so that the network can quickly adapt to new operation scenarios. The objective function of the improved SAC algorithm is: ; in, To improve the task objective function of the SAC algorithm within the meta-learning framework; Let be the next state at time t; The value function for the next state; For value network parameters; The expected value of the value function for the next state; Let be the state transition probability.
[0016] In step S4, the real-time identified oscillation frequency and damping ratio are obtained in the following way: Perform a frequency domain scan on the system loop impedance to find the resonant frequency corresponding to when the imaginary part of the system loop impedance crosses zero. The dominant oscillation frequency of the system is determined, and this dominant oscillation frequency is substituted into the expression for the damping ratio described in section 3.2 to calculate the current dominant damping ratio of the system. The dominant oscillation frequency and the dominant damping ratio are used as real-time inputs to the SAC algorithm to drive the online tuning of the virtual admittance parameters.
[0017] A grid-based photovoltaic-storage microgrid collaborative stability control system based on impedance adaptive tuning, comprising: The architecture establishment module is used to establish a multi-subgrid photovoltaic-storage microgrid architecture and enable the grid-type photovoltaic inverters in each subgrid to adopt inertial synchronous control. Impedance modeling module is used to establish the impedance model of the grid-type photovoltaic inverter containing virtual admittance, and to analyze the oscillation stability of the subsynchronous frequency band based on the impedance model; The mapping establishment module is used to establish a quantitative mapping relationship between the virtual admittance parameters of the virtual admittance element and the subsynchronous oscillation damping characteristics of the system based on the impedance model. The parameter tuning module is used to tune the virtual admittance parameters online using the real-time identified oscillation frequency and damping ratio as inputs, based on the quantization mapping relationship and employing the soft actor-critic SAC algorithm, thereby achieving adaptive suppression of broadband oscillations.
[0018] This invention discloses a method and system for coordinated stability control of a grid-type photovoltaic-storage microgrid based on impedance adaptive tuning, with the following technical advantages: 1) Overcomes the limitations of traditional impedance modeling, which relies on detailed system control parameters and is difficult to apply online: This invention establishes an impedance model for a grid-connected photovoltaic inverter with a virtual admittance element, and analyzes the oscillation stability of the subsynchronous frequency band using the generalized Nyquist stability criterion. It can achieve online identification of the dominant oscillation mode and damping ratio in the subsynchronous frequency band without relying on all detailed control parameters of the system. This solves the technical problems of existing impedance analysis methods that focus on the output characteristics of a single inverter, have complex modeling, and are difficult to apply online, providing a feasible basis for real-time monitoring and suppression of oscillations.
[0019] 2) Overcoming the problems of fixed-parameter virtual admittance being unable to adapt to changes in grid operating conditions and insufficient stability margin across all operating conditions: This invention establishes a quantitative mapping relationship between virtual admittance parameters and the subsynchronous oscillation damping characteristics of the system, and determines the critical boundary conditions for stability margin. It reveals the adjustment law of virtual resistance, virtual inductance and voltage feedback low-pass filter cutoff frequency on the system damping characteristics, so that parameter tuning has a clear theoretical basis and stability constraints. It solves the technical problems in the prior art of lacking a quantitative mapping relationship between control parameters and real-time oscillation characteristics of the system, and the separation between parameter tuning and oscillation identification.
[0020] 3) Achieve online adaptive tuning of virtual admittance parameters as the system oscillation characteristics change: This invention uses the real-time identified oscillation frequency and damping ratio as inputs and employs the soft actor-commentator SAC algorithm to tune the virtual admittance parameters online. This allows the virtual admittance parameters to automatically match and update with changes in grid strength and fluctuations in photovoltaic output and load. It overcomes the technical limitations of existing virtual admittance parameter optimization methods, where parameters remain fixed once tuned and have poor suppression effects during switching between weak and strong grids. This ensures that the system has sufficient stability margin under multiple operating conditions.
[0021] 4) Achieving broadband oscillation suppression from the source of impedance reshaping avoids the inherent defects of additional damping control: This invention improves system damping by directly reshaping the inverter output impedance through online adjustment of virtual admittance parameters, without the need for additional damping control channels. This overcomes the technical problems of existing additional damping control methods, such as difficulty in tuning phase compensation parameters, inability of a single suppression channel to handle time-varying oscillation frequencies, and suppression effectiveness depending on the accurate model of the controlled object. At the same time, the SAC algorithm is applicable to continuous action space and can achieve fine and continuous tuning of virtual admittance parameters, overcoming the problem of insufficient tuning accuracy in existing discrete action intelligent optimization methods.
[0022] 5) Improve the convergence efficiency and multi-scenario generalization ability of the intelligent tuning algorithm: This invention integrates a meta-learning mechanism into the SAC algorithm. Through the inner loop, the network parameters are quickly updated under specific operating scenarios, and the outer loop aggregates losses from multiple scenarios to optimize the initial parameters of the network. This allows the network to converge and adapt quickly with only a small number of samples under new operating scenarios, enhancing the exploratory nature and robustness of the agent. It solves the problems of low training efficiency and poor scene transfer adaptability of existing intelligent optimization methods, and meets the engineering application requirements of online real-time tuning.
[0023] 6) Achieve coordinated and stable control of multi-subgrid photovoltaic-storage microgrids to improve engineering practicality: This invention addresses the architecture of a multi-subgrid photovoltaic-storage microgrid with flexible DC interconnection. The grid-connected photovoltaic inverters within each subgrid employ inertial synchronous control, enabling self-synchronization and grid connection using DC-side voltage fluctuation information without the need for a phase-locked loop (PLL), effectively mitigating DC-side voltage fluctuations. Simultaneously, each subgrid is treated as an intelligent agent for collaborative tuning, overcoming the shortcomings of existing research that did not adequately consider the impact of power interaction between subgrids on oscillation characteristics in multi-subgrid interconnection scenarios. This significantly improves the adaptive capability and engineering practicality of multi-subgrid photovoltaic-storage microgrids in suppressing broadband oscillations during strong / weak grid switching and multi-subgrid collaborative operation scenarios, providing a highly practical technical solution for broadband oscillation management in grid-connected renewable energy systems. Attached Figure Description
[0024] The present invention will be further described below with reference to the accompanying drawings and examples; Figure 1 The topology diagram of a multi-subgrid photovoltaic-storage microgrid with flexible DC interconnection provided in an embodiment of the present invention is shown.
