Reactive compensation control method and device for weak power grid

Abstract

The invention relates to the technical field of reactive compensation, in particular to a reactive compensation control method and device for a weak power grid, and the method comprises the steps: carrying out the transient voltage monitoring of a plurality of monitoring nodes in a target weak power grid through transient voltage monitoring equipment, so as to obtain the transient voltage of the plurality of monitoring nodes, calculating a transient voltage stability index of the monitoring node based on the transient voltage of the monitoring node; and comparing the transient voltage stability index of the monitoring node with a preset stability threshold, and if the transient voltage stability index of the monitoring node is lower than the preset stability threshold, constructing a first coordination control model of the monitoring node. By monitoring the transient voltage stability index of the node and calculating and adjusting the compensation scheme according to the real-time data, the voltage stability of each node in the weak power grid can be effectively maintained, so that potential damage to equipment and users caused by voltage fluctuation can be avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of reactive power compensation, and in particular to a reactive power compensation control method and device for a weak power grid. Background Art

[0002] A weak power grid refers to a part or specific area in the power system with poor voltage stability and large voltage fluctuations. These areas may have poor voltage stability due to load changes, imperfect power grid topology or insufficient power supply, and are prone to voltage fluctuations or voltage deviations from the normal operating range. In these areas, various measures are usually required to improve the stability of the power grid to ensure the normal operation of power equipment and the safety of user equipment.

[0003] Traditional methods usually use fixed compensation settings or simple time scheduling to manage reactive compensation equipment. This method cannot be dynamically adjusted according to real-time grid conditions, which results in the inability to adjust the compensation in a timely and effective manner when the grid load and voltage change, which may cause voltage fluctuations or equipment damage. In addition, traditional methods often do not take into account the operating cost model and line loss cost model, so they cannot optimize the economic benefits. This means that the system may have unnecessary energy consumption or cost waste in reactive compensation, and cannot maximize the economy of system operation. In addition, traditional methods usually lack adaptive optimization capabilities for different grid conditions and cannot quickly respond to transient changes in the grid or load fluctuations, resulting in insufficient grid stability and equipment protection capabilities. Summary of the invention

[0004] The purpose of the present invention is to solve the problems existing in the background technology and to propose a reactive power compensation control method and device for a weak power grid.

[0005] The technical solution of the present invention is a reactive power compensation control method for a weak power grid, comprising: Performing transient voltage monitoring on multiple monitoring nodes in the target weak power grid through a transient voltage monitoring device to obtain transient voltages of the multiple monitoring nodes, and calculating transient voltage stability indicators of the monitoring nodes based on the transient voltages of the monitoring nodes; Comparing the transient voltage stability index of the monitoring node with a preset stability threshold, and if the transient voltage stability index of the monitoring node is lower than the preset stability threshold, constructing a first coordinated control model of the monitoring node; Solving the first coordinated control model by means of a genetic algorithm to obtain a first voltage compensation scheme, and controlling the first voltage regulating device based on the first voltage compensation scheme; If the transient voltage stability index of the monitoring node is not lower than a preset stability threshold, constructing a second coordinated control model of the monitoring node; The second coordinated control model is solved by deep reinforcement learning to obtain a second voltage compensation scheme, and the second voltage regulating device is controlled based on the second voltage compensation scheme.

[0006] Preferably, the transient voltage stability index calculation formula of the monitoring node is as follows: ; in, Indicates monitoring node The transient voltage stability index, Indicates monitoring node The transient voltage before the transient voltage drops, Indicates monitoring node exist The sampled transient voltage at time Indicates area The weight of the drop area of ​​the internal transient voltage curve, Indicates area The weight of the drop area of ​​the internal transient voltage curve, Indicates area The internal transient voltage is lower than the preset area Within the initial moment of the preset threshold voltage, Indicates area Internal transient voltage is higher than the preset area The initial moment within the preset threshold voltage.

