Data-model driving-based power distribution network medium-low voltage cooperative voltage control method

Abstract

The invention relates to a data-model driving-based power distribution network medium-low voltage cooperative voltage control method, which comprises the following steps that: 1) a medium-voltage distribution network acquires load parameters of each node and medium-voltage distribution network regulation and control object parameters, and establishes a medium-voltage distribution network control model; 2) solving optimization parameters of each node of the medium-voltage distribution network by using a genetic algorithm; 3) issuing the optimization parameters of the medium-voltage solution to the low-voltage distribution network through the power distribution cloud master station, and starting a DDPG voltage control algorithm of the low-voltage distribution network according to the parameters of the regulated and controlled equipment of the low-voltage transformer area and the optimization parameters of the medium-voltage solution; 4) after training is carried out through a large amount of data, the optimal photovoltaic and energy storage output is found to ensure that the voltage is maintained in a safe range; and 5) sending the voltage and power data of each node of the low-voltage transformer area and the optimal photovoltaic and energy storage power to a power distribution cloud master station, and correcting the optimized parameters of the medium-voltage solution to ensure that the voltage and line loss of the medium-low-voltage transformer area meet the requirements and ensure the active, safe and stable operation of the medium-low-voltage active power distribution network.

Description

Technical Field

[0001] This invention relates to a distribution network voltage control method, belonging to the field of distribution network operation and analysis. Background Technology

[0002] Photovoltaic power generation, as one of the most promising new energy power generation methods for large-scale application, is gradually replacing the fossil fuel-based energy structure. Under this new energy structure, the massive, intermittent, and random distributed photovoltaic (PV) grid integration has led to increasingly complex distribution network structures and a significant decrease in system voltage regulation capabilities, making power quality issues such as high / low voltage problems more prominent. Excess power from a high proportion of distributed PV in low-voltage residential distribution networks is fed back to the medium-voltage distribution network, further exacerbating voltage exceeding limits at medium-voltage nodes and heavy reverse load overload problems in transformer substations.

[0003] Existing research largely focuses on single voltage levels, such as medium or low voltage, without fully considering the coupling characteristics of medium and low voltage networks, making it difficult to achieve overall control performance at the medium and low voltage levels. Current model-based control methods rely on precise network parameters and have weak model generalization capabilities, making them difficult to deploy in real-world networks. Meanwhile, data-driven deep reinforcement learning algorithms face challenges in reward function design, and their predictions may not meet physical constraints such as system power balance, lacking physical interpretability. Therefore, to address the increasingly complex voltage limit exceedance issues in medium and low voltage distribution networks, combining physical models and data-driven approaches to solve the problem of coordinated medium and low voltage control is a pressing challenge. Summary of the Invention

[0004] This invention considers the differences in voltage regulation resources in medium and low voltage distribution networks, establishes a physical model for voltage regulation of traditional equipment such as on-load tap changers (OLTC) and static var compensators (SVC) in medium voltage distribution networks, trains the voltage dynamic response of photovoltaic inverters and energy storage in low voltage distribution networks based on a multi-agent reinforcement learning algorithm, and transforms the established voltage control mathematical model into a Markov decision process, thereby reducing the difficulty of power modeling for multiple voltage regulation devices.

[0005] A data-model-driven low-voltage coordinated voltage control method for distribution networks includes the following steps:

[0006] 1) Obtain the load parameters of each node in the medium-voltage distribution network and the parameters of the medium-voltage distribution network control objects, and establish a medium-voltage distribution network control model;

[0007] 2) Given the objective function and constraints of the medium-voltage distribution network, use a genetic algorithm to solve for the optimal parameters of each node in the medium-voltage distribution network;

[0008] 3) The optimized parameters obtained from the medium-voltage solution are sent to the low-voltage distribution network through the distribution cloud master station. Based on the parameters of the controlled equipment in the low-voltage distribution area and the optimized parameters obtained from the medium-voltage solution, the DDPG voltage control algorithm of the low-voltage distribution network is started.