[0025] Figure 2 The topology diagram of a single-stage photovoltaic inverter grid-connected system provided in an embodiment of the present invention is shown.
[0026] Figure 3 The graph shows the average round reward comparison between the improved SAC algorithm and the original SAC algorithm provided in this embodiment of the invention.
[0027] Figure 4 A comparison chart verifying the accuracy of the impedance model of the grid-type photovoltaic inverter provided in the embodiments of the present invention.
[0028] Figure 5 The Nyquist curves show a comparison of system stability before and after adaptive tuning of the virtual admittance parameters provided in this embodiment of the invention.
[0029] Figure 6 The overall flowchart of the coordinated stability control method for grid-type photovoltaic-storage microgrids based on impedance adaptive tuning provided in the embodiments of the present invention is shown. Detailed Implementation
[0030] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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.
[0031] Example 1: This embodiment provides a method for coordinated stability control of a grid-type photovoltaic-storage microgrid based on impedance adaptive tuning. Figure 6 The overall flowchart of the coordinated stability control method for grid-type photovoltaic-storage microgrids based on impedance adaptive tuning provided in this embodiment is as follows: Figure 6 As shown, the overall process of this method is as follows: First, a multi-subgrid photovoltaic-storage microgrid architecture with flexible DC interconnection is established. The grid-type photovoltaic inverters in each subgrid adopt inertial synchronous control to cope with DC side voltage fluctuations. Then, an impedance model of the grid-type photovoltaic inverter with virtual admittance is established, and the oscillation stability of the subsynchronous frequency band is analyzed based on the impedance model. Next, a quantitative mapping relationship between the virtual admittance parameters and the damping characteristics of the subsynchronous oscillation of the system is established based on the impedance model, and the stability margin boundary conditions of the virtual admittance parameters are determined. Then, with the real-time identified oscillation frequency and damping ratio as input, the virtual admittance parameters are tuned online according to the quantitative mapping relationship using the soft actor-commentator SAC algorithm, and the tuned values are sent to the grid-type photovoltaic inverters of each subgrid to achieve adaptive suppression of broadband oscillations. Finally, a MATLAB / Simulink time-domain simulation platform is built to verify the effectiveness of the adaptive suppression strategy in the scenarios of strong and weak grid switching and multi-subgrid collaborative operation.
[0032] Step S1: Establish a multi-subgrid photovoltaic-storage microgrid architecture with flexible DC interconnection. The grid-type photovoltaic inverters in each subgrid adopt inertial synchronous control to cope with DC side voltage fluctuations.
[0033] With the large-scale integration of renewable energy sources such as photovoltaics, the high proportion of new energy sources has led to a high degree of power electronics characteristics in the power system. Grid-based control, which can actively construct grid voltage and frequency, has become an important development direction in the photovoltaic power generation field. The grid-based photovoltaic-storage microgrid system in this embodiment includes n grid-based AC subgrids, each connected to the common DC transmission bus via interconnecting converters. Each subgrid consists of photovoltaic power generation devices, energy storage devices, and local loads. The energy storage converters and interconnecting converters within the subgrids operate in parallel using a virtual synchronous machine voltage source mode, providing voltage and frequency support for the subgrid. Figure 1 This is a topology diagram of a multi-subgrid photovoltaic-storage microgrid with flexible DC interconnection in this embodiment, where n is the number of subgrids.
[0034] For the energy storage converter in subnet i, its virtual rotor motion equation describes the supporting effect of virtual inertia and damping coefficient on frequency dynamics, and can be expressed as: ; ; Meanwhile, subnet i is connected to the common DC transmission bus via an interconnecting converter, and the virtual rotor motion equation of the interconnecting converter can be expressed as: ; ; Where i is the subnet number; t is the time; The rated angular frequency; The angular frequency of the AC bus voltage of the subnet; , These are the angular frequencies of the energy storage converter and the interconnect converter, respectively. , These are the phase differences between the energy storage converter, the interconnected converter, and the local load, respectively. , These are the virtual inertia and damping coefficient of the energy storage converter, respectively. , These are the virtual inertia and damping coefficient of the interconnected converter, respectively. , These are the virtual mechanical power reference values for the energy storage converter and the interconnected converter, respectively. , These are the actual electromagnetic power of the energy storage converter and the interconnect converter, respectively.
[0035] According to the law of conservation of power, the sum of the electromagnetic power output by the energy storage converter and the interconnecting converter within the subnet equals the equivalent load power of the subnet, which can be expressed as: ; in, , These are the output voltages of the energy storage converter and the interconnect converter, respectively. L is the equivalent load voltage of the local load; L is the equivalent line inductance between the converter and the local load. This represents the equivalent load power of the subnet. This refers to the local load power. Photovoltaic power.
[0036] All subnets are connected to the common DC transmission bus through their respective interconnecting converters. Based on power conservation, the dynamic equation of the DC bus can be obtained as follows: ; in, The DC bus voltage of the common DC transmission bus; This refers to the integrated capacitor on the common DC transmission bus. As can be seen from the DC bus dynamic equation, the electromagnetic power output by the interconnected converters of each sub-network jointly determines the dynamic change of the DC bus voltage. Power fluctuations in any sub-network will be coupled to other sub-networks through the common DC transmission bus; therefore, coordinated stability control is required among multiple sub-networks.