[0007] Preferably, constructing the first coordinated control model of the monitoring node includes: An operating cost model is constructed based on the costs generated by the reactive power and active power regulated by the first voltage regulating device, wherein the first voltage regulating device includes a reactive power compensation device and an energy storage device, and the operating cost model is as follows: ; in, represents the operating cost model, Indicates the reactive power of the reactive compensation equipment, Represents the reactive power of the energy storage device, Represents the active power of the energy storage device, represents the reactive power cost of reactive power compensation equipment, represents the reactive power cost of the energy storage device, represents the active power cost of the energy storage device, represents the electricity price of the energy storage device, Indicates monitoring node The number of reactive power compensation devices for voltage regulation, Indicates monitoring node The number of energy storage devices for voltage regulation, Indicates the voltage regulation duration; A line loss cost model is constructed based on the line loss cost between the monitoring node and the adjacent monitoring nodes. The line loss cost model is as follows: ; in, represents the line loss cost model, Indicates the real-time electricity price, Indicates branch The conductivity, Indicates monitoring node The voltage amplitude, Indicates adjacent monitoring nodes The voltage amplitude, Indicates monitoring node and adjacent monitoring nodes The phase angle difference between An economic optimization objective function is constructed based on the operating cost model and the line loss cost model. The economic optimization objective function is as follows: ; in, represents the economic optimization objective function; The economic optimization constraint conditions of the economic optimization objective function are constructed.

[0008] Preferably, the economic optimization constraint condition includes a first equality constraint condition and a first inequality constraint condition, and the first equality constraint condition is as follows: ; in, Indicates monitoring node The active power, Indicates monitoring node The reactive power, Indicates branch The admittance of Indicates the total number of monitoring nodes in the target weak power grid; The first inequality constraint is as follows: ; in, Indicates the adjustment range of reactive power compensation equipment. and Indicates the minimum and maximum values ​​of the reactive power compensation equipment adjustment range. and Respectively represent monitoring nodes The minimum and maximum values ​​of the voltage amplitude; The first coordinated control model is as follows: ; in, Represents a constraint.

[0009] Preferably, solving the first coordinated control model by a genetic algorithm to obtain a first voltage compensation scheme comprises the following steps: A1. Set the particle swarm population size to The acceleration factor is and And the maximum number of iterations is , and encode the first voltage regulating device to obtain a control variable, the control variable includes the number of the first voltage regulating device, the output power of the reactive compensation device and the energy storage power of the energy storage device, and the acceleration coefficient formula is as follows: ; in, and Represent the acceleration coefficient and The upper limit of and Represent the acceleration coefficient and The lower limit of Indicates the current iteration number; A2. Initialize the particle swarm to obtain particles, that is, A feasible solution is found, and the initialized particles are substituted into the power flow calculation to find the fitness value of each particle, and the fitness value of the particle is used as the current optimal position of the particle, and the current position of the particle with the smallest fitness is selected as the optimal position of the particle group and substituted into the next calculation; A3. Based on the number of particle swarm iterations, the inertia weight is updated, and the speed and position of each particle are updated. If the particle speed exceeds the preset speed threshold, the speed of the particle is set to the extreme value and substituted into the next calculation. The update formula of the inertia weight is as follows: ; in, represents the inertia weight, and Respectively represent the minimum and maximum values ​​of the inertia weight; A4. Calculate the fitness value of the updated particle and update the particle information, wherein the particle information includes the fitness value, the historical optimal position of the particle and the historical optimal position of the particle swarm; A5. If the number of iterations reaches the maximum number of iterations, stop the iteration and output the obtained first voltage compensation solution; otherwise, repeat steps A3 to A5.

[0010] Preferably, constructing the second coordinated control model of the monitoring node includes: Construct the transient voltage stability optimization objective function of the monitoring node, and the transient voltage stability optimization objective function is as follows: ; in, represents the optimization objective function of minimizing the transient voltage stability difference of the monitoring node, represents the objective function of maximizing the transient voltage stability of the monitoring node; The target weak grid voltage mean square error optimization objective function is constructed, and the target weak grid voltage mean square error optimization objective function is as follows: ; in, represents the target weak grid voltage mean square error optimization objective function; Voltage optimization constraints are constructed based on the transient voltage stability optimization objective function of the monitoring node and the target weak power grid voltage mean square error optimization objective function.

