A method, system, device and medium for coordinated control of a receiving power grid

Abstract

The application provides a kind of receiving end power grid coordination control method, system, equipment and medium, receiving end power grid includes joint control system, joint control system includes static synchronous compensator and synchronous compensator;Method includes: obtaining the state variable of joint control system, according to the preset time constant, the state variable is divided into slow mechanical state and fast electromagnetic state;Input the preset singular perturbation standard model, determine the boundary layer variable of fast electromagnetic state deviating from slow manifold;And according to slow manifold and boundary layer variable, construct overall singular perturbation state function;The overall singular perturbation state function is input into the preset feature encoder, and the low-dimensional core feature representing system stability and fast-slow dynamic coupling relationship is extracted;Input multi-time scale reinforcement learning agent, generate joint control instruction for coordinating static synchronous compensator and synchronous compensator, realize the voltage control of receiving end power grid.The application can improve the voltage control precision of receiving end power grid.

Description

Technical Field

[0001] This invention application relates to the field of power system control, and more particularly to a coordinated control method, system, device and medium for a receiving-end power grid. Background Technology

[0002] With the large-scale integration of new energy sources into the receiving-end power grid via power electronic devices, the system exhibits significant low inertia and weak damping characteristics, leading to a substantial increase in the risk of transient voltage instability. To enhance the dynamic voltage support and disturbance rejection capabilities of the receiving-end power grid, static synchronous compensators (STATCOMs) and synchronous condensers are typically deployed in combination. STATCOMs utilize power electronic devices to provide millisecond-level rapid reactive power response, while synchronous condensers leverage the characteristics of rotating electrical machines to provide critical short-circuit capacity and inertia support. Therefore, research on voltage coordination control methods for such joint control systems composed of heterogeneous reactive power compensation sources, aiming to achieve efficient coordination of reactive power output between the two types of equipment, is of great significance for ensuring the safe and stable operation of modern power systems.

[0003] Existing voltage coordination control schemes typically employ traditional independent control or proportional-integral-derivative (PID) coordination control based on the overall equipment category. This involves macroscopically classifying STATCOMs as fast-moving support equipment and synchronous condensers as slow-moving support equipment. However, this broad classification based on overall equipment category ignores the multi-timescale dynamic coupling effects of heterogeneous reactive power sources at the state variable level, particularly neglecting the physical fact that synchronous condensers also exhibit rapid electromagnetic transients. When disturbances occur in the receiving-end power grid, traditional control strategies cannot mathematically and accurately describe the interleaved states of fast and slow states within and between equipment, leading to problems such as system over-regulation and / or severe oscillations, which ultimately affect the voltage control accuracy of the receiving-end power grid. Summary of the Invention

[0004] This invention application provides a coordinated control method, system, device, and medium for a receiving-end power grid to solve the technical problem of how to improve the voltage control accuracy of the receiving-end power grid.

[0005] To address the aforementioned technical problems, this invention provides a coordinated control method for a receiving-end power grid, wherein the receiving-end power grid includes a joint control system, and the joint control system includes a static synchronous compensator and a synchronous condenser. The coordination and control method includes: The state variables of the joint control system are obtained, and the state variables are divided into slow-changing electromechanical state and fast-changing electromagnetic state according to a preset time constant. The slow-changing electromechanical state and the fast-changing electromagnetic state are input into a preset singular perturbation standard model to determine the boundary layer variables of the fast-changing electromagnetic state that deviate from the slow manifold; and the overall singular perturbation state function is constructed based on the slow manifold and the boundary layer variables. The overall singular perturbation state function is input into a preset feature encoder to extract low-dimensional core features that characterize the system's stability and fast-slow dynamic coupling relationship; The low-dimensional core features are input into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronizing compensator and the synchronous condenser; based on the joint control commands, the static synchronizing compensator and the synchronous condenser are coordinated and controlled to achieve voltage control of the receiving-end power grid.

[0006] As a preferred embodiment, the step of acquiring the state variables of the joint control system and dividing the state variables into slow-changing electromechanical states and fast-changing electromagnetic states according to a preset time constant includes: Multiple synchronous condenser state variables and multiple static synchronous compensator state variables are acquired; wherein, the multiple synchronous condenser state variables include the synchronous condenser's rotor power angle, rotor angular velocity, q-axis transient electromotive force, d-axis transient electromotive force, terminal voltage, stator current d-axis component, and stator current q-axis component; the multiple static synchronous compensator state variables include the static synchronous compensator's DC side voltage, output current d-axis component, and output current q-axis component; According to the preset time constant, the state variables of each synchronous condenser and the state variables of each static synchronous compensator are respectively divided into the slow-changing electromechanical state and the fast-changing electromagnetic state. The preset time constant is used to characterize the dynamic time scale of each state variable, so that the electromechanical state and electromagnetic state in the same device can be classified into slow-changing electromechanical state and fast-changing electromagnetic state, respectively.

[0007] As a preferred embodiment, the step of inputting the slowly varying electromechanical state and the rapidly varying electromagnetic state into a preset singular perturbation standard model, determining the boundary layer variables of the rapidly varying electromagnetic state deviating from the slow manifold, and constructing an overall singular perturbation state function based on the slow manifold and the boundary layer variables, including: Based on the slow-changing electromechanical state, a nonlinear state equation for the slow-changing subsystem is constructed; based on the fast-changing electromagnetic state, a nonlinear state equation for the fast-changing subsystem is constructed. Obtain pre-configured singular perturbation parameters, and input the singular perturbation parameters, the nonlinear state equation of the slow-varying subsystem, and the nonlinear state equation of the fast-varying subsystem into the singular perturbation standard model; wherein, the singular perturbation parameters represent the ratio of the fast-varying dynamic characteristic time constant to the slow-varying dynamic characteristic time constant; The slow manifold is obtained by solving the singular perturbation standard model under preset constraints. Based on the deviation of the rapidly changing electromagnetic state relative to the slow manifold, the boundary layer variables are determined, and the slow manifold and the boundary layer variables are used together as dynamic descriptors to construct the global singular perturbation state function.

[0008] As a preferred embodiment, the step of inputting the overall singular perturbation state function into a preset feature encoder to extract low-dimensional core features characterizing the system's stability and fast-slow dynamic coupling relationship includes: The overall state uncertainty entropy is obtained based on the analysis of the overall singular perturbation state function. The average composite entropy index is calculated based on the overall state uncertainty entropy and the entropy of the boundary layer variables. Under the constraint of the average composite entropy index, the overall singular perturbation state function is input into a preset feature encoder to obtain the low-dimensional core features; wherein, the low-dimensional core features are used to characterize the slow dynamic evolution trend of the synchronous condenser electromechanical system, the fast electromagnetic response characteristics of the static synchronous compensator, the degree of deviation of the fast-changing subsystem from the slow manifold, and the dynamic evolution trend of the slow-changing electromechanical state.