[0037] Figure 2 This is a topology diagram of the grid-connected system of a single-stage photovoltaic inverter in this embodiment. The output power of photovoltaic cells is random, and the power fluctuations caused by maximum power point tracking directly manifest as fluctuations in the DC-side voltage. The single-stage photovoltaic inverter eliminates a DC-DC conversion stage, but also loses the isolation and regulation capability of the preceding converter for the DC-side voltage; DC-side voltage fluctuations are directly transmitted to the grid-connected control stage. Therefore, in this embodiment, the grid-connected photovoltaic inverters in each subgrid use single-stage photovoltaic inverters and employ a synchronous control scheme based on DC capacitor inertia. This eliminates the need for a phase-locked loop and directly utilizes the DC-side voltage fluctuation information to achieve self-synchronization and grid connection. The inertial synchronous control equation can be expressed as: ; in, ω is the angular frequency of the dq rotating coordinate system under the Park transform; s is the Laplace operator; This is the measured value of the DC-side voltage of a single-stage photovoltaic inverter; This is the reference value for the DC-side voltage of a single-stage photovoltaic inverter; The damping coefficient for inertial synchronization; These are the inertial simulation coefficients; This is the DC voltage tracking coefficient.
[0038] The principle of the above-mentioned inertial synchronization control is as follows: when the photovoltaic output power increases, the measured value of the DC side voltage increases. Higher than the DC side voltage reference value At that time, the inertial synchronization control equation outputs a positive angular frequency deviation. This causes the single-stage photovoltaic inverter to accelerate its output power, leading to a drop in the DC-side voltage; when the photovoltaic output power decreases, the measured DC-side voltage value... Below the DC side voltage reference value At that time, the inertial synchronization control equation outputs a negative angular frequency deviation. This causes the single-stage photovoltaic inverter to slow down its output power, prompting the DC-side voltage to rise. Among these factors, the inertial simulation coefficient... Rotor inertia used to simulate synchronous generators; damping coefficient for inertial synchronization. DC voltage tracking coefficient is used to suppress oscillations during angular frequency adjustment. The tracking speed is used to adjust the DC-side voltage deviation, thereby enabling the single-stage photovoltaic inverter to simulate the inertial response characteristics of a synchronous generator to cope with DC-side voltage fluctuations.
[0039] Step S2: Establish an impedance model of the grid-type photovoltaic inverter with virtual admittance, and analyze the oscillation stability of the subsynchronous frequency band based on the impedance model.
[0040] The power synchronization loop, voltage loop, and current loop within a grid-connected photovoltaic inverter exhibit multi-timescale control interactions, which, when coupled with grid impedance, can easily trigger oscillations in the subsynchronous frequency band. Impedance analysis, which decouples the inverter from the grid and provides clear physical interpretation, is an effective tool for oscillation stability analysis. This step introduces a virtual admittance element into the control loop. By constructing a virtual impedance characteristic between the inverter's output voltage reference value and the inner current loop reference value, the inverter's output impedance is reshaped, providing adjustable control degrees of freedom for subsequent oscillation suppression.
[0041] First, by performing small-signal linearization on the active-frequency control equation and reactive-voltage control equation of the virtual synchronous machine, the small-signal model of the power outer loop can be obtained: ; ; in, The small-signal disturbance component of the output phase angle of the virtual synchronizer; 、 These are the transfer function matrices of the frequency loop for current disturbance and voltage disturbance, respectively; 、 These are the transfer function matrices of the voltage loop for current disturbances and voltage disturbances, respectively. 、 These are the small-signal disturbance components of the grid-connected point current and grid-connected point voltage on the dq axis in the control system coordinate system, respectively. This represents the small-signal disturbance component of the inverter output voltage reference value along the dq axis in the control system coordinate system; the superscript 'c' indicates the control system coordinate system, and the superscript indicates the reference value.
[0042] Furthermore, small-signal linearization is performed on the virtual admittance element to obtain the small-signal expression for the inner current loop reference value: ; ; in, The small-signal disturbance component of the inner current loop reference value of the dq axis in the control system coordinate system; The transfer function matrix of the virtual admittance element; The transfer function matrix of the voltage feedback low-pass filter in the virtual admittance stage; , These represent the virtual resistance and virtual inductance, respectively, for the virtual admittance. A voltage feedback low-pass filter can be a first-order low-pass filter, and its transfer function matrix is related to the cutoff frequency as follows: ; ; in, This is the cutoff angular frequency of the voltage feedback low-pass filter; This is the cutoff frequency of the voltage feedback low-pass filter; Pi is the mathematical constant. A voltage feedback low-pass filter is used to filter out high-frequency components in the grid connection point voltage sampling, preventing high-frequency noise from being amplified by the virtual admittance stage and entering the inner current loop.
[0043] The inner current loop uses PI control, and its small-signal equation can be expressed as: ; in, This refers to the small-signal disturbance component of the modulated voltage reference value at the inverter output on the dq axis in the system coordinate system. This is the transfer function matrix of the current inner-loop PI controller; For inverter-side filter inductance; This is the decoupling term for the inner current loop.
[0044] It should be noted that the control system of a grid-connected photovoltaic inverter uses its own synchronization link to establish a dq coordinate system as the control reference, while the electrical quantities of the power grid use a dq coordinate system established by the grid voltage as the reference. A steady-state phase angle difference exists between the two coordinate systems. Considering the steady-state phase angle difference between the control system coordinate system and the electrical system coordinate system, the voltage and current small-signal conversion relationship between the two coordinate systems is established as follows: ; ; in, This refers to the steady-state phase angle difference between the control system coordinate system and the electrical system coordinate system. 、 These are the small-signal disturbance components of the grid-connected point voltage and grid-connected point current along the dq axis in the electrical system coordinate system, respectively; the superscript 's' indicates the electrical system coordinate system. 、 , Let be the coordinate transformation matrix. By simultaneously establishing the small-signal equations and coordinate transformation relationships for the power outer loop, virtual admittance element, and current inner loop, the dq impedance model of the grid-type photovoltaic inverter can be obtained.