[0011] Preferably, the voltage optimization constraint condition includes a second equality constraint condition and a second inequality constraint condition, and the second equality constraint condition is as follows: ; in, represents the reactive power output of the i-th group of PVGU, Indicates the reactive power output of SVG. Indicates the reactive power actually generated by the power station; The second inequality constraint is as follows: ; in, represents the active output of the i-th group of PVGU, and Respectively represent the minimum and maximum reactive power output of the i-th group of PVGU, and Respectively represent the minimum and maximum values ​​of the reactive power output of SVG; The second coordinated control model is as follows: ; Preferably, solving the second coordinated control model by deep reinforcement learning to obtain a second voltage compensation scheme includes: A deep reinforcement learning model for coordinated control is established by using a Markov decision process, wherein the deep reinforcement learning model for coordinated control includes an Actor network and a Critic network, and the Actor network and the Critic network are initialized, and an experience pool is created; The Actor network outputs actions based on the current state and executes actions when outputting actions , the data of the interaction between the agent and the environment ,in and Represent the current state and the next state respectively. represents the reward function; Based on the experience sampling probability, experience data is obtained from the experience pool, and the network parameters of the Actor network and the Critic network are updated based on the experience data. The network parameters include importance sampling weights and priority probabilities. The experience sampling probability calculation formula is as follows: ; in, represents the experience sampling probability, Indicates playback to the buffer in the experience library Priority of the moment, Indicates playback to the buffer in the experience library Priority of the moment, represents the priority adjustment factor; The importance sampling weight update formula is as follows: ; in, represents the importance sampling weight, Represents the weight coefficient, which is used to control the degree of correction; The priority probability calculation formula is as follows: ; in, represents the priority probability, represents the difference between the target Q value and the estimated Q value, Represents a random positive number; The Actor network and the Critic network are updated based on the network parameters to obtain a target network, and the second coordinated control model is solved based on the target network to obtain a second voltage compensation scheme.

[0012] The technical solution of the present invention is a reactive power compensation control device for a weak power grid, which is applicable to the reactive power compensation control method for a weak power grid, and is characterized by comprising: A voltage monitoring module, wherein the voltage monitoring module is used to monitor the transient voltage of multiple monitoring nodes in the target weak power grid; An index calculation module, the index calculation module is used to calculate the transient voltage stability index of each monitoring node based on the transient voltage of the monitoring node, and compare the transient voltage stability index of the monitoring node with a preset stability threshold; A coordination control module, wherein the coordination control module is used to construct a first coordination control model of the monitoring node when the transient voltage stability index of the monitoring node is lower than a preset stability threshold, and to construct a second coordination control model of the monitoring node when the transient voltage stability index of the monitoring node is not lower than the preset stability threshold; A model solving module, the model solving module is used to solve the first coordinated control model by a genetic algorithm to obtain a first voltage compensation scheme, and to solve the second coordinated control model by deep reinforcement learning to obtain a second voltage compensation scheme; A voltage regulation control module is used to control the first voltage regulation device based on a first voltage compensation scheme, and to control the second voltage regulation device based on a second voltage compensation scheme.

[0013] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects: 1. The present invention can effectively maintain the voltage stability of each node in the weak power grid by monitoring the transient voltage stability index of the node and calculating and adjusting the compensation scheme according to the real-time data, thereby helping to avoid potential damage to equipment and users caused by voltage fluctuations. By constructing an operating cost model and a line loss cost model and performing optimization control based on the economic optimization objective function, the system operating cost can be effectively reduced. Not only the operating costs of reactive compensation equipment and energy storage equipment are taken into account, but also the line loss is optimized, thereby improving the economic benefits of the system.

[0014] 2. The present invention optimizes different voltage stability conditions respectively, so as to enhance the adaptability and responsiveness of reactive compensation. The genetic algorithm is suitable for the conditions below the stability threshold, while the deep reinforcement learning is used to deal with the conditions with better stability, so as to achieve the optimal voltage compensation scheme under various operating conditions. Through precise voltage control and economic optimization, not only the stability of the power grid can be improved, but also the reliability and anti-interference ability of the system can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic flow chart of an overall method in an embodiment of the present invention; Figure 2 A schematic diagram of the overall process of an embodiment of the present invention.