[0009] As a preferred embodiment, the step of inputting the low-dimensional core features into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronization compensator and the synchronous adjustment phase includes: The reactive current reference adjustment of the static synchronous compensator and the excitation reference adjustment of the synchronous condenser are obtained, and the reactive current reference adjustment and the excitation reference adjustment are used as the action space of the multi-time-scale reinforcement learning agent. The low-dimensional core features are input into a multi-timescale reinforcement learning agent. Based on a preset reward function, the joint control action consisting of the reactive current reference adjustment and the excitation reference adjustment in the action space is evaluated to obtain a first evaluation result. The preset reward function includes a voltage deviation penalty term, a voltage change rate penalty term, a synchronous condenser speed oscillation penalty term, and a boundary layer energy term. When the first evaluation result meets the preset requirements, a joint control command is obtained based on the output of the multi-timescale reinforcement learning agent.

[0010] As a preferred embodiment, the multi-timescale reinforcement learning agent includes an Actor network and a Critic network; the training method for the multi-timescale reinforcement learning agent includes: Acquire training data; wherein the training data includes low-dimensional feature samples; The Actor network outputs candidate joint control actions based on the low-dimensional feature samples; The Critic network evaluates the state-action pairs based on the low-dimensional feature samples, candidate joint control actions, and the reward function to obtain a second evaluation result; Based on the second evaluation result, the parameters of the preset feature encoder, Actor network and Critic network are iteratively updated to obtain the trained multi-timescale reinforcement learning agent.

[0011] As a preferred embodiment, the coordinated control of the static synchronous compensator and the synchronous condenser based on joint control commands includes: The joint control commands are divided into reactive current reference commands for static synchronous compensators and excitation reference commands for synchronous condensers. The static synchronous compensator is controlled according to the first preset control cycle based on the reactive current reference command; According to the second preset control cycle, the excitation reference command is smoothly applied to the excitation system of the synchronous condenser through zero-order hold, low-pass filtering or rate of change limiting, thereby realizing the control of the synchronous condenser.

[0012] The first preset control period and the second preset control period are determined based on the controllable bandwidth of the corresponding execution channel.

[0013] Accordingly, this application also provides a coordinated control system for a receiving-end power grid, the receiving-end power grid including a joint control system, the joint control system including a static synchronous compensator and a synchronous condenser; The coordinated control system includes a state partitioning module, a function construction module, a feature extraction module, and a coordinated control module; wherein... The state division module is used to acquire the state variables of the joint control system and divide the state variables into slow-changing electromechanical state and fast-changing electromagnetic state according to a preset time constant. The function construction module is used to input the slow-changing electromechanical state and the fast-changing electromagnetic state into a preset singular perturbation standard model, determine the boundary layer variables of the fast-changing electromagnetic state that deviate from the slow manifold, and construct the overall singular perturbation state function based on the slow manifold and the boundary layer variables. The feature extraction module is used to input the overall singular perturbation state function into a preset feature encoder to extract low-dimensional core features that characterize the system's stability and fast-slow dynamic coupling relationship. The coordination control module is used to input the low-dimensional core features into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronizing compensator and the synchronous condenser; based on the joint control commands, it performs coordinated control on the static synchronizing compensator and the synchronous condenser to achieve voltage control of the receiving-end power grid.

[0014] As a preferred embodiment, the state division module acquires the state variables of the joint control system and, based on a preset time constant, divides the state variables into slow-changing electromechanical states and fast-changing electromagnetic states, including: The state division module acquires multiple synchronous condenser state variables and multiple static synchronous compensator state variables; wherein, the multiple synchronous condenser state variables include the synchronous condenser's rotor power angle, rotor angular velocity, q-axis transient electromotive force, d-axis transient electromotive force, terminal voltage, stator current d-axis component, and stator current q-axis component; the multiple static synchronous compensator state variables include the static synchronous compensator's DC side voltage, output current d-axis component, and output current q-axis component; According to the preset time constant, the state variables of each synchronous condenser and the state variables of each static synchronous compensator are respectively divided into the slow-changing electromechanical state and the fast-changing electromagnetic state. The preset time constant is used to characterize the dynamic time scale of each state variable, so that the electromechanical state and electromagnetic state in the same device can be classified into slow-changing electromechanical state and fast-changing electromagnetic state, respectively.

[0015] As a preferred embodiment, the function construction module inputs the slowly varying electromechanical state and the rapidly varying electromagnetic state into a preset singular perturbation standard model, determines the boundary layer variables of the rapidly varying electromagnetic state that deviate from the slowly varying manifold, and constructs an overall singular perturbation state function based on the slowly varying manifold and the boundary layer variables, including: The function construction module constructs the nonlinear state equation of the slow-changing subsystem based on the slow-changing electromechanical state, and constructs the nonlinear state equation of the fast-changing subsystem based on the fast-changing electromagnetic state. Obtain pre-configured singular perturbation parameters, and input the singular perturbation parameters, the nonlinear state equation of the slow-varying subsystem, and the nonlinear state equation of the fast-varying subsystem into the singular perturbation standard model; wherein, the singular perturbation parameters represent the ratio of the fast-varying dynamic characteristic time constant to the slow-varying dynamic characteristic time constant; The slow manifold is obtained by solving the singular perturbation standard model under preset constraints. Based on the deviation of the rapidly changing electromagnetic state relative to the slow manifold, the boundary layer variables are determined, and the slow manifold and the boundary layer variables are used together as dynamic descriptors to construct the global singular perturbation state function.

[0016] As a preferred embodiment, the feature extraction module inputs the global singular perturbation state function into a preset feature encoder to extract low-dimensional core features characterizing the system's stability and fast-slow dynamic coupling relationship, including: The feature extraction module obtains the overall state uncertainty entropy based on the overall singular perturbation state function analysis. The average composite entropy index is calculated based on the overall state uncertainty entropy and the entropy of the boundary layer variables. Under the constraint of the average composite entropy index, the overall singular perturbation state function is input into a preset feature encoder to obtain the low-dimensional core features; wherein, the low-dimensional core features are used to characterize the slow dynamic evolution trend of the synchronous condenser electromechanical system, the fast electromagnetic response characteristics of the static synchronous compensator, the degree of deviation of the fast-changing subsystem from the slow manifold, and the dynamic evolution trend of the slow-changing electromechanical state.