[0045] Furthermore, since positive-sequence disturbances in a three-phase AC system will induce negative-sequence responses through the inverter control circuit, i.e., there is a positive-negative-sequence frequency coupling effect, based on the harmonic linearization principle, there is a frequency-shift coupling relationship between the sequence impedance components in the dq impedance model, and the frequency-shifted Laplace operator satisfies: ; ; in, , , respectively, are the positive-sequence and negative-sequence frequency-shifted Laplace operators; j is the imaginary unit. Considering the positive and negative-sequence frequency coupling effect, the dq impedance model is transformed into the equivalent single-input single-output positive-sequence and negative-sequence impedances on the AC side of a grid-connected photovoltaic inverter: ; ; in, , These are the equivalent single-input single-output positive-sequence impedance and negative-sequence impedance of a grid-connected photovoltaic inverter, respectively. , , , These are the positive-positive sequence component, positive-negative sequence coupling component, negative-positive sequence coupling component, and negative-negative sequence component of the improved sequence impedance of a grid-connected photovoltaic inverter, respectively. , , , These represent the positive-positive sequence component, positive-negative sequence coupled component, negative-positive sequence coupled component, and negative-negative sequence component of the improved power grid impedance, respectively. The superscripts p and n in the impedance symbols identify the positive and negative sequence components, respectively, and do not represent independent parameters. This equivalent treatment transforms the dual-input, dual-output impedance relationship of positive and negative sequence coupling into a single-input, single-output form, facilitating direct application of frequency domain stability criteria for analysis.
[0046] Based on the above equivalent sequence impedance model, the stability of a grid-connected photovoltaic inverter system can be analyzed using the generalized Nyquist stability criterion. The system hysteresis matrix is defined as follows: ; in, The system back-ratio matrix; 、 These are the sequence impedance matrices of the grid-connected photovoltaic inverter and the grid sequence impedance matrix, respectively, composed of positive-sequence and negative-sequence impedances. If the Nyquist curve of the characteristic trajectory of the system's hysteresis matrix encloses... If the system reaches a certain value, it is determined that the system is unstable; otherwise, it is determined that the system is stable.
[0047] Step S3: Based on the impedance model, establish a quantitative mapping relationship between the virtual admittance parameters and the subsynchronous oscillation damping characteristics of the system.
[0048] Based on the impedance model of the grid-type photovoltaic inverter established in S2, a quantitative mapping relationship between the virtual admittance parameter and the damping ratio of the subsynchronous oscillation of the system is further established, and the stability margin boundary conditions are determined so that the parameter tuning has a clear theoretical basis and stability constraints.
[0049] The system loop impedance is defined as the sum of the equivalent single-input single-output positive-sequence impedance and the equivalent impedance of the power grid: ; in, The system loop impedance; Let be the equivalent impedance of the power grid. The damping ratio of the system in the subsynchronous frequency band is determined by both the real and imaginary parts of the system loop impedance, and can be expressed as: ; in, Let f be the damping ratio of the system at frequency f; where f is the frequency. , These represent operations that extract the real part and operations that extract the imaginary part, respectively.
[0050] When the imaginary part of the system loop impedance crosses zero at the resonant frequency, that is, when the imaginary part of the system loop impedance crosses zero, the following condition is met: At that time, the system is at the resonant frequency There exists a potential oscillation mode at this point; if the following conditions are met... Then the damping ratio If the system exhibits negative damping, it is at risk of subsynchronous oscillation and instability; conversely, if the damping ratio is positive, the system is more stable. The system exhibits positive damping, effectively suppressing oscillations. The virtual admittance parameter adjusts the damping ratio by changing the real and imaginary parts of the inverter output impedance, thereby altering the real and imaginary parts of the system loop impedance. The critical boundary condition for stability margin is... At this point, the system is in a critically stable state, and the corresponding combination of virtual admittance parameters is the critical boundary of the stability margin. The adjustment rules of the virtual admittance parameters determined based on the quantization mapping relationship are shown in Table 1.
[0051]
[0052] The aforementioned critical boundary conditions limit the adjustable range of the cutoff frequency of the virtual resistance, virtual inductance, and voltage feedback low-pass filter, directly determining the feasible domain of the virtual admittance parameter, and providing a theoretical basis and stability constraints for the online tuning of S4.
[0053] Step S4: Using the real-time identified oscillation frequency and damping ratio as input, and based on the quantization mapping relationship, the SAC algorithm is used to tune the virtual admittance parameters online to achieve adaptive suppression of wideband oscillations.
[0054] Considering the randomness of photovoltaic power generation and electricity load, the fixed parameters determined by offline optimization cannot always guarantee optimal performance during actual operation. This step introduces a deep reinforcement learning method to perform online real-time tuning of the virtual admittance parameters, which includes the following five steps.
[0055] (1) Real-time identification of oscillation frequency and damping ratio: Perform frequency domain scanning on the system loop impedance to determine the oscillation frequency and damping ratio. resonant frequency The dominant oscillation frequency of the system is determined; the dominant oscillation frequency is substituted into the expression for the damping ratio described in S3 to calculate the current dominant damping ratio of the system. This transforms the abstract problem of oscillation stability into a quantizable input that can be acquired in real time, achieving a closed-loop connection between parameter tuning and oscillation characteristics.
