CN-121791344-B - Power distribution network voltage control method based on reactive virtual power plant

CN121791344BCN 121791344 BCN121791344 BCN 121791344BCN-121791344-B

Abstract

The invention relates to a power distribution network voltage control method based on a reactive virtual power plant, which comprises the steps of constructing a double-layer collaborative regulation model based on discrete and continuous reactive power regulation equipment in a power distribution network under the management of the reactive virtual power plant, establishing physical constraint of the power distribution network, converting optimal control of the discrete reactive power regulation equipment into first multi-agent reinforcement learning of a discrete action space strategy for solving under the condition of the physical constraint of the power distribution network, generating a daily scheduling plan and issuing the daily scheduling plan to a daily correction layer, converting control of the continuous reactive power regulation equipment into second multi-agent reinforcement learning of the continuous action space strategy for solving under the condition that the physical constraint of the power distribution network and the daily scheduling plan are taken as boundary constraint, generating a daily correction instruction, regulating the continuous reactive power regulation equipment based on the daily correction instruction to obtain node voltage in the power distribution network, and feeding the node voltage in the power distribution network back to the daily scheduling layer for closed loop collaborative regulation, so as to realize voltage optimal regulation.

Inventors

XIE BANGPENG
HAN DONG
YUAN YIMING
CHEN BIN
SHEN HAO
ZHAO ZHENYU
WANG MINHUA
GU LI
WANG JIAYU
SUN LEI

Assignees

国网上海市电力公司

Dates

Publication Date: 20260508
Application Date: 20260305

Claims (7)