[0009] 4) After training with a large amount of data, find the optimal output of photovoltaic and energy storage to ensure that the voltage is maintained within a safe range;

[0010] 5) Upload the voltage and power data of each node in the low-voltage distribution area, as well as the optimal photovoltaic and energy storage power, to the distribution cloud master station and correct the optimization parameters of the medium-voltage solution. Finally, ensure that the voltage and line loss of the medium and low-voltage distribution areas meet the requirements.

[0011] Step 1) The medium-voltage distribution network control model includes the physical models of OLTC and SVC.

[0012] OLTC model: This model uses a discrete on-load tap changer to control the secondary voltage of the transformer and maintain it within the allowable range during operation. The tap adjustment process is as follows:

[0013] (1)

[0014] (2)

[0015] (3)

[0016] (4)

[0017] In the formula, The secondary voltage value of the inverter Compared with reference value The difference; This refers to a discrete moment during OLTC runtime; For counters; This is a constant determined by the characteristics and voltage drop of the OLTC, when the OLTC changes from time... Start running, if the time is greater than or equal to When the counter reaches its maximum value, it will increment by one. This is a voltage dead zone to prevent the OLTC from operating unnecessaryly within the allowable voltage range; This refers to the transformer tap position; The number of taps changed for the OLTC tap; This refers to the delay time for actions during OLTC runtime.

[0018] SVC Model: The SVC model adopts the thyristor phase-controlled reactor model. The SVC is connected to the distribution network through an inverter. The equivalent transfer function of the control loop of the inverter in reactive power control mode is as follows:

[0019] (5)

[0020] In the formula, This is the difference between the current equivalent susceptance of the SVC inverter and the equivalent susceptance at the previous moment. To control the time constant of the loop; This is the difference between the current control variable and the control variable at the previous moment for the SVC inverter. This refers to the output voltage of the SVC inverter. This is the difference between the current reactive power output of the SVC inverter and the reactive power output at the previous moment.

[0021] Step 2)

[0022] 2.1) Objective Function: The overall control objective of the medium-voltage distribution network is the distribution network line loss. Minimum and node voltage deviation rates Minimum;

[0023] (6)

[0024] In the formula: The number of branches; and They are respectively time Active and reactive power of branch circuits; for Branch resistance; for time The voltage at the beginning of the branch;

[0025] (7)

[0026] In the formula, Represents any node in a medium-voltage distribution network; Indicates the total number of system nodes; Indicates node voltage; Indicates the nominal voltage; These represent the maximum and minimum values ​​of the node voltage, respectively.

[0027] The two objective functions are weighted separately, and a combined objective function is formed. for:

[0028] (8)

[0029] In the formula: and These are weighting coefficients, selected based on the optimization requirements of the medium-voltage distribution network;

[0030] 2.2) Constraints

[0031] a) Constraints of AC power flow equations

[0032] (9)

[0033] In the formula: and These are nodes Injected active and reactive power; and These are nodes The active and reactive power of the load; For nodes and nodes Phase difference of the voltage vector; and They are nodes and nodes The conductance and susceptance of the lines between them; This is the total number of nodes in the power grid;

[0034] b) Operational safety constraints

[0035] (10)

[0036] In the formula: Representing the power grid nodes The upper, real-time, and lower limits of the voltage;

[0037] c) SVC output constraint

[0038] (11)

[0039] (12)

[0040] In the formula, and They are nodes SVC output reactive power and output reactive power ramping; and They are nodes The upper and lower limits of reactive power output from the SVC; and They are respectively Time Node The upper and lower limits of reactive power ramping output of the upper SVC;

[0041] d) OLTC tap changer adjustment constraint

[0042] (13)

[0043] In the formula, This refers to the tap position of the voltage regulator; and These are the upper and lower limits of the tap position on the voltage regulator;