[0017] As a preferred embodiment, the coordination control module inputs the low-dimensional core features into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronization compensator and the synchronous condenser, including: The coordination control module obtains the reactive current reference adjustment of the static synchronous compensator and the excitation reference adjustment of the synchronous condenser, and uses the reactive current reference adjustment and the excitation reference adjustment as the action space of the multi-timescale reinforcement learning agent. The low-dimensional core features are input into a multi-timescale reinforcement learning agent. Based on a preset reward function, the joint control action consisting of the reactive current reference adjustment and the excitation reference adjustment in the action space is evaluated to obtain a first evaluation result. The preset reward function includes a voltage deviation penalty term, a voltage change rate penalty term, a synchronous condenser speed oscillation penalty term, and a boundary layer energy term. When the first evaluation result meets the preset requirements, a joint control command is obtained based on the output of the multi-timescale reinforcement learning agent.

[0018] As a preferred embodiment, the multi-timescale reinforcement learning agent includes an Actor network and a Critic network; the coordination control system further includes an agent training module, which is used for: Acquire training data; wherein the training data includes low-dimensional feature samples; The Actor network outputs candidate joint control actions based on the low-dimensional feature samples; The Critic network evaluates the state-action pairs based on the low-dimensional feature samples, candidate joint control actions, and the reward function to obtain a second evaluation result; Based on the second evaluation result, the parameters of the preset feature encoder, Actor network and Critic network are iteratively updated to obtain the trained multi-timescale reinforcement learning agent.

[0019] As a preferred embodiment, the coordination control module performs coordinated control of the static synchronous compensator and the synchronous condenser based on joint control commands, including: The coordination control module divides the joint control commands into reactive current reference commands for the static synchronous compensator and excitation reference commands for the synchronous condenser. The static synchronous compensator is controlled according to the first preset control cycle based on the reactive current reference command; According to the second preset control cycle, the excitation reference command is smoothly applied to the excitation system of the synchronous condenser through zero-order hold, low-pass filtering or rate of change limiting, thereby realizing the control of the synchronous condenser.

[0020] The first preset control period and the second preset control period are determined based on the controllable bandwidth of the corresponding execution channel.

[0021] Accordingly, this application also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the coordinated control method for the receiving-end power grid.

[0022] Accordingly, this application also provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the coordinated control method of the receiving-end power grid.

[0023] Compared with the prior art, this invention application has the following beneficial effects: This invention provides a coordinated control method, system, device, and medium for a receiving-end power grid. Compared to existing technologies that directly characterize the static synchronous compensator (SRC) as a fast-moving support device and the synchronous condenser (SCDC) as a slow-moving support device, this invention divides the slow-changing electromechanical state and the fast-changing electromagnetic state according to a preset time constant. It then uses the singular perturbation standard model to solve for the boundary layer variables of the fast-changing state deviating from the slow manifold, constructing a global singular perturbation state function. This not only allows for a refined and quantitative characterization of the rapid dynamic deviation of the SRC relative to the system's steady-state trajectory during transient processes but also does not neglect the internal characteristics of the SCDC. The physical fact of rapid electromagnetic transients also exists. Furthermore, by inputting the overall singular perturbation state function into a preset feature encoder, low-dimensional core features characterizing system stability and fast-slow dynamic coupling relationships are extracted. Compared with existing technologies, this does not ignore the multi-timescale dynamic coupling effect of heterogeneous reactive power sources at the state variable level. At the same time, the joint control commands generated by this invention for coordinating the static synchronous compensator and synchronous condenser can effectively avoid problems such as uneven reactive power distribution, over-regulation, and malignant system oscillations during emergency control of large disturbances, significantly improving the transient voltage stability and voltage control accuracy of the receiving-end power grid. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating an embodiment of the coordinated control method for the receiving-end power grid provided in this application.

[0025] Figure 2 This is a schematic diagram of an embodiment of the coordinated control system for the receiving-end power grid provided in this application. Detailed Implementation

[0026] 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.

[0027] Example 1 Please refer to Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the coordinated control method for the receiving-end power grid provided in this application.

[0028] In this embodiment, the receiving-end power grid includes a joint control system, which comprises a static synchronous compensator (SRC) and a synchronous condenser. The SRC provides rapid reactive power support; the synchronous condenser provides short-circuit capacity, inertia support, and excitation regulation support. This coordinated control method is applicable to receiving-end power grids with a high proportion of power electronic equipment connected, and is suitable for scenarios involving emergency voltage control and dynamic support at the receiving-end distribution side grid connection point.

[0029] Figure 1 The coordinated control method for the receiving-end power grid shown includes steps S101 to S104; each step is described in detail below: Step S101: Obtain the state variables of the joint control system, and divide the state variables into slow-changing electromechanical state and fast-changing electromagnetic state according to the preset time constant.

[0030] Specifically, the state variables of the joint control system may include the state variables of the synchronous condenser and the state variables of the static synchronous compensator.

[0031] In some preferred embodiments, step S101, which involves acquiring the state variables of the joint control system and dividing the state variables into slow-changing electromechanical states and fast-changing electromagnetic states according to a preset time constant, includes: Multiple synchronous condenser state variables and multiple static synchronous compensator state variables are acquired; wherein, the multiple synchronous condenser state variables include the synchronous condenser's rotor power angle, rotor angular velocity, q-axis transient electromotive force, d-axis transient electromotive force, terminal voltage, stator current d-axis component, and stator current q-axis component; the multiple static synchronous compensator state variables include the static synchronous compensator's DC side voltage, output current d-axis component, and output current q-axis component; According to the preset time constant, the state variables of each synchronous condenser and the state variables of each static synchronous compensator are respectively divided into the slow-changing electromechanical state and the fast-changing electromagnetic state. The preset time constant is used to characterize the dynamic time scale of each state variable, so that the electromechanical state and electromagnetic state in the same device can be classified into slow-changing electromechanical state and fast-changing electromagnetic state, respectively.

[0032] In this embodiment, the slowly varying electromechanical state can be represented by a slowly varying state vector. This slowly varying state vector characterizes the synchronous condenser rotor motion, excitation regulation, and terminal voltage-related electromechanical states, and can be expressed as: ; Where x is a slowly changing state vector. To synchronize the camera rotor's power angle, ω is the angular velocity of the rotor. The transient potential along the q-axis. The transient potential along the d-axis is... To synchronize the voltage at the camera's terminals.

[0033] Furthermore, the rapidly changing electromagnetic state can be represented by a rapidly changing state vector, which can be used to characterize the electromagnetic transients of STATCOM, the electromagnetic transients of the network, and the electromagnetic fast dynamics of the stator side of the synchronous condenser.