[0016] Figure numerals: 1. Voltage monitoring module; 2. Index calculation module; 3. Coordination control module; 4. Model solving module; 5. Voltage regulation control module. DETAILED DESCRIPTION

[0017] Embodiment 1, as Figure 1 As shown, the present invention proposes a reactive power compensation control method for a weak power grid, comprising: S1. Perform transient voltage monitoring on multiple monitoring nodes in the target weak power grid through a transient voltage monitoring device to obtain transient voltages of the multiple monitoring nodes, and calculate transient voltage stability indicators of the monitoring nodes based on the transient voltages of the monitoring nodes; S2. Compare the transient voltage stability index of the monitoring node with a preset stability threshold. If the transient voltage stability index of the monitoring node is lower than the preset stability threshold, construct a first coordinated control model of the monitoring node. S3. Solving the first coordinated control model by using a genetic algorithm to obtain a first voltage compensation scheme, and controlling the first voltage regulating device based on the first voltage compensation scheme; S4. If the transient voltage stability index of the monitoring node is not lower than the preset stability threshold, constructing a second coordinated control model of the monitoring node; S5. Solve the second coordinated control model through deep reinforcement learning to obtain a second voltage compensation scheme, and control the second voltage regulating device based on the second voltage compensation scheme.

[0018] In the present invention, genetic algorithm is an evolutionary computing technology that simulates the process of biological evolution in nature. Its basic idea is to screen out the optimal solution from a population by simulating processes such as natural selection, crossover and mutation. In power systems, genetic algorithms are usually used to solve complex optimization problems, such as optimizing the output distribution of generators in power grid scheduling, minimizing transmission losses or optimizing the stability of power grids; deep reinforcement learning is a branch of machine learning that focuses on learning decision-making strategies through the interaction between agents and the environment so that agents can maximize the expected cumulative rewards in unknown environments. It combines the capabilities of deep learning and the framework of reinforcement learning, and can handle complex, high-dimensional state spaces and action spaces.

[0019] Embodiment 2, a reactive power compensation control method for a weak power grid proposed by the present invention, compared with embodiment 1, this embodiment further includes: a transient voltage stability index calculation formula of a monitoring node is as follows: ; in, Indicates monitoring node The transient voltage stability index, Indicates monitoring node The transient voltage before the transient voltage drops, Indicates monitoring node exist The sampled transient voltage at time Indicates area The weight of the drop area of ​​the internal transient voltage curve, Indicates area The weight of the drop area of ​​the internal transient voltage curve, Indicates area The internal transient voltage is lower than the preset area Within the initial moment of the preset threshold voltage, Indicates area Internal transient voltage is higher than the preset area The initial moment within the preset threshold voltage.

[0020] In an optional embodiment, constructing a first coordinated control model of the monitoring node includes: An operating cost model is constructed based on the costs generated by the reactive power and active power regulated by the first voltage regulating device. The first voltage regulating device includes a reactive compensation device and an energy storage device. The operating cost model is as follows: ; in, represents the operating cost model, Indicates the reactive power of the reactive compensation equipment, Represents the reactive power of the energy storage device, Represents the active power of the energy storage device, represents the reactive power cost of reactive power compensation equipment, represents the reactive power cost of the energy storage device, represents the active power cost of the energy storage device, represents the electricity price of the energy storage device, Indicates monitoring node The number of reactive power compensation devices for voltage regulation, Indicates monitoring node The number of energy storage devices for voltage regulation, Indicates the voltage regulation duration; A line loss cost model is constructed based on the line loss cost between the monitoring node and the adjacent monitoring nodes. The line loss cost model is as follows: ; in, represents the line loss cost model, Indicates the real-time electricity price, Indicates branch The conductivity, Indicates monitoring node The voltage amplitude, Indicates adjacent monitoring nodes The voltage amplitude, Indicates monitoring node and adjacent monitoring nodes The phase angle difference between The economic optimization objective function is constructed based on the operation cost model and the line loss cost model. The economic optimization objective function is as follows: ; in, represents the economic optimization objective function; Construct the economic optimization constraints of the economic optimization objective function.