[0044] 3. The low-voltage coordinated voltage control method for distribution networks based on data-model driven technology according to claim 2, characterized in that: in step 2)

[0045] 2.3) Solving the medium-voltage distribution network optimization model using a genetic algorithm

[0046] ① Input system data, including grid parameters, OLTC parameters, load parameters, energy storage parameters, and SVC parameters, and encode the number of OLTC taps and the reactive power capacity of SVC;

[0047] ② Based on the set genetic algorithm parameters, an initial population is generated. The comprehensive objective function of the medium-voltage distribution network is calculated according to the OLTC gear position and SVC output power of each individual, and the fitness is calculated on this basis.

[0048] ③ Utilize the elite preservation strategy to perform selection, segmented crossover, and segmented mutation operations on the corresponding population, selecting individuals that meet the constraints to enter the next generation of the population;

[0049] ④ Calculate the fitness function value and save the optimal individual;

[0050] ⑤ Based on the set termination condition, determine whether to end the loop. If yes, output the result; otherwise, go to step .

[0051] Step 3): The DDPG voltage control algorithm for low-voltage distribution networks includes...

[0052] a) State space

[0053] The DDPG voltage control algorithm is used to control voltage within a defined low-voltage distribution network area where voltage exceedance nodes are located. Real-time data collection of power output from each node in this low-voltage distribution network area is used to monitor the power output adjustment of voltage regulating equipment. The state space of the DDPG voltage control algorithm is designed as follows:

[0054] (14)

[0055] In the formula, For nodes per-unit voltage value For nodes The active power; For nodes reactive power; , This represents the number of busbar nodes within the distribution network area.

[0056] b) Action space

[0057] The active and reactive power output and energy storage output of distributed photovoltaic systems within the low-voltage distribution network area where the voltage cross-node is located are adjusted to achieve low-voltage voltage control. The action space of the DDPG voltage control algorithm is designed as follows:

[0058] (15)

[0059] When only photovoltaic reactive power regulation is used When photovoltaic reactive power regulation fails to achieve the desired voltage control effect, photovoltaic active power reduction is applied. ,at the same time, ,in, and They are nodes Changes in active and reactive power output of distributed photovoltaic systems; For nodes Changes in active power output of the upper energy storage system;

[0060] c) Reward function

[0061] The DDPG voltage control algorithm controls node voltage by adjusting the distributed photovoltaic and energy storage connected to each node. The control objective is to stabilize the voltage within a safe threshold. The instantaneous reward function is set as the sum of the quadratic form of the voltage exceeding the limit for each node and the quadratic form of the active and reactive power regulation of the distributed photovoltaic output and the active power regulation of the energy storage output, i.e.:

[0062] (16)

[0063] In the formula, bus node Voltage exceeding the limit; This refers to the number of busbar nodes within a low-voltage distribution network area. , These represent the number of distributed photovoltaic (PV) systems and energy storage systems within the low-voltage distribution network area, respectively, and the matrix... , , Both are weight matrices. For selection coefficients, when photovoltaic active power reduction is not adopted. When photovoltaic active power reduction is added .

[0064] The constraints of the low-voltage distribution network DDPG algorithm in step 3)

[0065] In addition to the AC power flow equation constraints and operational safety constraints in step 2, constraints also need to be applied to photovoltaic power output, energy storage power output, and state of charge (SOC), namely:

[0066] 1) Constraints of photovoltaic inverters

[0067] Reactive power output of photovoltaic inverter It is related to its apparent power and rated active power, and its constraints are as follows:

[0068] (17)

[0069] In the formula: For photovoltaic Apparent power; For photovoltaic The active power;

[0070] 2) Energy storage constraints

[0071] Active power regulation of energy storage needs to consider not only the device capacity but also the state of charge (SOC) of the energy storage, and its constraints are as follows:

[0072] (18)

[0073] In the formula: for Real-time energy storage SOC: for Real-time energy storage SOC: for Changes in SOC (State of Charge) of stored energy at any given time; for Real-time energy storage active power: For energy storage device capacity; For time intervals; For energy storage SOC; and Energy storage SOC lower and upper limits; For energy storage Output active power; and Energy storage The minimum and maximum values ​​of the output active power.