[0056] (2) Construction of state space, action space and reward function: Each subnet is regarded as an agent. The state space of agent i is constructed with the real-time identified oscillation frequency and damping ratio as the core: ; in, The state of agent i; The deviation between the AC frequency of agent i and the rated frequency; DC bus voltage Deviation from the rated voltage of the DC bus; The variable is the integral time variable. Frequency deviation, voltage deviation, and their integral terms are introduced into the state space simultaneously, enabling the agent to perceive both the current oscillation state and the cumulative trend of the deviation.
[0057] The action space of agent i is a combination of virtual admittance parameters: ; in, The action of agent i is defined by the critical boundary conditions determined by S3.
[0058] The reward function aims to minimize the system frequency deviation and voltage deviation, while introducing a penalty term to constrain the system to operate within the safety boundary: ; in, The cumulative reward of agent i during the reward statistics period; , This is the reward coefficient; The rated voltage of the DC bus is T; T is the duration of reward statistics. The penalty for agent i is applied when the combination of virtual admittance parameters causes the system to exceed the stability boundary, guiding the agent to always choose actions within the stable domain. The integral term in the reward function is time-weighted, which encourages the agent to prioritize eliminating deviations with longer durations, balancing the speed and stability of oscillation suppression.
[0059] (3) SAC Algorithm Solution: The Soft Actor-Critic SAC algorithm is a deep reinforcement learning algorithm applicable to continuous action spaces, capable of continuously and finely tuning the cutoff frequency of virtual resistors, virtual inductors, and voltage feedback low-pass filters. Unlike standard reinforcement learning, which only maximizes the cumulative policy reward, the SAC algorithm adds an entropy term to the objective function, maximizing the entropy of the cumulative expected reward and the action distribution, balancing reward maximization and policy randomness, and enhancing the agent's exploratory and robust properties. Its objective expression is: ; in, Here, E represents the objective function of the SAC algorithm; E is the expected value operation. , These represent the state and action at time t, respectively. Let be the immediate reward function after the action is performed at time t; Discount factor; H is the temperature coefficient; H is the entropy function of the strategy. For the strategy in the state The probability distribution of the next action selection.
[0060] (4) Improved Meta-Learning Mechanism: To improve training efficiency and generalization ability, the meta-learning mechanism is integrated into the SAC algorithm to obtain an improved SAC algorithm. After the meta-learning is integrated into the SAC framework, each microgrid operation scenario is regarded as an independent task. That is, different operation scenarios of multi-subgrid photovoltaic-storage microgrids are constructed into multiple independent tasks. The operation scenarios include strong and weak grid switching scenarios, photovoltaic power output fluctuation scenarios, and load fluctuation scenarios. The network parameters are quickly updated in specific operation scenarios through the inner loop. That is, for each independent task, the network initial parameters are updated according to the task objective function based on the sample corresponding to the independent task, starting from the initial network parameters, to obtain the task network parameters corresponding to the independent task. The network initial parameters are optimized by aggregating the losses of multiple scenarios through the outer loop. That is, the losses of multiple independent tasks are aggregated, and the network initial parameters are optimized according to the aggregated losses, so that the optimized network initial parameters are in the neighborhood of the task network parameters corresponding to each independent task, so that the network can quickly adapt to new scenarios. The task objective function of the SAC algorithm under the meta-learning framework is: ; in, The objective function is the task objective function. These are the task network parameters updated via the inner loop; Let be the next state at time t; The value function for the next state; The expected value of the value function for the next state; Let be the state transition probability. The objective function of this task maximizes the weighted sum of the cumulative reward and the policy entropy, thereby enabling the network to achieve faster convergence and better adaptation with fewer samples in the task of real-time tuning of virtual admittance parameters.
[0061] During online adaptation, the network is fine-tuned using the optimized initial parameters and samples from the current operating scenario. The fine-tuned network then outputs the optimal virtual admittance parameters online based on the real-time identified oscillation frequency and damping ratio. The main hyperparameter settings of the improved SAC algorithm are shown in Table 2.
[0062]
[0063] It should be noted that the values in Table 2 are merely exemplary values for this embodiment, and those skilled in the art can adjust them according to the actual system scale and operating conditions.
[0064] Figure 3 This is a graph comparing the average round reward of the improved SAC algorithm and the original SAC algorithm in this embodiment. After introducing the meta-learning mechanism, the improved SAC algorithm accelerates the accumulation of experience, and the training start point is advanced from 500 rounds to 300 rounds. During training, the reward value of the improved SAC algorithm rises more smoothly, the convergence speed is faster, and the average reward after convergence is significantly better than that of the original algorithm, and the fluctuation phenomenon of reward drop caused by random sampling in the original algorithm is not observed.
[0065] (5) Parameter distribution and closed-loop verification: SAC algorithm outputs virtual resistance Virtual inductance and the cutoff frequency of the voltage feedback low-pass filter The system sets the setpoint and distributes it to the virtual admittance stage of the grid-connected photovoltaic inverters in each subgrid. After updating the parameters according to the setpoint, the virtual admittance stage re-scans the system loop impedance in the frequency domain and determines the dominant oscillation frequency and dominant damping ratio. If the newly determined dominant damping ratio does not meet the requirements... When the positive damping condition is met, the SAC algorithm re-tunes the virtual admittance parameters based on the re-determined results. This closed-loop iteration allows the virtual admittance parameters to automatically match and update with the system's oscillation characteristics until the system exhibits positive damping. This overcomes the technical challenge that traditional fixed-parameter virtual admittance cannot adapt to the switching scenarios between weak and strong power grids, and achieves adaptive suppression of wideband oscillations.
[0066] Step S5: Build a MATLAB / Simulink time-domain simulation platform to verify the effectiveness of the adaptive suppression strategy under the scenarios of strong and weak power grid switching and multi-subgrid collaborative operation.