[0021] In an optional embodiment, the economic optimization constraint condition includes a first equality constraint condition and a first inequality constraint condition, and the first equality constraint condition is as follows: ; in, Indicates monitoring node The active power, Indicates monitoring node The reactive power, Indicates branch The admittance of Indicates the total number of monitoring nodes in the target weak power grid; The first inequality constraint is as follows: ; in, Indicates the adjustment range of reactive power compensation equipment. and Indicates the minimum and maximum values ​​of the reactive power compensation equipment adjustment range. and Respectively represent monitoring nodes The minimum and maximum values ​​of the voltage amplitude; The first coordinated control model is as follows: ; in, Represents a constraint.

[0022] In an optional embodiment, solving the first coordinated control model by a genetic algorithm to obtain a first voltage compensation scheme includes the following steps: A1. Set the particle swarm population size to The acceleration factor is and And the maximum number of iterations is , and encode the first voltage regulating device to obtain the control variables, which include the number of the first voltage regulating device, the output power of the reactive compensation device and the energy storage power of the energy storage device. The acceleration coefficient formula is as follows: ; in, and Represent the acceleration coefficient and The upper limit of and Represent the acceleration coefficient and The lower limit of Indicates the current iteration number; A2. Initialize the particle swarm to obtain particles, that is, A feasible solution is found, and the initialized particles are substituted into the power flow calculation to find the fitness value of each particle, and the fitness value of the particle is used as the current optimal position of the particle, and the current position of the particle with the smallest fitness is selected as the optimal position of the particle group and substituted into the next calculation; A3. Based on the number of particle swarm iterations, the inertia weight is updated, and the speed and position of each particle are updated. If the particle speed exceeds the preset speed threshold, the particle speed is set to the extreme value and substituted into the next calculation. The update formula of the inertia weight is as follows: ; in, represents the inertia weight, and Respectively represent the minimum and maximum values ​​of the inertia weight; A4. Calculate the fitness value of the updated particles and update the particle information, which includes the fitness value, the historical optimal position of the particle and the historical optimal position of the particle group; A5. If the number of iterations reaches the maximum number of iterations, stop the iteration and output the obtained first voltage compensation solution; otherwise, repeat steps A3 to A5.

[0023] In an optional embodiment, constructing a second coordinated control model of the monitoring node includes: Construct the transient voltage stability optimization objective function of the monitoring node. The transient voltage stability optimization objective function is as follows: ; in, represents the optimization objective function of minimizing the transient voltage stability difference of the monitoring node, represents the objective function of maximizing the transient voltage stability of the monitoring node; The target weak grid voltage mean square error optimization objective function is constructed. The target weak grid voltage mean square error optimization objective function is as follows: ; in, represents the target weak grid voltage mean square error optimization objective function; Voltage optimization constraints are constructed based on the transient voltage stability optimization objective function of the monitoring node and the target weak grid voltage mean square error optimization objective function.

[0024] In an optional embodiment, the voltage optimization constraint includes a second equality constraint and a second inequality constraint, and the second equality constraint is as follows: ; in, represents the reactive power output of the i-th group of PVGU, Indicates the reactive power output of SVG. Indicates the reactive power actually generated by the power station; The second inequality constraint is as follows: ; in, represents the active output of the i-th group of PVGU, and Respectively represent the minimum and maximum reactive power output of the i-th group of PVGU, and Respectively represent the minimum and maximum values ​​of the reactive power output of SVG; The second coordinated control model is as follows: ; In an optional embodiment, solving the second coordinated control model by deep reinforcement learning to obtain a second voltage compensation scheme includes: A deep reinforcement learning model for coordinated control is established using the Markov decision process. The deep reinforcement learning model for coordinated control includes an Actor network and a Critic network. The Actor network and the Critic network are initialized, and an experience pool is created. The Actor network outputs actions based on the current state and executes actions when outputting actions , the data of the interaction between the agent and the environment ,in and Represent the current state and the next state respectively. represents the reward function; Based on the experience sampling probability, experience data is obtained from the experience pool, and the network parameters of the Actor network and the Critic network are updated based on the experience data. The network parameters include importance sampling weights and priority probabilities. The experience sampling probability calculation formula is as follows: ; in, represents the experience sampling probability, Indicates playback to the buffer in the experience library Priority of the moment, Indicates playback to the buffer in the experience library Priority of the moment, represents the priority adjustment factor; The importance sampling weight update formula is as follows: ; in, represents the importance sampling weight, Represents the weight coefficient, which is used to control the degree of correction; The priority probability calculation formula is as follows: ; in, represents the priority probability, represents the difference between the target Q value and the estimated Q value, Represents a random positive number; The Actor network and the Critic network are updated based on the network parameters to obtain a target network, and the second coordinated control model is solved based on the target network to obtain a second voltage compensation scheme.