[0074] In step 5), the uploaded low-voltage electrical parameters are combined with the medium-voltage electrical parameters and the objective function to recalculate and adjust the previous calculation results.

[0075] To address the shortcomings of existing technologies that fail to consider coordinated voltage control in medium- and low-voltage distribution networks, resulting in poor global regulation and deficiencies in physical models or data-driven methods, this invention proposes a data-model-driven coordinated voltage control method for medium- and low-voltage distribution networks. This method comprehensively considers voltage regulation resources involved in medium- and low-voltage distribution networks, including OLTC, SVC, energy storage, and photovoltaics; it also provides methods for coordinating medium- and low-voltage distribution networks to achieve voltage control of the regional power grid, such as...Figure 1 As shown, this invention combines the advantages of both data-driven and physical models, making it better suited to scenarios where a high proportion of renewable energy is integrated into medium- and low-voltage distribution networks. The proposed collaborative control method exhibits excellent voltage control performance, effectively reducing distribution network losses and voltage deviations while simultaneously improving photovoltaic absorption and balancing capabilities, ensuring the active, safe, and stable operation of medium- and low-voltage active distribution networks. Attached Figure Description

[0076] Figure 1 It is a collaborative voltage control architecture for medium and low voltage distribution networks;

[0077] Figure 2 Flowchart of a data-model-driven collaborative voltage control method for medium and low voltage distribution networks. Detailed Implementation

[0078] Figure 1 The text mentions photovoltaic inverters and energy storage in low-voltage distribution areas. Since there are multiple related devices, each device is considered an intelligent agent, and multiple devices constitute a multi-agent system. A multi-agent system refers to the controlled devices.

[0079] The flowchart of this invention patent is as follows Figure 2 As shown, to clearly illustrate the implementation and advantages of the technical method of this invention, the implementation process is described in detail. The implementation steps are as follows:

[0080] Step 1: Physical Model of OLTC and SVC in Medium Voltage Distribution Network

[0081] 1) OLTC Model: This model employs a discrete on-load tap changer to control the secondary voltage of the transformer, ensuring that the secondary voltage remains within acceptable limits during operation. The tap changer adjustment process is as follows:

[0082] (1)

[0083] (2)

[0084] (3)

[0085] (4)

[0086] In the formula, The secondary voltage value of the inverter Compared with reference value The difference; This refers to a discrete moment during OLTC runtime; For counters; This is a constant determined by the characteristics and voltage drop of the OLTC, when the OLTC changes from time... Start running, if the time is greater than or equal to When the voltage is high, the counter will increment by one; this is a voltage dead zone to prevent the OLTC from operating unnecessaryly within the allowable voltage range. This refers to the transformer tap position; The number of taps changed for the OLTC tap; This is the delay time for actions during OLTC runtime.

[0087] 2) SVC Model: The SVC model adopts the Thyristor Controlled Reactor (TCR) model. The SVC is connected to the distribution network through an inverter. The equivalent transfer function of the inverter's control loop in reactive power control mode is as follows:

[0088] (5)

[0089] In the formula, This is the difference between the current equivalent susceptance of the SVC inverter and the equivalent susceptance at the previous moment. To control the time constant of the loop; This is the difference between the current control variable and the control variable at the previous moment for the SVC inverter. This refers to the output voltage of the SVC inverter. This is the difference between the current reactive power output of the SVC inverter and the reactive power output at the previous moment.

[0090] Step 2: Objective function, constraints, and solution method for medium-voltage distribution network

[0091] 1) Objective function: The overall control objective of the medium-voltage distribution network is the distribution network line loss. Minimum and node voltage deviation rates Minimum.