1. The power distribution network voltage control method based on the reactive virtual power plant is characterized by comprising the following steps of: s1, constructing a double-layer cooperative regulation model comprising a day-ahead dispatching layer and an intra-day correction layer based on discrete-type reactive power regulation equipment and continuous-type reactive power regulation equipment in a power distribution network under reactive power virtual power plant management, and constructing physical constraint of the power distribution network; s2, under the physical constraint condition of the power distribution network, the day-ahead dispatching layer converts the optimal control of the discrete reactive power regulation equipment into first multi-agent reinforcement learning of a discrete action space strategy for solving, and generates a day-ahead dispatching plan and sends the day-ahead dispatching plan to the day-ahead correction layer; The step S2 includes: based on the physical constraint of the power distribution network, constructing a daily scheduling optimization objective function with the aim of minimizing the total node voltage deviation in the power distribution network in a scheduling period, and constructing the maximum allowable action frequency constraint of each discrete reactive power regulating device in the scheduling period; The daily scheduling optimization is modeled as a Markov game process, and a corresponding discrete intelligent agent is defined for each discrete reactive power regulation device to obtain a first multi-intelligent agent, wherein the action space of each intelligent agent is discrete gear or switching operation, the state space comprises the current state of the device and the accumulated action times, and the reward function comprises the comprehensive evaluation of node voltage deviation, device action cost and voltage out-of-limit conditions in the power distribution network; Reinforcement learning is carried out on the first multi-agent by utilizing MASAC-discete algorithm, and an agent cooperation strategy of the Discrete reactive power regulation equipment is obtained ; Generating a day-ahead dispatching plan containing gears or switching states of future dispatching cycles of each discrete reactive power regulating device based on an agent cooperative strategy of the discrete reactive power regulating device; the daily schedule optimization objective function is as follows: Wherein, the For the period index of time, Is the number of time periods; For the index of the reactive virtual power plant, The total number of the reactive virtual power plants is; Indexing the nodes in the distribution network, Is a reactive virtual power plant A set of nodes in the administered power distribution network; To at the same time Nodes in a power distribution network with time periods belonging to reactive virtual power plants Voltage per unit value of (2); S3, under the dual condition that physical constraint of the power distribution network and a day-ahead dispatching plan are used as boundary constraint, the day-ahead correction layer converts optimal control of the continuous reactive power regulation equipment into second multi-agent reinforcement learning of a continuous action space strategy to solve the problem, and generates a day-ahead correction instruction; The step S3 includes: Based on the physical constraint of the power distribution network, the day-ahead scheduling plan as a boundary condition and an uncertainty scene set, constructing an intra-day correction optimization objective function with the aim of minimizing expected voltage deviation; Modeling the daily correction optimization as a Markov game process, and defining a corresponding continuous intelligent agent for each continuous reactive power regulating device to obtain a second intelligent agent, wherein the action space of each intelligent agent is continuous reactive power output, the state space comprises local active or reactive load and photovoltaic active output, and the reward function comprises node voltage deviation in the power distribution network and punishment against corresponding safe operation constraint on the action of the continuous reactive power regulating device; reinforcement learning is carried out on the second agent by utilizing MASAC algorithm, and the cooperative control strategy of the agent of the continuous reactive power regulation equipment is obtained ; Generating an intra-day correction instruction of each continuous reactive power regulating device based on a cooperative control strategy of an intelligent agent of the continuous reactive power regulating device; Adjusting the continuous reactive power adjusting equipment based on the daily adjustment instruction to obtain node voltage of the power distribution network under the management of the reactive virtual power plant; Comparing the node voltage of the power distribution network with the node voltage of the power distribution network corresponding to the daily scheduling plan to obtain node voltage deviation information, and re-carrying out first multi-agent reinforcement learning solution on the basis of the node voltage deviation information by the daily scheduling layer to re-generate the daily scheduling plan; wherein one distribution network is managed by at least one reactive virtual power plant.
2. The reactive virtual power plant-based power distribution network voltage control method according to claim 1, wherein the discrete reactive power regulation device comprises at least one of an on-load voltage regulating transformer OLTC, a capacitor bank CB, a voltage regulator VR; the continuous reactive power conditioning apparatus includes at least one of a photovoltaic inverter and an energy storage inverter.
3. The reactive virtual power plant-based power distribution network voltage control method according to claim 1, wherein a power distribution network physical constraint is established based on DistFlow tide equations, and the power distribution network physical constraint comprises a node power balance constraint and a voltage balance constraint.
4. The reactive virtual power plant-based power distribution network voltage control method according to claim 1, wherein under the physical constraint of the power distribution network, the maximum allowable action frequency constraint of each discrete reactive power regulation device of a day-ahead dispatching layer in the dispatching period is as follows: Wherein, the 、 And 0, 1 Variable, delta OLTC (t) is the time period An indication variable of whether on-load tap operation of the step-down transformer occurs, M OLTC is the maximum number of operations allowed by the OLTC in the programming cycle, delta CB (t) is the on-time period Whether an indication variable of the CB switching action of the capacitor bank occurs or not, wherein M CB is the maximum action number allowed by the CB in the planning period; To be in a period of If an indication of voltage regulator VR range adjustment actions has occurred, M VR is the maximum number of actions allowed by VR during the programming period.
5. The reactive virtual power plant-based power distribution network voltage control method according to claim 4, wherein the intelligent agent cooperation strategy of the discrete reactive power regulation device Comprises an OLTC gear adjustment quantity and a CB switching state change quantity And VR shift adjustment The following are provided: The local observation state space of the discrete agent comprises the following steps of Tap position of on-load tap changer OLTC Switching state of capacitor bank CB Gear position of voltage regulator VR Cut off to Cumulative number of actions of period OLTC Cumulative number of actions of CB And the number of cumulative actions of VR The following are provided: Reward function The following are provided: Wherein, the The voltage deviation square sum of the whole power distribution network is represented and used for measuring the voltage quality; Representation of The motion vector of the time period comprises an OLTC gear adjustment amount, a CB switching state and a VR gear adjustment amount, and two norms Characterizing an operational cost of the device; The voltage out-of-limit indication function is adopted, when the voltage exceeds the allowable range, the value is 1, and otherwise, the value is 0; 、 Respectively is And A weight coefficient; the day-ahead schedule is expressed as follows: wherein T OLTC (t)、S CB (T) and L VR (T) are respectively at the present The period is determined by the discrete layer before date by actual values of OLTC tap shift, CB switching state, and VR shift adjustment.
6. The reactive virtual power plant-based power distribution network voltage control method according to claim 1, wherein the optimization objective function based on the Ω day correction layer is based on a scene set Ω including source, network, load, storage uncertainty factors as follows: wherein ω is an index of the scene of the uncertainty; representing expected values for all possible scenarios ω in the scenario set Ω for the uncertainty factor; The uncertain factors in each scene comprise source side distributed power supply output fluctuation, network side line parameter and topology disturbance, load random change on the load side and energy storage side SOC deviation and efficiency change; The power constraint of the continuous reactive power regulation equipment of the intra-day correction layer comprises photovoltaic inverter power constraint and energy storage inverter power constraint; the photovoltaic inverter power constraint is as follows: Wherein, the And Active power output and reactive power output of the photovoltaic inverter in a t period are respectively; the upper limit of active power output of the photovoltaic system in a t-period maximum power point tracking mode is set; rated apparent power for the photovoltaic inverter; The lowest power factor allowed for the photovoltaic access point; the energy storage inverter power constraint is as follows: Wherein, the And Active power and reactive power output of the energy storage inverter in the period t are respectively; And The maximum discharge power and the maximum charge power of the energy storage inverter are respectively; is the rated apparent power of the energy storage inverter.
7. The reactive virtual power plant-based power distribution network voltage control method according to claim 6, characterized by a coordinated control strategy of the agents of the continuous reactive power regulating device Including reactive power output of photovoltaic inverter And reactive power output of energy storage inverter The following are provided: Wherein, the A local observation state space for the continuous intelligent body; active load including t period Reactive load And photovoltaic active output The following are provided: Each continuous agent Is a reward function of (2) The following are provided: Wherein, the In order to provide a continuous quantity of agent, Is a continuous intelligent body A set of nodes in a distribution network, Is a node At the position of The voltage amplitude per unit value of the time period, Is a continuous intelligent body Action violation penalty coefficients of (a).