[0034] The rapidly changing state vector z described above can be represented as: ; In the formula, i d For the d-axis component of the stator current of the synchronous modulator, i q The current is the q-axis component of the stator current of the synchronous condenser. This current quantity is used to characterize the electromagnetic transients on the stator side of the synchronous condenser. Its time constant is significantly smaller than that of the rotor electromechanical transients, therefore it is classified into the fast-changing state vector; V dc This refers to the DC side voltage of the static synchronous compensator. The d-axis component of the output current of the static synchronous compensator. This refers to the q-axis component of the output current of the static synchronous compensator.

[0035] It should be noted that the fast and slow division in this embodiment is based on the time scale of the state variables, rather than simply defining the static synchronous compensator as a fast-changing device and the synchronous condenser as a slow-changing device.

[0036] Although synchronous condensers exhibit primarily slow electromechanical dynamics, their response time constants for electromagnetic quantities such as stator current are significantly smaller than those for electromechanical quantities such as rotor power angle and rotor angular velocity. Therefore, they can be classified as rapidly changing electromagnetic states. This classification avoids the problem of traditional control schemes neglecting the rapid electromagnetic transients within synchronous condensers.

[0037] Step S102: Input the slow-changing electromechanical state and the fast-changing electromagnetic state into a preset singular perturbation standard model, determine the boundary layer variables of the fast-changing electromagnetic state that deviate from the slow manifold, and construct the overall singular perturbation state function based on the slow manifold and the boundary layer variables.

[0038] Following the slow-changing electromechanical state and the fast-changing electromagnetic state obtained in step S101, this embodiment constructs the nonlinear state equations for the slow-changing subsystem and the fast-changing subsystem, respectively. The nonlinear state equation for the slow-changing subsystem can be composed of the synchronous condenser electromechanical swing equation, the excitation system, and the voltage / reactive power relationship; the nonlinear state equation for the fast-changing subsystem can be composed of the STATCOM current inner loop, the DC voltage loop, the grid-connected reactor, and the electromagnetic transient equations on the stator side of the synchronous condenser.

[0039] In some preferred embodiments, step S102 involves inputting the slowly varying electromechanical state and the rapidly varying electromagnetic state into a preset singular perturbation standard model to determine the boundary layer variables in which the rapidly varying electromagnetic state deviates from the slow manifold; and constructing an overall singular perturbation state function based on the slow manifold and the boundary layer variables, including: Based on the slow-changing electromechanical state, a nonlinear state equation for the slow-changing subsystem is constructed; based on the fast-changing electromagnetic state, a nonlinear state equation for the fast-changing subsystem is constructed. Obtain pre-configured singular perturbation parameters, and input the singular perturbation parameters, the nonlinear state equation of the slow-varying subsystem, and the nonlinear state equation of the fast-varying subsystem into the singular perturbation standard model; wherein, the singular perturbation parameters represent the ratio of the fast-varying dynamic characteristic time constant to the slow-varying dynamic characteristic time constant; The slow manifold is obtained by solving the singular perturbation standard model under preset constraints. Based on the deviation of the rapidly changing electromagnetic state relative to the slow manifold, the boundary layer variables are determined, and the slow manifold and the boundary layer variables are used together as dynamic descriptors to construct the global singular perturbation state function.

[0040] Specifically, in this embodiment, the joint control input vector may include the synchronous condenser excitation voltage and STATCOM reactive current command, and external disturbances may include disturbances such as faults, load changes, or fluctuations in new energy output.

[0041] The singular perturbation parameters, the nonlinear state equations of the slowly varying subsystem, and the nonlinear state equations of the fast varying subsystem are input into a preset singular perturbation standard model, which can be expressed as: ; In the formula, The nonlinear state equation is composed of the electromechanical swing equation of the synchronous condenser, the excitation system, and the voltage / reactive power relationship. The nonlinear fast-changing state equations consist of the STATCOM current inner loop / DC voltage loop, the electromagnetic transient equations of the grid-connected reactor and the stator side of the synchronous condenser. The nonlinear state equations of the slowly varying subsystem are as follows: The nonlinear state equations of the fast variable subsystem are as follows: For singular perturbation parameters, The joint control input vector includes the excitation voltage of the synchronous modulator. Reactive current command of static synchronous compensator , External disturbances (such as faults or sudden load changes).

[0042] Furthermore, the singular perturbation standard model is solved under preset constraints to obtain a slow manifold. The slow manifold can be understood as the quasi-steady-state trajectory formed by a fast-changing subsystem when it converges sufficiently fast relative to a slow-changing subsystem. The preset constraints can include slow manifold solving constraints, system stability constraints, variable value constraints, or boundary conditions.

[0043] It should be noted that in singular perturbation analysis, the singular perturbation parameter is brought close to zero for theoretical decomposition and approximate analysis, and does not mean that the fast dynamic time constant in the actual physical system is actually zero.

[0044] Boundary layer variables are determined based on the deviation of the rapidly changing electromagnetic state from the slow manifold. These boundary layer variables reflect the degree to which the rapidly changing state deviates from the slow manifold and characterize the rapid dynamic deviation of the STATCOM from the system's steady-state trajectory during transient processes. The slow manifold and boundary layer variables are used together as dynamic descriptors to construct a global singular perturbation state function. This global singular perturbation state function is not a simple concatenation of state vectors but a dynamic description including the deviation between the fast and slow manifolds, which can be expressed as: ; ; In the formula, Let be the global singular perturbation state function, and t denote that the function is related to time. For order At that time, the nonlinear fast-changing state equation satisfied by the fast-changing subsystem The solution, in a real physical system, It is usually a very small positive number greater than 0, rather than actually being 0. In the analysis, let... The physical meaning of this is not that the time constant of the fast-changing subsystem is actually zero, but rather that, relative to the slow-changing subsystem, the convergence process of the fast-changing subsystem is fast enough that it can be approximated as reaching a quasi-steady state instantaneously, thus achieving the theoretical decomposition of fast and slow dynamics.

[0045] Therefore, the global singular perturbation state function output in step S102 can be passed to step S103 for subsequent low-dimensional core feature extraction.

[0046] Step S103: Input the overall singular perturbation state function into a preset feature encoder to extract low-dimensional core features that characterize the system stability and fast-slow dynamic coupling relationship.

[0047] Following the global singular perturbation state function obtained in step S102, since this function contains high-dimensional coupled information such as slowly varying electromechanical states, rapidly varying electromagnetic states, and boundary layer biases, directly using it for control decisions could easily increase the state dimension of the reinforcement learning agent and reduce decision-making efficiency. Therefore, this embodiment sets a preset feature encoder to compress and represent the global singular perturbation state function to obtain low-dimensional core features.