[0025] It should be noted that the Markov decision process (MDP) is a mathematical framework used to describe a decision problem, in which the decision is based on the current state and the goal is to maximize the expected cumulative reward through a series of decisions. MDP includes elements such as state set, action set, state transition probability, and reward function. In deep reinforcement learning, the Actor network is a neural network used to learn the policy function, that is, the function that maps the state to the action. The Critic network is also a neural network, which is used in deep reinforcement learning to learn the value function and evaluate the long-term return of each state. The experience pool is used to store the experience data of the interaction between the agent and the environment, including state, action, reward, next state, etc. These data are used to train the Actor and Critic networks. In order to improve the stability of deep reinforcement learning, two sets of network structures are usually used: one is the current network for updating, and the other is the target network for estimating the target Q value. The parameters of the target network are updated by gradually updating the parameters of the current network.

[0026] Embodiment three, as Figure 2 As shown, a reactive power compensation control device for a weak power grid, which is applicable to the reactive power compensation control method for a weak power grid mentioned above, comprises: Voltage monitoring module 1, voltage monitoring module 1 is used to monitor the transient voltage of multiple monitoring nodes in the target weak power grid; The indicator calculation module 2 is used to calculate the transient voltage stability index of each monitoring node based on the transient voltage of the monitoring node, and compare the transient voltage stability index of the monitoring node with a preset stability threshold; Coordination control module 3, coordination control module 3 is used to construct a first coordination control model of the monitoring node when the transient voltage stability index of the monitoring node is lower than the preset stability threshold, and to construct a second coordination control model of the monitoring node when the transient voltage stability index of the monitoring node is not lower than the preset stability threshold; A model solving module 4, the model solving module 4 is used to solve the first coordinated control model by a genetic algorithm to obtain a first voltage compensation scheme, and to solve the second coordinated control model by deep reinforcement learning to obtain a second voltage compensation scheme; The voltage regulation control module 5 is used to control the first voltage regulation device based on the first voltage compensation scheme, and to control the second voltage regulation device based on the second voltage compensation scheme.

[0027] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A reactive power compensation control method for a weak power grid, It is characterized in that include: Performing transient voltage monitoring on multiple monitoring nodes in the target weak power grid through a transient voltage monitoring device to obtain transient voltages of the multiple monitoring nodes, and calculating transient voltage stability indicators of the monitoring nodes based on the transient voltages of the monitoring nodes; Comparing the transient voltage stability index of the monitoring node with a preset stability threshold, and if the transient voltage stability index of the monitoring node is lower than the preset stability threshold, constructing a first coordinated control model of the monitoring node; Solving the first coordinated control model by a genetic algorithm to obtain a first voltage compensation scheme, and controlling the first voltage regulating device based on the first voltage compensation scheme; If the transient voltage stability index of the monitoring node is not lower than a preset stability threshold, constructing a second coordinated control model of the monitoring node; The second coordinated control model is solved by deep reinforcement learning to obtain a second voltage compensation scheme, and the second voltage regulating device is controlled based on the second voltage compensation scheme.