Description

Power distribution network voltage control method based on reactive virtual power plant Technical Field The invention relates to the field of intelligent power grid and power distribution network regulation and control, in particular to a power distribution network voltage control method based on a reactive virtual power plant. Background With the continuous pushing of 'double carbon', the construction of a novel power system with cleanness, low carbon, safety and high efficiency has become an important direction of energy transformation. Under the background, the virtual power plant is taken as a key technical path for comprehensively planning distributed resources and improving the system regulation capability, and increasingly shows the important roles of improving the operation efficiency of an energy system, promoting new energy consumption and promoting energy conservation and emission reduction. The reactive virtual power plant is a special form of the virtual power plant, focuses on coordination and management of reactive power, can effectively integrate multiple types of distributed reactive power resources, and constructs a reactive power regulation system with 'source-network-load-storage' cooperation, so that the voltage stability of a power distribution network is improved, the flexibility of system operation is enhanced, and the green low-carbon development of a power system is promoted. At present, research on reactive virtual power plants at home and abroad is still in a starting stage, and particularly under the condition of high-proportion renewable energy access, the optimal regulation mechanism is still immature. The existing method mostly adopts a centralized control architecture, but can play a role in a small-scale system, and is faced with the actual demands of distributed resource proliferation and increasingly complex power grid structure, and has obvious limitations in aspects of expandability, dynamic adaptability and data privacy protection. Specifically, the current reactive power regulation strategy still lacks a dynamic matching mechanism between heterogeneous resources, and is difficult to realize cooperative response of equipment and time scales under uncertain disturbance, so that the effect of actually regulating and controlling node voltage in a power distribution network of a reactive virtual power plant in a large-scale application is limited. Disclosure of Invention In view of the above analysis, the embodiment of the invention aims to provide a power distribution network voltage control method based on a reactive virtual power plant, which is used for solving the technical problem that node voltage is difficult to control stably due to lack of a cooperative response mechanism across equipment and time scales in the conventional reactive virtual power plant control method in a complex power distribution network accessed by a high-proportion distributed power supply. The invention provides a power distribution network voltage control method based on a reactive virtual power plant, which comprises the following steps: s1, constructing a double-layer cooperative regulation model comprising a day-ahead dispatching layer and an intra-day correction layer based on discrete-type reactive power regulation equipment and continuous-type reactive power regulation equipment in a power distribution network under reactive power virtual power plant management, and constructing physical constraint of the power distribution network; s2, under the physical constraint condition of the power distribution network, the day-ahead dispatching layer converts the optimal control of the discrete reactive power regulation equipment into first multi-agent reinforcement learning of a discrete action space strategy for solving, and generates a day-ahead dispatching plan and sends the day-ahead dispatching plan to the day-ahead correction layer; S3, under the double conditions that the physical constraint of the power distribution network and the day-ahead dispatching plan are used as boundary constraint, the day-ahead correction layer converts the optimal control of the continuous reactive power regulation equipment into the second multi-agent reinforcement learning of the continuous action space strategy to solve the problem, and generates a day-ahead correction instruction; wherein one distribution network is managed by at least one reactive virtual power plant. Further, the discrete reactive power regulation equipment comprises at least one of an on-load regulating transformer OLTC, a capacitor bank CB and a voltage regulator VR; the continuous reactive power conditioning apparatus includes at least one of a photovoltaic inverter and an energy storage inverter. Further, physical constraints of the power distribution network are established based on DistFlow tide equations, wherein the physical constraints of the power distribution network comprise node power balance constraints and voltage balan