[0048] In some preferred embodiments, step S103, which involves inputting the global singular perturbation state function into a preset feature encoder to extract low-dimensional core features characterizing the system's stability and fast-slow dynamic coupling relationship, includes: The overall state uncertainty entropy is obtained based on the analysis of the overall singular perturbation state function. The average composite entropy index is calculated based on the overall state uncertainty entropy and the entropy of the boundary layer variables. Under the constraint of the average composite entropy index, the overall singular perturbation state function is input into a preset feature encoder to obtain the low-dimensional core features; wherein, the low-dimensional core features are used to characterize the slow dynamic evolution trend of the synchronous condenser electromechanical system, the fast electromagnetic response characteristics of the static synchronous compensator, the degree of deviation of the fast-changing subsystem from the slow manifold, and the dynamic evolution trend of the slow-changing electromechanical state.

[0049] In this embodiment, the average composite entropy index is used to quantify redundant, fluctuating, and uncertain information in high-dimensional perturbation states and boundary layer variables, and to constrain the preset feature encoder to extract the low-dimensional core features most representative of system stability. The average composite entropy index can be expressed as: ; In the formula, H avg The above average composite entropy, Let be the entropy of the uncertainty of the overall state described above. Boundary layer variables Entropy, specifically referring to the uncertainty of rapidly changing dynamics deviating from a slow manifold, is intended to punish excessive rapid oscillations. For policy entropy, , and These are the weighting coefficients.

[0050] For example, the average composite entropy index may include an overall state uncertainty term, a boundary layer variable uncertainty term, and a policy entropy term. The overall state uncertainty term characterizes the uncertainty of the overall singular perturbation state; the boundary layer variable uncertainty term characterizes the degree of uncertainty of rapid dynamic deviations from the slow manifold, thus suppressing excessively rapid oscillations; the policy entropy term can be used during the reinforcement learning training phase to maintain moderate exploration and reduce the risk of the policy getting trapped in local optima. The entropy of the boundary layer variables can be expressed as: ; In the formula, This represents the probability density distribution of the boundary layer variables.

[0051] Furthermore, a parameterized preset feature encoder is set up, and the high-dimensional global singular perturbation state function is input into the preset feature encoder to output a low-dimensional core feature vector. The training objective is to find the optimal encoder parameters such that the encoded low-dimensional core features minimize the average composite entropy index. It can be represented as: ; In the formula, The parameter is The feature encoder.

[0052] The purpose of this feature encoder is to take high-dimensional input. The output is a low-dimensional feature vector. The training objective is to find the optimal parameters. This allows the encoded features to minimize the aforementioned average composite entropy.

[0053] In this way, the preset feature encoder can filter out the slow-changing trend features and fast-changing impact features that are most critical to voltage recovery from complex singular perturbation states, generating a denoised and compressed state representation. The low-dimensional core features are used to characterize the slow dynamic evolution trend of the synchronous condenser electromechanical system, the fast electromagnetic response features of the static synchronous compensator, and the degree to which the fast and slow subsystems deviate from the slow manifold, and serve as the input to the multi-timescale reinforcement learning agent in step S104.

[0054] Step S104: Input the low-dimensional core features into the multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronizing compensator and the synchronous condenser; based on the joint control commands, coordinate the control of the static synchronizing compensator and the synchronous condenser to achieve voltage control of the receiving-end power grid.

[0055] In this step, the low-dimensional core features are used as the state input of the multi-timescale reinforcement learning agent. The low-dimensional core features are not a single physical quantity, but a state representation formed by fusing and compressing key stability information such as the slow dynamic evolution trend of the synchronous condenser electromechanical system, the fast electromagnetic response characteristics of STATCOM, and the degree of deviation of the fast and slow subsystems from the slow manifold. It can comprehensively reflect the system's transient voltage support requirements, dynamic recovery trend, and the degree of coordination between fast and slow subsystems.

[0056] In some preferred embodiments, step S104, which involves inputting the low-dimensional core features into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the stationary synchronization compensator and the synchronous adjustment phase, includes: The reactive current reference adjustment of the static synchronous compensator and the excitation reference adjustment of the synchronous condenser are obtained, and the reactive current reference adjustment and the excitation reference adjustment are used as the action space of the multi-time-scale reinforcement learning agent. The low-dimensional core features are input into a multi-timescale reinforcement learning agent. Based on a preset reward function, the joint control action consisting of the reactive current reference adjustment and the excitation reference adjustment in the action space is evaluated to obtain a first evaluation result. The preset reward function includes a voltage deviation penalty term, a voltage change rate penalty term, a synchronous condenser speed oscillation penalty term, and a boundary layer energy term. When the first evaluation result meets the preset requirements, a joint control command is obtained based on the output of the multi-timescale reinforcement learning agent.

[0057] For example, the reactive current reference adjustment of the static synchronous compensator corresponds to the fast channel control quantity, and the excitation reference adjustment of the synchronous condenser corresponds to the slow channel control quantity. The control cycle of the two can be determined according to the controllable bandwidth of the corresponding execution channel.

[0058] The aforementioned preset requirements can be that the reward value is greater than a preset threshold, the control action meets safety constraints, or the first evaluation result meets the convergence condition.

[0059] Furthermore, a preset reward function is designed to evaluate the joint control actions in the action space and obtain a first evaluation result. The preset reward function may include a voltage deviation penalty term, a voltage change rate penalty term, a synchronous condenser speed oscillation penalty term, and a boundary layer energy term. The voltage deviation penalty term is used to constrain the grid connection point bus voltage to approach the voltage reference value; the voltage change rate penalty term is used to suppress excessively rapid voltage changes at the grid connection point, prompting the STATCOM to quickly suppress voltage abrupt changes and smooth the voltage recovery process; the synchronous condenser speed oscillation penalty term is used to provide damping support; and the boundary layer energy term is used to penalize the degree of incoordination between fast and slow systems, i.e., the deviation of the fast-changing state relative to the slow manifold. The preset reward function R... tIt can be represented as: ; In the formula, This is the voltage reference value. For real-time measurement of voltage at the grid connection point bus, To penalize voltage deviation, This is a penalty term for the square of the voltage change rate, used to penalize situations where the voltage at the grid connection point changes too rapidly, encouraging the static synchronous compensator to quickly suppress voltage spikes and smooth the voltage recovery process. To punish the oscillation of the synchronous synchrotron speed, damping support is provided, and the boundary layer energy term is applied. The degree of incoordination between fast and slow systems (i.e., the deviation between the action of the static synchronizing compensator and its steady-state target). , , and These are the weighting coefficients.

[0060] in , representing the deviation of the rapidly changing state relative to the slow manifold, can be calculated online from the original singular perturbation state, system measurements, or state estimation results.