[0092] (6)

[0093] In the formula: The number of branches; and They are respectively time Active and reactive power of branch circuits; for Branch resistance; for time The voltage at the beginning of the branch.

[0094] (7)

[0095] In the formula, Represents any node in a medium-voltage distribution network; Indicates the total number of system nodes; Indicates node voltage; Indicates the nominal voltage; These represent the maximum and minimum values ​​of the node voltage, respectively.

[0096] Therefore, weights are assigned to the two objective functions mentioned above, and a comprehensive objective function is formed. for:

[0097] (8)

[0098] In the formula: and This is a weighting coefficient, which can be selected according to the optimization needs of the medium-voltage distribution network.

[0099] 2) Constraints

[0100] a) Constraints of AC power flow equations

[0101] (9)

[0102] In the formula: and These are nodes Injected active and reactive power; and These are nodes The active and reactive power of the load; For nodes and nodes Phase difference of the voltage vector; and They are nodes and nodes The conductance and susceptance of the lines between them; It represents the total number of nodes in the power grid.

[0103] b) Operational safety constraints

[0104] (10)

[0105] In the formula: Representing the power grid nodes The voltage upper limit, real-time value, and lower limit value.

[0106] c) SVC output constraint

[0107] (11)

[0108] (12)

[0109] In the formula, and They are nodes SVC output reactive power and output reactive power ramping; and They are nodes The upper and lower limits of reactive power output from the SVC; and They are respectively Time Node The upper limit of reactive power ramping is set for the SVC output.

[0110] d) OLTC tap changer adjustment constraint

[0111] (13)

[0112] In the formula, This refers to the tap position of the voltage regulator; and These are the upper and lower limits of the tap position on the voltage regulator.

[0113] 3) Solve the medium-voltage distribution network optimization model using a genetic algorithm.

[0114] Input system data, including grid parameters, OLTC parameters, load parameters, energy storage parameters, and SVC parameters, and encode the number of OLTC taps and the reactive power capacity of SVC.

[0115]  Based on the set genetic algorithm parameters, an initial population is generated. The comprehensive objective function of the medium-voltage distribution network is calculated based on the OLTC gear position and SVC output power of each individual, and the fitness is calculated on this basis.

[0116] Using the elite preservation strategy, select, perform segmented crossover and segmented mutation operations on the corresponding population, and select individuals that meet the constraints to enter the next generation of the population.

[0117]  Calculate the fitness function value and save the optimal individual.

[0118]  Determine whether to end the loop based on the set termination condition. If yes, output the result; otherwise, go to .

[0119] Step 3: Deep Deterministic Strategy Gradient Algorithm (DDPG) for Low-Voltage Distribution Networks

[0120] a) State space

[0121] This patent employs the DDPG algorithm primarily for voltage control within a defined low-voltage distribution network area where voltage exceedance nodes are located. Real-time data collection of power output from each node in this area is necessary to monitor and adjust the power output of the voltage regulating equipment. Therefore, the state space of the DDPG algorithm is designed as follows:

[0122] (14)

[0123] In the formula, For nodes per-unit voltage value For nodes The active power; For nodes reactive power; , This represents the number of busbar nodes within the distribution network area.

[0124] b) Action space

[0125] The active and reactive power output and energy storage output of the distributed photovoltaic system within the control area where the node is located are adjusted, thereby achieving control of the low voltage.

[0126] Therefore, the action space of the DDPG algorithm is designed as follows:

[0127] (15)

[0128] When only photovoltaic reactive power regulation is used When reactive power regulation fails to achieve the desired voltage control effect, photovoltaic active power reduction is applied. .at the same time, .in, and They are nodes Changes in active and reactive power output of distributed photovoltaic systems; For nodes Changes in the active power output of the energy storage system.

