CN-122026339-A - Distribution network dynamic reconstruction method considering distributed power supply uncertainty

CN122026339ACN 122026339 ACN122026339 ACN 122026339ACN-122026339-A

Abstract

The invention discloses a dynamic reconstruction method of a power distribution network, which comprises the steps of constructing a mathematical model of dynamic reconstruction of the power distribution network, taking a reconstruction problem corresponding to an optimization target with minimum total daily network loss as a learning object, describing uncertainty of DG and load, constructing a double random scene to form a scene set covering an uncertainty space, solving dynamic reconstruction optimization of one certainty for each generated scene to obtain an optimal switch combination under the scene, constructing a data set according to acquired data, training DNN, and carrying out rapid forward propagation on the trained DNN according to real-time measurement to realize rapid prediction and decision of an optimal reconstruction scheme. The invention can rapidly give out the optimal action scheme of the tie switch and the sectionalizing switch on the premise of meeting the operation constraint of voltage, current, radial topology and the like, thereby achieving the aims of reducing network loss, improving voltage distribution and improving the operation economy of the power distribution network.

Inventors

TANG ZHENYU
XU DONGLIANG
MA GANG
Wang Chaomeng
Bian Lijie
Han Rushuai
YU XIUYONG
LV ZHENYU

Assignees

南京师范大学

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (9)

1. A dynamic reconfiguration method of a power distribution network considering uncertainty of a distributed power supply is characterized by comprising the following steps: s1, constructing a mathematical model of dynamic reconstruction of a power distribution network, and taking a reconstruction problem corresponding to an optimization target with minimum daily total network loss as a learning object; s2, describing uncertainty of DG and load, constructing a double random scene, and forming a scene set covering an uncertainty space; s3, solving one-time deterministic dynamic reconstruction optimization for each scene generated in the step S2 according to the constructed mathematical model of dynamic reconstruction of the power distribution network to obtain an optimal switch combination in the scene and using the optimal switch combination as a label of scene data; s4, constructing a data set according to the data acquired in the step S3, and training the constructed DNN based on the data set; S5, carrying out rapid forward propagation on the trained DNN according to real-time measurement, and realizing rapid prediction and decision-making on the optimal reconstruction scheme.
2. The method for dynamically reconstructing a power distribution network according to claim 1, wherein the mathematical model of the dynamic reconstruction of the power distribution network in step S1 comprises the following objective functions: ; Wherein, the T is a time period index, and T is a total time period number; for the length of each period; The calculation formula of the total active power loss of the time system is as follows: ; Wherein, the Is the set of all branches in the system; Is a binary variable representing a branch At the position of A switch state at a moment; Is a branch Is a conductive material; And Respectively are Time node Sum node Is set to the voltage amplitude of (1); Is a node Sum node Is provided.
3. The dynamic reconfiguration method of the power distribution network according to claim 2, wherein the mathematical model of the dynamic reconfiguration of the power distribution network in step S1 includes constraint conditions for considering stable operation of the power distribution network, including power flow constraint, operation safety constraint, topology constraint and DG output constraint, and is specifically expressed as follows: The power flow constraint is that for any node i in the power distribution network, at any time t, the injection power of the node i must be equal to the outflow power, namely, a power balance equation is satisfied: ; Wherein, the And Respectively nodes At the position of Active and reactive power injection at time; And Active and reactive loads, respectively; is connected with the node A set of connected nodes; And Corresponding nodes in the node admittance matrix Sum node Is a conductivity and susceptance of (a); the operation safety constraint comprises node voltage constraint and branch current constraint, and specifically comprises the following steps: node voltage constraint, namely the voltage amplitude of all load nodes must be within an allowable range; ; branch current constraint, namely the current of all closed branches cannot exceed the thermal stability limit, namely the current carrying capacity; ; Wherein, the And The lower and upper limits of the node voltage, respectively; Is a branch Maximum allowable current of (2); Is a set of all nodes; Topological structure constraint that the reconstructed power distribution network must keep a radial structure, and all load nodes cannot be isolated; DG force constraint: ; Wherein, the Is a collection of all nodes that have DG installed; Is that Time node At DG, the maximum available active force.
4. A method for dynamically reconstructing a power distribution network in consideration of uncertainty of distributed power supply according to claim 3, wherein said step S2 uses latin hypercube sampling to generate independent 24-hour operation scenarios, each generated scenario comprising a maximum available power sequence DG for 24 hours And a load sequence By generating N s such scenes, a set of scenes that can represent the annual operating characteristics of the system is obtained.
5. The method for dynamically reconstructing a power distribution network according to claim 4, wherein said maximum available power sequence of DG in step S2 Reference to the theoretical maximum output of the photovoltaic unit at time t The calculation model of (a) is obtained as follows: ; Wherein, the Rated output power under standard test conditions; Is the illumination intensity under standard test conditions; Is the battery temperature under standard test conditions; is the actual illumination intensity; Is the power temperature coefficient; is the actual operating temperature of the photovoltaic cell.
6. The method for dynamically reconstructing a power distribution network according to claim 5, wherein in step S3, the improved genetic algorithm IGA is adopted as an offline solver for solving, and the solving process comprises: A1, adopting natural number coding, wherein a chromosome consists of states of all tie switches, and ensuring that the switch operation meets the constraint of a radial network topology structure; a2, adopting an objective function As a fitness function, i.e. minimizing the total daily energy consumption; a3, adopting a roulette selection method and combining an elite retention strategy, and designing special crossing and mutation operators for ensuring topological constraint so as to improve the searching efficiency.
7. The method for dynamically reconstructing a power distribution network according to claim 6, wherein in step S3, an offline solver is operated one by one on N s scenes to obtain an original dataset containing N s samples, and the final constructed dataset N s sample pairs are included, and the structure is as follows: ; wherein the features are input Represents the first The system state of each scene comprises a 24-hour load curve of all nodes in the scene and a 24-hour maximum output curve of DG; ; Output label Is corresponding to a scene Is calculated by an offline solver, is a binary vector, and represents the optimal states of all switches; 。
8. The method according to claim 7, wherein the DNN in the step S4 is composed of an input layer, a plurality of hidden layers and an output layer, the data is propagated unidirectionally from the input layer to the output layer, and for the first layer in the network, the data is output Recursively calculated by the following formula: ; Wherein, the Is the output of layer 1, for the first layer, Namely, the input characteristics of the network; And The weight matrix and the bias vector of the first layer are respectively; Is a linear weighted input of the first layer; is an activation function of the first layer and is responsible for introducing nonlinear transformation; The output layer comprises M neurons, each neuron corresponds to one switch, and the aim of the output layer is to predict the states of all M switches in the network.
9. The method for dynamically reconstructing a power distribution network according to claim 8, wherein the output layer in DNN of step S4 adopts a Sigmoid activation function: ; Wherein, the The probability of the ith switch closing, which is a model prediction; is a linear result combination of the DNN output layers.