[0061] When the first evaluation result meets the preset requirements, a joint control command is obtained based on the output of the multi-timescale reinforcement learning agent. The joint control command may include the reactive current reference command of the static synchronizing compensator and the excitation reference command of the synchronous condenser. The mapping relationship of the jointly controlled command output by the trained agent based on low-dimensional core features can be expressed as: ; In the formula, This is the reactive current reference command for the static synchronous compensator. For synchronous condenser excitation reference command, It is a trained policy network or control policy mapping, whose function is to generate corresponding joint control actions (joint control instructions) based on the core feature states of the input.

[0062] Preferably, the multi-timescale reinforcement learning agent includes an Actor network and a Critic network; the training method for the multi-timescale reinforcement learning agent includes: Acquire training data; wherein the training data includes low-dimensional feature samples; The Actor network outputs candidate joint control actions based on the low-dimensional feature samples; The Critic network evaluates the state-action pairs based on the low-dimensional feature samples, candidate joint control actions, and the reward function to obtain a second evaluation result; Based on the second evaluation result, the parameters of the preset feature encoder, Actor network and Critic network are iteratively updated to obtain the trained multi-timescale reinforcement learning agent.

[0063] Specifically, the multi-timescale reinforcement learning agent can be trained using a deep deterministic policy gradient algorithm applicable to continuous action spaces.

[0064] During training, the parameters of the feature encoder, Actor network, and Critic network are optimized collaboratively through iterative methods. Furthermore, exploration constraints or entropy regularization mechanisms related to the policy distribution can be introduced to maintain a degree of randomness while optimizing the cumulative reward, thereby reducing the risk of getting trapped in local optima. Thus, the trained joint control policy can improve voltage support performance while avoiding instability in control actions due to premature policy convergence.

[0065] Preferably, the coordinated control of the static synchronizing compensator and the synchronous condenser based on joint control commands includes: The joint control commands are divided into reactive current reference commands for static synchronous compensators and excitation reference commands for synchronous condensers. The static synchronous compensator is controlled according to the first preset control cycle based on the reactive current reference command; According to the second preset control cycle, the excitation reference command is smoothly applied to the excitation system of the synchronous condenser through zero-order hold, low-pass filtering or rate of change limiting, so as to realize the control of the synchronous condenser. The first preset control period and the second preset control period are determined based on the controllable bandwidth of the corresponding execution channel.

[0066] In this embodiment, the reactive current reference command of the static synchronous compensator can be updated in each fast control cycle (corresponding to the first preset control cycle) and tracked by the STATCOM local current inner loop to achieve millisecond-level fast reactive power support.

[0067] The excitation reference command of the synchronous condenser can be updated in each slow control cycle (corresponding to the second preset control cycle), and applied smoothly to the synchronous condenser excitation system between adjacent slow control cycles through zero-order hold, low-pass filtering or rate of change limiting, so as to avoid oscillation caused by high-frequency excitation commands.

[0068] It should be noted that the fast / slow division in step S101 refers to the time scale of the state variable, while the fast / slow control cycle in step S104 refers to the controllable bandwidth of the actuator. The two correspond to each other but are not the same.

[0069] The stator current of a synchronous condenser can be in a rapidly changing electromagnetic state, but the external control channel of the synchronous condenser is mainly the excitation system, which is a slow-bandwidth object overall. Therefore, this embodiment adopts a unified strategy output and a multi-rate execution mechanism for different channels, so that the STATCOM reference command is updated and tracked quickly according to the fast control cycle, and the synchronous condenser excitation reference is updated and applied smoothly according to the slow control cycle, thereby realizing the coordinated control of the two types of reactive power support equipment on the time axis.

[0070] Through the above steps, this embodiment can use singular perturbation theory to distinguish between fast and slow dynamics in the mathematical model and establish the relationship between them. By using average composite entropy constraints to extract low-dimensional core features that can characterize the transient stability of the system, and then using multi-timescale reinforcement learning agents to generate joint control commands, the static synchronous compensator can play a role in rapid reactive power support in the early stage of disturbance, and the synchronous condenser can play a role in inertia and damping support in the subsequent recovery stage. This reduces the risk of uneven reactive power distribution, over-regulation and system oscillation, and improves the transient voltage stability and voltage control accuracy of the receiving-end power grid.

[0071] Accordingly, such as Figure 2 As shown, this application of the present invention also provides a coordinated control system 200 for a receiving-end power grid, the receiving-end power grid including a joint control system, the joint control system including a static synchronous compensator and a synchronous condenser; The coordinated control system 200 includes a state partitioning module 201, a function construction module 202, a feature extraction module 203, and a coordinated control module 204; wherein... The state division module 201 is used to acquire the state variables of the joint control system and divide the state variables into slow-changing electromechanical state and fast-changing electromagnetic state according to a preset time constant. The function construction module 202 is used to input the slow-changing electromechanical state and the fast-changing electromagnetic state into a preset singular perturbation standard model, determine the boundary layer variables of the fast-changing electromagnetic state that deviate from the slow manifold, and construct the overall singular perturbation state function based on the slow manifold and the boundary layer variables. The feature extraction module 203 is used to input the overall singular perturbation state function into a preset feature encoder to extract low-dimensional core features that characterize the system stability and fast-slow dynamic coupling relationship. The coordination control module 204 is used to input the low-dimensional core features into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronizing compensator and the synchronous condenser; based on the joint control commands, it performs coordinated control on the static synchronizing compensator and the synchronous condenser to achieve voltage control of the receiving-end power grid.

[0072] As a preferred embodiment, the state division module 201 acquires the state variables of the joint control system and, based on a preset time constant, divides the state variables into slow-changing electromechanical states and fast-changing electromagnetic states, including: The state division module 201 acquires multiple synchronous condenser state variables and multiple static synchronous compensator state variables; wherein, the multiple synchronous condenser state variables include the synchronous condenser's rotor power angle, rotor angular velocity, q-axis transient electromotive force, d-axis transient electromotive force, terminal voltage, stator current d-axis component, and stator current q-axis component; the multiple static synchronous compensator state variables include the static synchronous compensator's DC side voltage, output current d-axis component, and output current q-axis component; According to the preset time constant, the state variables of each synchronous condenser and the state variables of each static synchronous compensator are respectively divided into the slow-changing electromechanical state and the fast-changing electromagnetic state. The preset time constant is used to characterize the dynamic time scale of each state variable, so that the electromechanical state and electromagnetic state in the same device can be classified into slow-changing electromechanical state and fast-changing electromagnetic state, respectively.