[0129] c) Reward function

[0130] The DDPG algorithm controls node voltage by adjusting the distributed photovoltaic (PV) and energy storage connections at each node. The control objective is to stabilize the voltage within a safe threshold. Therefore, the instantaneous reward function is set as the sum of the quadratic forms of the voltage exceeding the limit at each node and the quadratic forms of the active and reactive power regulation of the distributed PV output and the active power regulation of the energy storage output, i.e.:

[0131] (16)

[0132] In the formula, bus node Voltage exceeding the limit; To control the number of busbar nodes within the area; , These represent the number of distributed photovoltaic (PV) and energy storage units within the control area, respectively. (Matrix) , , All are weight matrices. For selection coefficients, when photovoltaic active power reduction is not adopted. When photovoltaic active power reduction is added .

[0133] Step 4: Constraints of the DDPG algorithm for low-voltage distribution networks

[0134] In addition to the AC power flow equation constraints and operational safety constraints in step 2, constraints also need to be applied to photovoltaic power output, energy storage power output, and state of charge (SOC), namely:

[0135] 1) Constraints of photovoltaic inverters

[0136] Reactive power output of photovoltaic inverter It is related to its apparent power and rated active power, therefore its constraints are as follows:

[0137] (17)

[0138] In the formula: For photovoltaic Apparent power; For photovoltaic The active power;

[0139] 2) Energy storage constraints

[0140] Active power regulation of energy storage needs to consider not only the device capacity but also the state of charge (SOC) of the energy storage. Therefore, its constraints are as follows:

[0141] (18)

[0142] In the formula: for Real-time energy storage SOC: for Real-time energy storage SOC: for Changes in SOC (State of Charge) of stored energy at any given time; for Real-time energy storage active power: For energy storage device capacity; For time intervals; For energy storage SOC; and Energy storage SOC lower and upper limits; For energy storage Output active power; and Energy storage The minimum and maximum values ​​of the output active power.

[0143] Step 5: Coordinated Voltage Control Strategy for Medium and Low Voltage Distribution Networks

[0144] 1) Obtain the load parameters of each node in the medium-voltage distribution network, the parameters of the medium-voltage distribution network control objects (OLTC on-load tap changer and SVC), and establish a medium-voltage distribution network control model;

[0145] 2) Use a genetic algorithm to solve for the optimal voltage and power data of each node in the medium-voltage distribution network;

[0146] 3) The medium-voltage solution data is sent down to the low-voltage distribution network through the distribution cloud master station. Based on the parameters of the multi-agent (photovoltaic and energy storage) in the low-voltage distribution area and the medium-voltage solution data, the DDPG voltage control algorithm of the low-voltage distribution network is started.

[0147] 4) After training with a large amount of data, find the optimal output of photovoltaic and energy storage to ensure that the voltage is maintained within a safe range; the data includes: 1. Real-time data such as voltage, current, and power collected by monitoring equipment; 2. Historical electrical data of each node and device;

[0148] 5) Upload the voltage and power data of each node in the low-voltage distribution area, as well as the optimal photovoltaic and energy storage power, to the distribution cloud master station and correct the optimization results of the medium-voltage distribution network. Finally, ensure that the voltage and line loss of the medium and low-voltage distribution areas meet the requirements.

Claims

1. A data-model-driven low-voltage coordinated voltage control method for distribution networks, characterized in that: Includes the following steps: 1) Obtain the load parameters of each node in the medium-voltage distribution network and the parameters of the medium-voltage distribution network control objects, and establish a medium-voltage distribution network control model; 2) Given the objective function and constraints of the medium-voltage distribution network, use a genetic algorithm to solve for the optimal parameters of each node in the medium-voltage distribution network; 3) The optimized parameters obtained from the medium-voltage solution are sent to the low-voltage distribution network through the distribution cloud master station. Based on the parameters of the controlled equipment in the low-voltage distribution area and the optimized parameters obtained from the medium-voltage solution, the DDPG voltage control algorithm of the low-voltage distribution network is started. 4) After training with a large amount of data, find the optimal output of photovoltaic and energy storage to ensure that the voltage is maintained within a safe range; 5) Upload the voltage and power data of each node in the low-voltage distribution area, as well as the optimal photovoltaic and energy storage power, to the distribution cloud master station and correct the optimization parameters of the medium-voltage solution. Finally, ensure that the voltage and line loss of the medium and low-voltage distribution areas meet the requirements.