[0067] A MATLAB / Simulink time-domain simulation platform was built. In a strong-weak grid switching scenario, the grid strength switching was simulated by changing the grid's equivalent impedance. In a multi-subgrid collaborative operation scenario, the power interaction between subgrids was simulated by changing the photovoltaic output and local load of each subgrid. Under these scenarios, a fixed-parameter virtual admittance scheme and the adaptive tuning scheme of this invention were applied respectively. The oscillation performance of the system was compared to verify the effectiveness of the adaptive suppression strategy of this invention. This step relies on the impedance model of the grid-type photovoltaic inverter with virtual admittance and the generalized Nyquist stability criterion. By identifying the dominant oscillation mode and damping ratio in the subsynchronous frequency band online, a quantitative mapping relationship between the virtual admittance parameters and the system damping characteristics is established. An improved SAC algorithm is used to automatically tune the optimal virtual admittance parameters in real time to match the current oscillation mode, improving the adaptive capability and engineering practicality of the multi-subgrid photovoltaic-storage microgrid in broadband oscillation suppression.
[0068] Example 2: This embodiment provides a grid-based photovoltaic-storage microgrid collaborative stability control system based on impedance adaptive tuning, including: The architecture establishment module is used to establish a multi-subgrid photovoltaic-storage microgrid architecture and enable the grid-type photovoltaic inverters in each subgrid to adopt inertial synchronous control, specifically executing the processing described in S1 of Example 1; Impedance modeling module is used to establish the impedance model of the grid-type photovoltaic inverter containing virtual admittance, and analyze the oscillation stability of the subsynchronous frequency band based on the impedance model, specifically performing the processing described in S2 of Example 1; The mapping establishment module is used to establish a quantitative mapping relationship between the virtual admittance parameters of the virtual admittance element and the subsynchronous oscillation damping characteristics of the system based on the impedance model, specifically performing the processing described in S3 of Embodiment 1; The parameter tuning module is used to tune the virtual admittance parameters online using the real-time identified oscillation frequency and damping ratio as inputs, according to the quantization mapping relationship, and employing the soft actor-critic SAC algorithm to achieve adaptive suppression of wideband oscillations. Specifically, it performs the processing described in S4 of Embodiment 1.
[0069] The above modules can be implemented through software, hardware, or a combination of both. For example, these modules can be integrated into the controllers of the grid-type photovoltaic inverters in each subgrid, or they can be integrated into the central controller of the microgrid and send the virtual admittance parameter settings to the grid-type photovoltaic inverters in each subgrid via a communication link.
[0070] Example 3: This embodiment establishes a time-domain simulation model of a grid-type photovoltaic inverter based on MATLAB / Simulink, and verifies the theoretical impedance model established by S2 by sweep frequency measurement impedance. Figure 4 This diagram shows a comparison of the accuracy verification of the impedance model for the grid-type photovoltaic inverter in this embodiment. It presents the amplitude-frequency response curve and phase-frequency response curve of the impedance. The solid line represents the calculated result of the theoretical impedance model, and the scatter points represent the measured result of the time-domain simulated swept-frequency model. The comparison results show that the amplitude and phase characteristics of the theoretical impedance and the simulated swept-frequency impedance of the grid-type photovoltaic inverter are basically consistent over a wide frequency range, verifying the accuracy of the established impedance model.
[0071] Figure 5 This is a comparison of the Nyquist curves of system stability before and after adaptive tuning of the virtual admittance parameters in this embodiment. The frequency domain Nyquist curve is used to characterize the system stability criterion. When the imaginary part of the criterion crosses the frequency axis, the magnitude of the corresponding real part reflects the system's damping level. Before tuning, with fixed virtual admittance parameters, the imaginary part of the system crosses the frequency axis at 324Hz, and the corresponding real part is less than -1, indicating negative damping and a subsynchronous oscillation instability state. After optimizing the virtual admittance parameters using the adaptive tuning strategy of this invention, the corresponding real part increases to greater than -1, the system obtains positive damping, and the subsynchronous oscillation is effectively suppressed. The results show that the adaptive tuning strategy of virtual admittance parameters of this invention can significantly improve the subsynchronous frequency band damping characteristics of the system, and can effectively suppress subsynchronous oscillations in scenarios of strong and weak grid switching and multi-subgrid collaborative operation, providing a highly practical technical solution for broadband oscillation management in grid-connected new energy systems.
Claims
1. A network configuration type photovoltaic energy storage micro-grid cooperative stability control method based on impedance adaptive tuning, characterized in that Includes the following steps: S1: Establish a multi-subgrid photovoltaic-storage microgrid architecture, with inertial synchronous control adopted for the grid-type photovoltaic inverters in each subgrid; S2: Establish an impedance model for a grid-type photovoltaic inverter containing a virtual admittance element, and analyze the oscillation stability of the subsynchronous frequency band based on the impedance model; S3: Based on the impedance model, establish a quantitative mapping relationship between the virtual admittance parameters of the virtual admittance element and the subsynchronous oscillation damping characteristics of the system; S4: Using the real-time identified oscillation frequency and damping ratio as input, and based on the quantization mapping relationship, the virtual admittance parameters are tuned online using the soft actor-critic SAC algorithm to achieve adaptive suppression of broadband oscillations.