Description

Distribution network dynamic reconstruction method considering distributed power supply uncertainty Technical Field The invention belongs to the field of power grids, relates to a power distribution network operation optimization and network reconstruction control technology containing a distributed power supply, and particularly relates to a power distribution network dynamic reconstruction method considering uncertainty of the distributed power supply. Background With the rapid development of new energy sources such as photovoltaic, wind power and the like, the permeability of a distributed power supply on the side of a power distribution network is continuously improved, and the power distribution network is gradually changed from a passive network with traditional unidirectional power supply to an active power distribution network form comprising multi-source access and flexible regulation. Under the background of high-proportion access of new energy, the output of the distributed power supply has obvious randomness and fluctuation, and meanwhile, the load time-varying characteristics are superposed, so that the problems of frequent change of power flow distribution, increase of voltage out-of-limit risks, increase of network loss and the like of the power distribution network are more likely to occur on different time scales, and higher requirements are put forward on the safety, economy and instantaneity of operation scheduling of the power distribution network. Although the traditional methods based on genetic algorithm, particle swarm and the like can obtain a better solution, on-line iterative solution consumes a long time under the conditions of high-proportion distributed power supply access and frequent change of working conditions, and the requirement of dynamic reconstruction on instantaneity is difficult to meet. Disclosure of Invention The invention aims to provide a dynamic reconstruction method of the power distribution network, which takes the uncertainty of the distributed power supply into account, in order to reduce the uncertainty influence such as output fluctuation, load change and the like caused by high-proportion access of the distributed power supply to the power distribution network, and can rapidly give out an optimal action scheme of a tie switch and a sectionalizing switch on the premise of meeting the operation constraint such as voltage, current, radial topology and the like, thereby achieving the aims of reducing network loss, improving voltage distribution and improving the operation economy of the power distribution network. The invention provides a dynamic reconstruction method of a power distribution network considering uncertainty of a distributed power supply, which comprises the following steps of: s1, constructing a mathematical model of dynamic reconstruction of a power distribution network, and taking a reconstruction problem corresponding to an optimization target with minimum daily total network loss as a learning object; s2, describing uncertainty of DG and load, constructing a double random scene, and forming a scene set covering an uncertainty space; s3, solving one-time deterministic dynamic reconstruction optimization for each scene generated in the step S2 according to the constructed mathematical model of dynamic reconstruction of the power distribution network to obtain an optimal switch combination in the scene and using the optimal switch combination as a label of scene data; s4, constructing a data set according to the data acquired in the step S3, and training the constructed DNN based on the data set; S5, carrying out rapid forward propagation on the trained DNN according to real-time measurement, and realizing rapid prediction and decision-making on the optimal reconstruction scheme. Further, the mathematical model of the dynamic reconstruction of the power distribution network in the step S1 includes the following objective functions: Wherein, the T is a time period index, and T is a total time period number; for the length of each period; The calculation formula of the total active power loss of the time system is as follows: Wherein, the Is the set of all branches in the system; Is a binary variable representing a branch At the position ofA switch state at a moment; Is a branch Is a conductive material; And Respectively areTime nodeSum nodeIs set to the voltage amplitude of (1); Is a node Sum nodeIs provided. Further, the mathematical model of the dynamic reconfiguration of the power distribution network in the step S1 includes constraint conditions considering stable operation of the power distribution network, including power flow constraint, operation safety constraint, topology structure constraint and DG output constraint, and is specifically expressed as follows: The power flow constraint is that for any node i in the power distribution network, at any time t, the injection power of the node i must be equal to the outflow power, namel