[0073] As a preferred embodiment, the function construction module 202 inputs the slowly varying electromechanical state and the rapidly varying electromagnetic state into a preset singular perturbation standard model, determines the boundary layer variables where the rapidly varying electromagnetic state deviates from the slowly varying manifold, and constructs an overall singular perturbation state function based on the slowly varying manifold and the boundary layer variables, including: The function construction module 202 constructs the nonlinear state equation of the slow-changing subsystem based on the slow-changing electromechanical state, and constructs the nonlinear state equation of the fast-changing subsystem based on the fast-changing electromagnetic state. Obtain pre-configured singular perturbation parameters, and input the singular perturbation parameters, the nonlinear state equation of the slow-varying subsystem, and the nonlinear state equation of the fast-varying subsystem into the singular perturbation standard model; wherein, the singular perturbation parameters represent the ratio of the fast-varying dynamic characteristic time constant to the slow-varying dynamic characteristic time constant; The slow manifold is obtained by solving the singular perturbation standard model under preset constraints. Based on the deviation of the rapidly changing electromagnetic state relative to the slow manifold, the boundary layer variables are determined, and the slow manifold and the boundary layer variables are used together as dynamic descriptors to construct the global singular perturbation state function.

[0074] As a preferred embodiment, the feature extraction module 203 inputs the overall singular perturbation state function into a preset feature encoder to extract low-dimensional core features characterizing the system's stability and fast-slow dynamic coupling relationship, including: The feature extraction module 203 obtains the overall state uncertainty entropy based on the overall singular perturbation state function analysis. The average composite entropy index is calculated based on the overall state uncertainty entropy and the entropy of the boundary layer variables. Under the constraint of the average composite entropy index, the overall singular perturbation state function is input into a preset feature encoder to obtain the low-dimensional core features; wherein, the low-dimensional core features are used to characterize the slow dynamic evolution trend of the synchronous condenser electromechanical system, the fast electromagnetic response characteristics of the static synchronous compensator, the degree of deviation of the fast-changing subsystem from the slow manifold, and the dynamic evolution trend of the slow-changing electromechanical state.

[0075] As a preferred embodiment, the coordination control module 204 inputs the low-dimensional core features into the multi-timescale reinforcement learning agent to generate joint control commands for coordinating the stationary synchronization compensator and the synchronous adjustment phase, including: The coordination control module 204 acquires the reactive current reference adjustment amount of the static synchronous compensator and the excitation reference adjustment amount of the synchronous condenser, and uses the reactive current reference adjustment amount and the excitation reference adjustment amount as the action space of the multi-timescale reinforcement learning agent. The low-dimensional core features are input into a multi-timescale reinforcement learning agent. Based on a preset reward function, the joint control action consisting of the reactive current reference adjustment and the excitation reference adjustment in the action space is evaluated to obtain a first evaluation result. The preset reward function includes a voltage deviation penalty term, a voltage change rate penalty term, a synchronous condenser speed oscillation penalty term, and a boundary layer energy term. When the first evaluation result meets the preset requirements, a joint control command is obtained based on the output of the multi-timescale reinforcement learning agent.

[0076] As a preferred embodiment, the multi-timescale reinforcement learning agent includes an Actor network and a Critic network; the coordination control system 200 further includes an agent training module, which is used for: Acquire training data; wherein the training data includes low-dimensional feature samples; The Actor network outputs candidate joint control actions based on the low-dimensional feature samples; The Critic network evaluates the state-action pairs based on the low-dimensional feature samples, candidate joint control actions, and the reward function to obtain a second evaluation result; Based on the second evaluation result, the parameters of the preset feature encoder, Actor network and Critic network are iteratively updated to obtain the trained multi-timescale reinforcement learning agent.

[0077] As a preferred embodiment, the coordination control module 204 performs coordinated control of the static synchronous compensator and the synchronous condenser based on joint control commands, including: The coordination control module 204 divides the joint control commands into reactive current reference commands for the static synchronous compensator and excitation reference commands for the synchronous condenser. The static synchronous compensator is controlled according to the first preset control cycle based on the reactive current reference command; According to the second preset control cycle, the excitation reference command is smoothly applied to the excitation system of the synchronous condenser through zero-order hold, low-pass filtering or rate of change limiting, thereby realizing the control of the synchronous condenser.

[0078] The first preset control period and the second preset control period are determined based on the controllable bandwidth of the corresponding execution channel.

[0079] Accordingly, this application also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the coordinated control method for the receiving-end power grid.

[0080] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal, connecting various parts of the terminal via various interfaces and lines.

[0081] The memory can be used to store the computer program. The processor implements various functions of the terminal by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0082] Accordingly, this application also provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the coordinated control method of the receiving-end power grid.

[0083] Wherein, if the modules of the device / terminal equipment / system integration are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0084] Compared with the prior art, this invention application has the following beneficial effects: This invention provides a coordinated control method, system, device, and medium for a receiving-end power grid. Compared to existing technologies that directly characterize the static synchronous compensator (SRC) as a fast-moving support device and the synchronous condenser (SCDC) as a slow-moving support device, this invention divides the slow-changing electromechanical state and the fast-changing electromagnetic state according to a preset time constant. It then uses the singular perturbation standard model to solve for the boundary layer variables of the fast-changing state deviating from the slow manifold, constructing a global singular perturbation state function. This not only allows for a refined and quantitative characterization of the rapid dynamic deviation of the SRC relative to the system's steady-state trajectory during transient processes but also does not neglect the internal characteristics of the SCDC. The physical fact of rapid electromagnetic transients also exists. Furthermore, by inputting the overall singular perturbation state function into a preset feature encoder, low-dimensional core features characterizing system stability and fast-slow dynamic coupling relationships are extracted. Compared with existing technologies, this does not ignore the multi-timescale dynamic coupling effect of heterogeneous reactive power sources at the state variable level. At the same time, the joint control commands generated by this invention for coordinating the static synchronous compensator and synchronous condenser can effectively avoid problems such as uneven reactive power distribution, over-regulation, and malignant system oscillations during emergency control of large disturbances, significantly improving the transient voltage stability and voltage control accuracy of the receiving-end power grid.

[0085] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A coordinated control method for a receiving-end power grid, characterized in that, The receiving-end power grid includes a joint control system, which includes a static synchronous compensator and a synchronous condenser. The coordination and control method includes: The state variables of the joint control system are obtained, and the state variables are divided into slow-changing electromechanical state and fast-changing electromagnetic state according to a preset time constant. The slow-changing electromechanical state and the fast-changing electromagnetic state are input into a preset singular perturbation standard model to determine the boundary layer variables of the fast-changing electromagnetic state that deviate from the slow manifold; and the overall singular perturbation state function is constructed based on the slow manifold and the boundary layer variables. The overall singular perturbation state function is input into a preset feature encoder to extract low-dimensional core features that characterize the system's stability and fast-slow dynamic coupling relationship; The low-dimensional core features are input into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronizing compensator and the synchronous condenser; based on the joint control commands, the static synchronizing compensator and the synchronous condenser are coordinated and controlled to achieve voltage control of the receiving-end power grid.