2. A reactive power compensation control method for a weak power grid according to claim 1, Features: The calculation formula of the transient voltage stability index of the monitoring node is as follows: ； in, Indicates monitoring node The transient voltage stability index, Indicates monitoring node The transient voltage before the transient voltage drops, Indicates monitoring node exist The sampled transient voltage at time Indicates area The weight of the drop area of ​​the internal transient voltage curve, Indicates area The weight of the drop area of ​​the internal transient voltage curve, Indicates area The internal transient voltage is lower than the preset area Within the initial moment of the preset threshold voltage, Indicates area Internal transient voltage is higher than the preset area The initial moment within the preset threshold voltage.

3. The reactive power compensation control method for a weak power grid according to claim 1, Features: Constructing a first coordinated control model of the monitoring node, including: An operating cost model is constructed based on the costs generated by the reactive power and active power regulated by the first voltage regulating device, wherein the first voltage regulating device includes a reactive power compensation device and an energy storage device, and the operating cost model is as follows: ； in, represents the operating cost model, Indicates the reactive power of the reactive compensation equipment. Represents the reactive power of the energy storage device, Represents the active power of the energy storage device, represents the reactive power cost of reactive power compensation equipment, represents the reactive power cost of the energy storage device, represents the active power cost of the energy storage device, represents the electricity price of the energy storage device, Indicates monitoring node The number of reactive power compensation devices for voltage regulation, Indicates monitoring node The number of energy storage devices for voltage regulation, Indicates the voltage regulation duration; A line loss cost model is constructed based on the line loss cost between the monitoring node and the adjacent monitoring nodes. The line loss cost model is as follows: ； in, represents the line loss cost model, Indicates the real-time electricity price, Shizhi Road The conductivity, Indicates monitoring node The voltage amplitude, Indicates adjacent monitoring nodes The voltage amplitude, Indicates monitoring node and adjacent monitoring nodes The phase angle difference between An economic optimization objective function is constructed based on the operating cost model and the line loss cost model. The economic optimization objective function is as follows: ； in, represents the economic optimization objective function; The economic optimization constraint conditions of the economic optimization objective function are constructed.

4. A reactive power compensation control method for a weak power grid according to claim 3, Features: The economic optimization constraint condition includes a first equality constraint condition and a first inequality constraint condition, and the first equality constraint condition is as follows: ； in, Indicates monitoring node The active power, Indicates monitoring node The reactive power, Indicates branch The admittance of Indicates the total number of monitoring nodes in the target weak power grid; The first inequality constraint is as follows: ； in, Indicates the adjustment range of reactive power compensation equipment. and Indicates the minimum and maximum values ​​of the reactive power compensation equipment adjustment range. and Respectively represent monitoring nodes The minimum and maximum values ​​of the voltage amplitude; The first coordinated control model is as follows: ； in, Represents a constraint.

5. A reactive power compensation control method for a weak power grid according to claim 4, Features: Solving the first coordinated control model by a genetic algorithm to obtain a first voltage compensation scheme includes the following steps: A1. Set the particle swarm population size to The acceleration factor is and And the maximum number of iterations is , and encode the first voltage regulating device to obtain a control variable, the control variable includes the number of the first voltage regulating device, the output power of the reactive compensation device and the energy storage power of the energy storage device, and the acceleration coefficient formula is as follows: ； in, and Represent the acceleration coefficient and The upper limit of and Represent the acceleration coefficient and The lower limit of Indicates the current iteration number; A2. Initialize the particle swarm to obtain particles, that is, A feasible solution is found, and the initialized particles are substituted into the power flow calculation to find the fitness value of each particle, and the fitness value of the particle is used as the current optimal position of the particle, and the current position of the particle with the smallest fitness is selected as the optimal position of the particle group and substituted into the next calculation; A3. Based on the number of particle swarm iterations, the inertia weight is updated, and the speed and position of each particle are updated. If the particle speed exceeds the preset speed threshold, the speed of the particle is set to the extreme value and substituted into the next calculation. The update formula of the inertia weight is as follows: ； in, represents the inertia weight, and Respectively represent the minimum and maximum values ​​of the inertia weight; A4. Calculate the fitness value of the updated particle and update the particle information, wherein the particle information includes the fitness value, the historical optimal position of the particle and the historical optimal position of the particle swarm; A5. If the number of iterations reaches the maximum number of iterations, stop the iteration and output the obtained first voltage compensation solution; otherwise, repeat steps A3 to A5.