2. The low-voltage coordinated voltage control system for distribution networks based on data-model driven technology as described in claim 1 The manufacturing method is characterized by: Step 1) The medium-voltage distribution network control model includes the physical models of OLTC and SVC. OLTC model: This model uses a discrete on-load tap changer to control the secondary voltage of the transformer and maintain it within the allowable range during operation. The tap adjustment process is as follows: (1)； (2)； (3)； (4)； In the formula, The secondary voltage value of the inverter Compared with reference value The difference; This refers to a discrete moment during OLTC runtime; For counters; This is a constant determined by the characteristics and voltage drop of the OLTC, when the OLTC changes from time... Start running, if the time is greater than or equal to When the counter reaches its maximum value, it will increment by one. This is a voltage dead zone to prevent the OLTC from operating unnecessaryly within the allowable voltage range; This refers to the transformer tap position; The number of taps changed for the OLTC tap; This refers to the delay time for actions during OLTC runtime. SVC Model: The SVC model adopts the thyristor phase-controlled reactor model. The SVC is connected to the distribution network through an inverter. The equivalent transfer function of the control loop of the inverter in reactive power control mode is as follows: (5)； In the formula, This is the difference between the current equivalent susceptance of the SVC inverter and the equivalent susceptance at the previous moment. To control the time constant of the loop; This is the difference between the current control variable and the control variable at the previous moment for the SVC inverter. This refers to the output voltage of the SVC inverter. This is the difference between the current reactive power output of the SVC inverter and the reactive power output at the previous moment.

3. The low-voltage coordinated voltage control method for distribution networks based on data-model driven technology according to claim 1, characterized in that: Step 2) 2.1) Objective Function: The overall control objective of the medium-voltage distribution network is the distribution network line loss. Minimum and node voltage deviation rates Minimum; （6） ； In the formula: The number of branches; and They are respectively time Active and reactive power of branch circuits; for Branch resistance; for time The voltage at the beginning of the branch; （7） ； In the formula, Represents any node in a medium-voltage distribution network; Indicates the total number of system nodes; Indicates node voltage; Indicates the nominal voltage; These represent the maximum and minimum values ​​of the node voltage, respectively. The two objective functions are weighted separately, and a combined objective function is formed. for: （8） ； In the formula: and These are weighting coefficients, selected based on the optimization requirements of the medium-voltage distribution network; 2.2) Constraints a) Constraints of AC power flow equations （9） ； In the formula: and These are nodes Injected active and reactive power; and These are nodes The active and reactive power of the load; For nodes and nodes Phase difference of the voltage vector; and They are nodes and nodes The conductance and susceptance of the lines between them; This is the total number of nodes in the power grid; b) Operational safety constraints （10） ； In the formula: Representing power grid nodes The upper, real-time, and lower limits of the voltage; c) SVC output constraint （11） ； （12） ； In the formula, and They are nodes SVC output reactive power and output reactive power ramping; and They are nodes The upper and lower limits of reactive power output from the SVC; and They are respectively Time Node The upper and lower limits of reactive power ramping output of the upper SVC; d) OLTC tap changer adjustment constraint （13） ； In the formula, This refers to the tap position of the voltage regulator; and These are the upper and lower limits of the tap position of the voltage regulator.