2. The coordinated control method for a receiving-end power grid as described in claim 1, characterized in that, The process of acquiring the state variables of the joint control system and dividing the state variables into slow-changing electromechanical states and fast-changing electromagnetic states according to a preset time constant includes: Multiple synchronous condenser state variables and multiple static synchronous compensator state variables are acquired; wherein, the multiple synchronous condenser state variables include the synchronous condenser's rotor power angle, rotor angular velocity, q-axis transient electromotive force, d-axis transient electromotive force, terminal voltage, stator current d-axis component, and stator current q-axis component; the multiple static synchronous compensator state variables include the static synchronous compensator's DC side voltage, output current d-axis component, and output current q-axis component; According to the preset time constant, the state variables of each synchronous condenser and the state variables of each static synchronous compensator are respectively divided into the slow-changing electromechanical state and the fast-changing electromagnetic state. The preset time constant is used to characterize the dynamic time scale of each state variable, so that the electromechanical state and electromagnetic state in the same device are respectively classified into slow-changing electromechanical state and fast-changing electromagnetic state.

3. The coordinated control method for a receiving-end power grid as described in claim 2, characterized in that, The slow-changing electromechanical state and the fast-changing electromagnetic state are input into a preset singular perturbation standard model to determine the boundary layer variables in which the fast-changing electromagnetic state deviates from the slow manifold. Based on the slow manifold and boundary layer variables, a global singular perturbation state function is constructed, including: Based on the slow-changing electromechanical state, a nonlinear state equation for the slow-changing subsystem is constructed; based on the fast-changing electromagnetic state, a nonlinear state equation for the fast-changing subsystem is constructed. Obtain pre-configured singular perturbation parameters, and input the singular perturbation parameters, the nonlinear state equation of the slow-varying subsystem, and the nonlinear state equation of the fast-varying subsystem into the singular perturbation standard model; wherein, the singular perturbation parameters represent the ratio of the fast-varying dynamic characteristic time constant to the slow-varying dynamic characteristic time constant; The slow manifold is obtained by solving the singular perturbation standard model under preset constraints. Based on the deviation of the rapidly changing electromagnetic state relative to the slow manifold, the boundary layer variables are determined, and the slow manifold and the boundary layer variables are used together as dynamic descriptors to construct the global singular perturbation state function.

4. The coordinated control method for a receiving-end power grid as described in claim 3, characterized in that, The step of inputting the global singular perturbation state function into a preset feature encoder to extract low-dimensional core features characterizing the system's stability and fast-slow dynamic coupling relationship includes: The overall state uncertainty entropy is obtained based on the analysis of the overall singular perturbation state function. The average composite entropy index is calculated based on the overall state uncertainty entropy and the entropy of the boundary layer variables. Under the constraint of the average composite entropy index, the overall singular perturbation state function is input into a preset feature encoder to obtain the low-dimensional core features; wherein, the low-dimensional core features are used to characterize the electromechanical slow dynamic evolution trend of the synchronous condenser, the fast electromagnetic response characteristics of the static synchronous compensator, the degree of deviation of the fast-changing subsystem from the slow manifold, and the dynamic evolution trend of the slow-changing electromechanical state.

5. The coordinated control method for a receiving-end power grid as described in claim 4, characterized in that, The step of inputting the low-dimensional core features into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronization compensator and the synchronous adjustment phase includes: The reactive current reference adjustment of the static synchronous compensator and the excitation reference adjustment of the synchronous condenser are obtained, and the reactive current reference adjustment and the excitation reference adjustment are used as the action space of the multi-time-scale reinforcement learning agent. The low-dimensional core features are input into a multi-timescale reinforcement learning agent. Based on a preset reward function, the joint control action consisting of the reactive current reference adjustment and the excitation reference adjustment in the action space is evaluated to obtain a first evaluation result. The preset reward function includes a voltage deviation penalty term, a voltage change rate penalty term, a synchronous condenser speed oscillation penalty term, and a boundary layer energy term. When the first evaluation result meets the preset requirements, a joint control command is obtained based on the output of the multi-timescale reinforcement learning agent.

6. The coordinated control method for a receiving-end power grid as described in claim 5, characterized in that, The multi-timescale reinforcement learning agent includes an Actor network and a Critic network; The training method for the multi-timescale reinforcement learning agent includes: Acquire training data; wherein the training data includes low-dimensional feature samples; The Actor network outputs candidate joint control actions based on the low-dimensional feature samples; The Critic network evaluates the state-action pairs based on the low-dimensional feature samples, candidate joint control actions, and the reward function to obtain a second evaluation result; Based on the second evaluation result, the parameters of the preset feature encoder, Actor network and Critic network are iteratively updated to obtain the trained multi-timescale reinforcement learning agent.

7. The coordinated control method for a receiving-end power grid as described in claim 5, characterized in that, The coordinated control of the static synchronizing compensator and the synchronous condenser based on joint control commands includes: The joint control commands are divided into reactive current reference commands for static synchronous compensators and excitation reference commands for synchronous condensers. The static synchronous compensator is controlled according to the first preset control cycle based on the reactive current reference command; According to the second preset control cycle, the excitation reference command is smoothly applied to the excitation system of the synchronous condenser through zero-order hold, low-pass filtering or rate of change limiting, so as to realize the control of the synchronous condenser. The first preset control period and the second preset control period are determined based on the controllable bandwidth of the corresponding execution channel.

8. A coordinated control system for a receiving-end power grid, characterized in that, The receiving-end power grid includes a joint control system, which includes a static synchronous compensator and a synchronous condenser. The coordinated control system includes a state partitioning module, a function construction module, a feature extraction module, and a coordinated control module; wherein... The state division module is used to acquire the state variables of the joint control system and divide the state variables into slow-changing electromechanical state and fast-changing electromagnetic state according to a preset time constant. The function construction module is used to input the slow-changing electromechanical state and the fast-changing electromagnetic state into a preset singular perturbation standard model, determine the boundary layer variables of the fast-changing electromagnetic state that deviate from the slow manifold, and construct the overall singular perturbation state function based on the slow manifold and the boundary layer variables. The feature extraction module is used to input the overall singular perturbation state function into a preset feature encoder to extract low-dimensional core features that characterize the system's stability and fast-slow dynamic coupling relationship. The coordination control module is used to input the low-dimensional core features into a multi-timescale reinforcement learning agent to generate joint control commands for coordinating the static synchronizing compensator and the synchronous condenser; based on the joint control commands, it performs coordinated control on the static synchronizing compensator and the synchronous condenser to achieve voltage control of the receiving-end power grid.

9. A terminal device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the coordinated control method for the receiving-end power grid as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the coordinated control method for the receiving-end power grid as described in any one of claims 1 to 7.

Images