2. The network configuration type photovoltaic energy storage micro-grid cooperative stability control method based on impedance adaptive tuning according to claim 1, characterized in that: S1 includes: 1.1: Construct a photovoltaic-storage microgrid topology containing n grid-type AC subgrids. Each subgrid is connected to the common DC transmission bus through interconnecting converters. Each subgrid consists of photovoltaic power generation devices, energy storage devices, and local loads. The energy storage converters and interconnecting converters in the subgrids are all connected in parallel using virtual synchronous machine voltage source mode. Here, n is the number of subgrids. 1.2: Based on the photovoltaic-storage microgrid topology constructed in 1.1, establish a subnetwork. The virtual rotor motion equation of the energy storage converter: ; ; And establish subnets The virtual rotor motion equation of the interconnected converter: ; ; in, Subnet number; For time; The rated angular frequency; The angular frequency of the AC bus voltage of the subnet; , These are the angular frequencies of the energy storage converter and the interconnect converter, respectively. , These are the phase differences between the energy storage converter, the interconnected converter, and the local load, respectively. , These are the virtual inertia and damping coefficient of the energy storage converter, respectively. , These are the virtual inertia and damping coefficient of the interconnected converter, respectively. , These are the virtual mechanical power reference values for the energy storage converter and the interconnected converter, respectively. , These are the actual electromagnetic power of the energy storage converter and the interconnect converter, respectively. 1.3: Based on the virtual rotor motion equations established in 1.2, and according to the power conservation within the subnet, the subnet is determined. The equivalent load power satisfies: ; Furthermore, based on the power conservation between each sub-network and the common DC transmission bus, the dynamic equation of the DC bus is established: ; in, , These are the output voltages of the energy storage converter and the interconnect converter, respectively. L is the equivalent load voltage of the local load; L is the equivalent line inductance between the converter and the local load. This represents the equivalent load power of the subnet. This refers to the local load power. Photovoltaic power; The DC bus voltage of the common DC transmission bus; An integrated capacitor on a common DC transmission bus.
3. The method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning according to claim 2, characterized in that: In S1, the grid-type photovoltaic inverters in each subgrid are single-stage photovoltaic inverters, and the inertial synchronization control includes: 1.4: Real-time acquisition of measured DC-side voltage values of the single-stage photovoltaic inverter. ; 1.5: Substituting the squared deviation between the measured DC-side voltage and the reference DC-side voltage into the inertial synchronization control equation, the angular frequency of the dq rotating coordinate system is obtained: ; in, Let be the angular frequency of the dq rotating coordinate system under the Park transform; For the Laplace operator; This is the measured value of the DC-side voltage of a single-stage photovoltaic inverter; This is the reference value for the DC-side voltage of a single-stage photovoltaic inverter; The damping coefficient for inertial synchronization; These are the inertial simulation coefficients; This is the DC voltage tracking coefficient; 1.6: Adjust the power output of the single-stage photovoltaic inverter according to the angular frequency obtained in 1.5: When the photovoltaic output power increases, causing the DC side voltage to rise, the inertial synchronization control equation outputs a positive frequency deviation, which makes the inverter accelerate the output of power; conversely, it makes the inverter decelerate the output of power, so as to simulate the inertial response characteristics of the synchronous generator.
4. The method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning according to claim 3, characterized in that: S2 includes: 2.1: Small-signal linearization is performed on the active-frequency control equation and reactive-voltage control equation of the virtual synchronous machine to obtain the small-signal model of the power outer loop: ; ; Where s is the Laplace operator; The small-signal disturbance component of the output phase angle of the virtual synchronizer; 、 These are the transfer function matrices of the frequency loop for current disturbance and voltage disturbance, respectively; 、 These are the transfer function matrices of the voltage loop for current disturbances and voltage disturbances, respectively. 、 These are the small-signal disturbance components of the grid-connected point current and grid-connected point voltage on the dq axis in the control system coordinate system, respectively. This represents the small-signal disturbance component of the inverter output voltage reference value along the dq axis in the control system coordinate system; the superscript c indicates the control system coordinate system, and the superscript indicates the reference value. 2.2: Based on the small-signal disturbance component of the inverter output voltage reference value obtained in 2.1, the virtual admittance element is linearized using small-signal method to obtain the small-signal expression for the inner current loop reference value: ; ; in, The small-signal disturbance component of the inner current loop reference value of the dq axis in the control system coordinate system; The transfer function matrix of the virtual admittance element; The transfer function matrix of the voltage feedback low-pass filter in the virtual admittance stage; , These are the virtual resistance and virtual inductance of the virtual admittance, respectively; 2.3: Based on the current inner loop reference value obtained in 2.2, combined with the small-signal equation of the current inner loop PI control: ; Based on the steady-state phase angle difference between the control system coordinate system and the electrical system coordinate system, a small-signal transformation relationship between the two coordinate systems is established: ; ; By combining the equations in 2.1 to 2.3, the dq impedance model of the grid-type photovoltaic inverter is obtained; in, This refers to the small-signal disturbance component of the modulated voltage reference value at the inverter output on the dq axis in the system coordinate system. This is the transfer function matrix of the current inner-loop PI controller; The rated angular frequency; For inverter-side filter inductance; This refers to the steady-state phase angle difference between the control system coordinate system and the electrical system coordinate system. , These are the small-signal disturbance components of the grid-connected point voltage and grid-connected point current along the dq axis in the electrical system coordinate system, respectively; the superscript 's' indicates the electrical system coordinate system. , , This is the coordinate transformation matrix.
5. The method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning according to claim 4, characterized in that: S2 further includes: S2.4: Convert the dq impedance model obtained in S2.3 into equivalent single-input single-output positive-sequence impedance and negative-sequence impedance: ; ; in, , These are the equivalent single-input single-output positive-sequence impedance and negative-sequence impedance of a grid-connected photovoltaic inverter, respectively. , These are the Laplace operators after positive and negative frequency shifts, respectively; , , , These are the positive-positive sequence component, positive-negative sequence coupling component, negative-positive sequence coupling component, and negative-negative sequence component of the improved sequence impedance of a grid-connected photovoltaic inverter, respectively. , , , These represent the positive-positive sequence component, positive-negative sequence coupled component, negative-positive sequence coupled component, and negative-negative sequence component of the improved power grid impedance, respectively; the superscripts p and n in the impedance symbol are the identifiers of the positive and negative sequence components, respectively, and do not represent independent parameters; S2.5: Based on the equivalent single-input single-output positive-sequence impedance and negative-sequence impedance obtained in S2.4, construct the sequence impedance matrix and define the system hysteresis matrix: ; in: The system back-ratio matrix; 、 These are the sequence impedance matrix of the grid-connected photovoltaic inverter and the grid sequence impedance matrix, respectively, which are composed of positive sequence impedance and negative sequence impedance; j is the imaginary unit.