6. A reactive power compensation control method for a weak power grid according to claim 4, Features: Constructing a second coordinated control model of the monitoring node, including: Construct the transient voltage stability optimization objective function of the monitoring node, and the transient voltage stability optimization objective function is as follows: ； in, represents the optimization objective function of minimizing the transient voltage stability difference of the monitoring node, represents the objective function of maximizing the transient voltage stability of the monitoring node; The target weak grid voltage mean square error optimization objective function is constructed, and the target weak grid voltage mean square error optimization objective function is as follows: ； in, represents the target weak grid voltage mean square error optimization objective function; Voltage optimization constraints are constructed based on the transient voltage stability optimization objective function of the monitoring node and the target weak power grid voltage mean square error optimization objective function.

7. A reactive power compensation control method for a weak power grid according to claim 6, Features: The voltage optimization constraint condition includes a second equality constraint condition and a second inequality constraint condition, and the second equality constraint condition is as follows: ； in, represents the reactive power output of the i-th group of PVGU, Indicates the reactive power output of SVG. Indicates the reactive power actually generated by the power station; The second inequality constraint is as follows: ； in, represents the active output of the i-th group of PVGU, and Respectively represent the minimum and maximum reactive power output of the i-th group of PVGU, and Respectively represent the minimum and maximum values ​​of the reactive power output of SVG; The second coordinated control model is as follows: 。 8. A reactive power compensation control method for a weak power grid according to claim 7, Features: The second coordinated control model is solved by deep reinforcement learning to obtain a second voltage compensation scheme, including: A deep reinforcement learning model for coordinated control is established by using a Markov decision process, wherein the deep reinforcement learning model for coordinated control includes an Actor network and a Critic network, and the Actor network and the Critic network are initialized, and an experience pool is created; The Actor network outputs actions based on the current state and executes actions when outputting actions , the data of the interaction between the agent and the environment ,in and Represent the current state and the next state respectively. represents the reward function; Based on the experience sampling probability, experience data is obtained from the experience pool, and the network parameters of the Actor network and the Critic network are updated based on the experience data. The network parameters include importance sampling weights and priority probabilities. The experience sampling probability calculation formula is as follows: ； in, represents the experience sampling probability, Indicates playback to the buffer in the experience library Priority of the moment, Indicates playback to the buffer in the experience library Priority of the moment, represents the priority adjustment factor; The importance sampling weight update formula is as follows: ； in, represents the importance sampling weight, Represents the weight coefficient, which is used to control the degree of correction; The priority probability calculation formula is as follows: ； in, represents the priority probability, represents the difference between the target Q value and the estimated Q value, Represents a random positive number; The Actor network and the Critic network are updated based on the network parameters to obtain a target network, and the second coordinated control model is solved based on the target network to obtain a second voltage compensation scheme.

9. A reactive power compensation control device for a weak power grid, which is applicable to a reactive power compensation control method for a weak power grid as claimed in any one of claims 1 to 8. It is characterized in that include: A voltage monitoring module (1), the voltage monitoring module (1) being used to monitor transient voltages of a plurality of monitoring nodes in a target weak power grid; An index calculation module (2), the index calculation module (2) being used to calculate a transient voltage stability index of each monitoring node based on the transient voltage of the monitoring node, and to compare the transient voltage stability index of the monitoring node with a preset stability threshold; A coordination control module (3), the coordination control module (3) being used to construct a first coordination control model of the monitoring node when the transient voltage stability index of the monitoring node is lower than a preset stability threshold, and to construct a second coordination control model of the monitoring node when the transient voltage stability index of the monitoring node is not lower than the preset stability threshold; A model solving module (4), the model solving module (4) being used to solve the first coordinated control model by means of a genetic algorithm to obtain a first voltage compensation scheme, and to solve the second coordinated control model by means of deep reinforcement learning to obtain a second voltage compensation scheme; A voltage regulation control module (5) is used to control a first voltage regulation device based on a first voltage compensation scheme, and to control a second voltage regulation device based on a second voltage compensation scheme.

Images