4. The low-voltage coordinated voltage control method for distribution networks based on data-model driven technology according to claim 2, characterized in that: Step 2) 2.3) Solving the medium-voltage distribution network optimization model using a genetic algorithm ① Input system data, including grid parameters, OLTC parameters, load parameters, energy storage parameters, and SVC parameters, and encode the number of OLTC taps and the reactive power capacity of SVC; ② Based on the set genetic algorithm parameters, an initial population is generated. The comprehensive objective function of the medium-voltage distribution network is calculated according to the OLTC gear position and SVC output power of each individual, and the fitness is calculated on this basis. ③ Utilize the elite preservation strategy to perform selection, segmented crossover, and segmented mutation operations on the corresponding population, selecting individuals that meet the constraints to enter the next generation of the population; ④ Calculate the fitness function value and save the optimal individual; ⑤ Based on the set termination condition, determine whether to end the loop. If yes, output the result; otherwise, go to step .

5. The low-voltage coordinated voltage control method for distribution networks based on data-model driven technology according to claim 1, characterized in that: Step 3): The DDPG voltage control algorithm for low-voltage distribution networks includes... a) State space The DDPG voltage control algorithm is used to control voltage within a defined low-voltage distribution network area where voltage exceedance nodes are located. Real-time data collection of power output from each node in this low-voltage distribution network area is used to monitor the power output adjustment of voltage regulating equipment. The state space of the DDPG voltage control algorithm is designed as follows: （14） ； In the formula, For nodes per-unit voltage value For nodes The active power; For nodes reactive power; , This represents the number of busbar nodes within the distribution network area. b) Action space The active and reactive power output and energy storage output of distributed photovoltaic systems within the low-voltage distribution network area where the voltage cross-node is located are adjusted to achieve low-voltage voltage control. The action space of the DDPG voltage control algorithm is designed as follows: （15） ； When only photovoltaic reactive power regulation is used When photovoltaic reactive power regulation fails to achieve the desired voltage control effect, photovoltaic active power reduction is applied. ,at the same time, ,in, and They are nodes Changes in active and reactive power output of distributed photovoltaic systems; For nodes Changes in active power output of the upper energy storage; c) Reward function The DDPG voltage control algorithm controls node voltage by adjusting the distributed photovoltaic and energy storage connected to each node. The control objective is to stabilize the voltage within a safe threshold. The instantaneous reward function is set as the sum of the quadratic form of the voltage exceeding the limit for each node and the quadratic form of the active and reactive power regulation of the distributed photovoltaic output and the active power regulation of the energy storage output, i.e.: （16） ； In the formula, bus node Voltage exceeding the limit; This refers to the number of busbar nodes within a low-voltage distribution network area. , These represent the number of distributed photovoltaic (PV) systems and energy storage systems within the low-voltage distribution network area, respectively, and the matrix... , , Both are weight matrices. For selection coefficients, when photovoltaic active power reduction is not adopted. When photovoltaic active power reduction is added .

6. The low-voltage coordinated voltage control method for distribution networks based on data-model driven technology according to claim 1, characterized in that: The constraints of the low-voltage distribution network DDPG algorithm in step 3) In addition to the AC power flow equation constraints and operational safety constraints in step 2, constraints also need to be applied to photovoltaic power output, energy storage power output, and state of charge (SOC), namely: 1) Constraints of photovoltaic inverters Reactive power output of photovoltaic inverter It is related to its apparent power and rated active power, and its constraints are as follows: （17） ； In the formula: For photovoltaic Apparent power; For photovoltaic The active power; 2) Energy storage constraints Active power regulation of energy storage needs to consider not only the device capacity but also the energy storage state of charge (SOC), and its constraints are as follows: （18） ； In the formula: for Real-time energy storage SOC: for Real-time energy storage SOC: for Changes in SOC (State of Charge) of stored energy at any given time; for Real-time energy storage active power: For energy storage device capacity; For time intervals; For energy storage SOC; and Energy storage SOC lower and upper limits; For energy storage Output active power; and Energy storage The minimum and maximum values ​​of the output active power.

7. The low-voltage coordinated voltage control method for distribution networks based on data-model driven technology according to claim 1, characterized in that: In step 5), the uploaded low-voltage electrical parameters are combined with the medium-voltage electrical parameters and the objective function to recalculate and adjust the previous calculation results.

Images