6. The method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning according to claim 5, characterized in that: S3 includes: 3.1: Define the system loop impedance as the sum of the equivalent single-input single-output positive-sequence impedance and the equivalent impedance of the power grid: ; in, The system loop impedance; The equivalent impedance of the power grid; 3.2: Based on the system loop impedance defined in 3.1, the damping ratio of the system in the subsynchronous frequency band is determined by the real and imaginary parts of the system loop impedance: ; in, For the system at frequency Damping ratio at the point; For frequency; , These represent the operations of taking the real part and the operations of taking the imaginary part, respectively. The imaginary unit; Pi; 3.3: Based on the damping ratio determined in 3.2, when the imaginary part of the system loop impedance crosses zero at the resonant frequency, it satisfies... And satisfy At that time, the system was determined to exhibit negative damping, indicating a risk of subsynchronous oscillation instability. Conversely, the system is judged to exhibit positive damping; and with As critical boundary conditions for stability margin, the adjustable ranges of the virtual resistance, virtual inductance, and low-pass filter cutoff frequency are determined, and the quantitative mapping relationship between the virtual admittance parameters and the subsynchronous oscillation damping characteristics of the system is obtained. It is the resonant frequency corresponding to the imaginary part of the system loop impedance crossing zero.
7. The method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning according to claim 6, characterized in that: S4 includes: 4.1: Treat each subnet as an agent and construct the state space of agent i: ; in, The state of agent i; The deviation between the AC frequency of agent i and the rated frequency; DC bus voltage Deviation from the rated voltage of the DC bus; For integration time; 4.2: Based on the state space constructed in 4.1, a reward function is constructed with the goal of minimizing the system frequency deviation and voltage deviation, and a penalty term is introduced to constrain the system to operate within the stability boundary: ; in, The cumulative reward of agent i during the reward statistics period; , This is the reward coefficient; The rated voltage of the DC bus is T; T is the duration of reward statistics. The penalty for agent i is imposed when the combination of virtual admittance parameters causes the system to exceed the stability boundary; 4.3: Based on the reward function constructed in 4.2, the SAC algorithm is used to solve the weighted objective of maximizing the cumulative expected reward and the action distribution entropy, and the optimal virtual admittance parameters are output online. The objective expression of the SAC algorithm is: ; in, Here, E represents the objective function of the SAC algorithm; E is the expected value operation. , These represent the state and action at time t, respectively. Let be the immediate reward function after the action is performed at time t; Discount factor; H is the temperature coefficient; H is the entropy function of the strategy. For the strategy in the state The probability distribution of the next action selection.
8. The method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning according to claim 7, characterized in that: In section 4.3, the SAC algorithm is used to tune the optimal parameters of the virtual admittance online, including: The meta-learning mechanism is integrated into the SAC algorithm to obtain the improved SAC algorithm. The improved SAC algorithm treats each microgrid operation scenario as an independent task, updates network parameters under a specific operation scenario through an inner loop, and optimizes the initial network parameters by aggregating the losses of multiple operation scenarios through an outer loop, so that the network can quickly adapt to new operation scenarios. The objective function of the improved SAC algorithm is: ; in, To improve the task objective function of the SAC algorithm within the meta-learning framework; Let be the next state at time t; The value function for the next state; For value network parameters; The expected value of the value function for the next state; Let be the state transition probability.
9. The method for coordinated stability control of grid-type photovoltaic-storage microgrids based on impedance adaptive tuning according to claim 8, characterized in that: In step S4, the real-time identified oscillation frequency and damping ratio are obtained in the following way: Perform a frequency domain scan on the system loop impedance to find the resonant frequency corresponding to when the imaginary part of the system loop impedance crosses zero. The dominant oscillation frequency of the system is determined, and this dominant oscillation frequency is substituted into the expression for the damping ratio described in section 3.2 to calculate the current dominant damping ratio of the system. The dominant oscillation frequency and the dominant damping ratio are used as real-time inputs to the SAC algorithm to drive the online tuning of the virtual admittance parameters.
10. A grid-type photovoltaic-storage microgrid collaborative stability control system based on impedance adaptive tuning, characterized in that, The system includes: The architecture establishment module is used to establish a multi-subgrid photovoltaic-storage microgrid architecture and enable the grid-type photovoltaic inverters in each subgrid to adopt inertial synchronous control. Impedance modeling module is used to establish the impedance model of the grid-type photovoltaic inverter containing virtual admittance, and to analyze the oscillation stability of the subsynchronous frequency band based on the impedance model; The mapping establishment module is used to establish a quantitative mapping relationship between the virtual admittance parameters of the virtual admittance element and the subsynchronous oscillation damping characteristics of the system based on the impedance model. The parameter tuning module is used to tune the virtual admittance parameters online using the real-time identified oscillation frequency and damping ratio as inputs, based on the quantization mapping relationship and employing the soft actor-critic SAC algorithm, thereby achieving adaptive suppression of broadband oscillations.
Citation Information
Patent Citations
Subsynchronous oscillation suppression method and device of photovoltaic delivery system and controller
CN115483689A
Subsynchronous oscillation suppression method of photovoltaic grid-connected system based on data driving
CN116613749A
Virtual admittance-based power oscillation suppression method and device for network-forming converter
CN117713085A
Subsynchronous oscillation suppression method, device and equipment and storage medium
CN118889391A
