CN115495986A - Self-healing distributed network reconstruction method for power distribution network - Google Patents

Self-healing distributed network reconstruction method for power distribution network Download PDF

Info

Publication number
CN115495986A
CN115495986A CN202211168721.4A CN202211168721A CN115495986A CN 115495986 A CN115495986 A CN 115495986A CN 202211168721 A CN202211168721 A CN 202211168721A CN 115495986 A CN115495986 A CN 115495986A
Authority
CN
China
Prior art keywords
network
healing
power
node
self
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211168721.4A
Other languages
Chinese (zh)
Inventor
徐谦
俞楚天
胡哲晟
孙轶恺
张利军
王蕾
戴攀
徐威涛
耿磊
杨凯
袁翔
范明霞
朱超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd
Original Assignee
Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd filed Critical Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd
Priority to CN202211168721.4A priority Critical patent/CN115495986A/en
Publication of CN115495986A publication Critical patent/CN115495986A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
    • G06Q10/06375—Prediction of business process outcome or impact based on a proposed change
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/04—Power grid distribution networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Evolutionary Computation (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Quality & Reliability (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention provides a self-healing distributed network reconstruction method for a power distribution network. The scheme is used for acquiring the whole network topology and node operation information through a distributed consensus protocol. Under the provided power distribution network reconstruction method, the power distribution network has a self-healing function, can realize economic operation in a normal operation state, and can realize automatic reconstruction and self-healing in an N-1 and N-2 line fault state without an additional external trigger signal. In order to cope with the fluctuation of the high proportion of renewable energy, the scheme adopts forward-looking rolling optimization, and combines the power generation data generated by the generated countermeasure network and historical data to be used as a prediction basis. Simulation results show that the scheme can automatically realize economic operation optimization and self-healing under normal state, single-point fault and two-point fault.

Description

Self-healing distributed network reconstruction method for power distribution network
Technical Field
The invention relates to the technical field of power distribution network reconstruction, in particular to a self-healing distributed network reconstruction method for a power distribution network.
Background
With the increasing permeability of renewable energy in a power Distribution Network (DNs), planning and operation of the power distribution network face huge pressure, and it is especially important to ensure safe and reliable operation. Power Distribution Network Reconfiguration (DNR) is a key approach to optimizing power flow distribution and improving power supply reliability and economy by changing the state of remotely controllable switches.
Because DNs is complex in structure, various in equipment and susceptible to external factors, DNs is inevitably failed, which usually results in complete or partial power failure of the whole system and large-scale economic loss. DNs has a large number of normally closed section switches and a small number of normally open tie switches. DNR adjusts the topology of the power distribution network primarily by changing the on or off state of certain switches. According to the actual operation condition of the system, the network structure is adjusted by changing the switch state, so that the network loss can be reduced, the load can be balanced, the power supply voltage quality can be improved, the overload can be eliminated, and the economical efficiency of the system can be improved. Therefore, DNR has become one of the important means for optimizing the operation of power distribution systems. It can be divided into two categories: reconfiguration in normal situations and reconfiguration after a failure is called power restoration.
The reconstruction of the fault recovery of the power distribution network is a complex mixed integer nonlinear combination optimization problem. This is an extremely complex real-time decision-making process involving a large number of uncertainties. The existing linear numerical analysis technology is difficult to provide an effective solution, and a dispatcher has to make a decision according to field experience. When the system is in failure, various information is crowded in a short time, and a dispatcher needs to control the situation as much as possible, avoid the expansion of accidents and recover as much load as possible in a short time. At this time, the dispatcher must not only have substantial expertise and rich operating experience, be familiar with various operating modes, operating procedures and safety margin parameters, but also be able to keep cool and awake. In fact, in the face of such complex and urgent situations, it is difficult for the dispatcher to take correct and effective measures according to their limited experience and inaccurate intuition, so as to reduce the accident losses as much as possible.
Optimization of DNR aims to achieve several goals, including reducing network active power loss, balancing load, improving the reliability of DNs, and improving power supply quality. With the aim of reducing the system operation cost, a DNR model is established. The system operating costs include electricity purchase costs, switch operating costs, and distributed electricity operating costs. In fact, reducing network losses may also reduce the cost of purchasing electricity from the upper grid of the distribution grid, thereby reducing system operating costs. In conventional distribution networks without Distributed Generation (DGs), it is believed that there is no substantial difference between reducing network losses and reducing operating costs. The conventional method reconfiguration method considers the switching operation number in the reconfiguration target, and takes the minimum switching operation number in the reconfiguration period as one of the target functions. Still other approaches propose a new DNs reconfiguration strategy that takes into account the N-1 security standard and takes reduced security risk as one of the optimization objectives.
However, most existing literature on reconstruction methods is centralized, which may not be suitable for current microgrid architectures. In recent years, with the proliferation of distributed power generation systems, power distributed power generation systems have changed greatly, for example, the introduction of renewable energy sources (photovoltaic power generation, wind power generation, etc.) has increased the fluctuation of distributed power generation systems, which has brought about a great challenge to distributed power generation systems. DG is connected to DNs in a small scale and distributed manner. Due to environmental protection and easiness in installation, compared with a centralized power supply system, the DN has the advantages of higher power supply efficiency and the like. Meanwhile, the grid-connected mode of the distributed power generation system also obviously improves the traditional energy structure. In fact, distributed DNR is more compliant with DNs, which is in line with future trends.
Disclosure of Invention
The invention aims to provide a self-healing distributed network reconstruction method for a power distribution network, which is based on a distributed structure of distributed network reconstruction and provides a distributed optimization method for processing the problem of distributed network reconstruction so that the distributed network has a self-healing function.
The method comprises the steps of firstly constructing a network reconfiguration planning problem (DNR) mathematical model with complex constraints in the power distribution network, then establishing a power flow constraint condition of the DNR model, designing a communication rule discovered by global information by adopting a consensus algorithm, and finally providing a prospective prediction method based on a generation countermeasure network (GAN), introducing a prospective rolling optimization method, generating more data by utilizing the GAN, and improving the robustness of prediction.
In order to achieve the purpose, the technical scheme of the invention is as follows: a self-healing distributed network reconstruction method for a power distribution network comprises the following steps:
step 1: constructing a network reconfiguration planning problem (DNR) mathematical model with complex constraints in the power distribution network;
step 2: establishing a power flow constraint condition of the DNR model;
and step 3: designing a communication rule for global information discovery by adopting a consensus algorithm;
and 4, step 4: a prospective prediction method based on GAN is provided.
The network reconfiguration planning problem model in the step 1 specifically comprises the following steps:
Figure BDA0003862628330000021
wherein N is the node number of DN system and uses undirected communication topology
Figure BDA0003862628330000022
Representing DGs, V =1,2, …, m, …, N is a set of nodes of the topology,
Figure BDA0003862628330000023
is a set of edges that make up the topology. (i, j) ∈ denotes a communication edge between node i and node j. Node 0 represents an upstream substation and m represents the number of dispatchable micro gas turbines (MTs), photovoltaics (PV) or Wind Turbines (WTs). For schedulable MTs, set is defined as N MT 。P 0 (t) the amount of electricity purchased from the upstream substation at time t, corresponding to a price of C G 。
Figure BDA0003862628330000024
And
Figure BDA0003862628330000025
the scheduled power generated by the ith MT and the transmission loss at time t at nodes i and j, respectively, C MT And C Loss Respectively their unit cost. Load shedding
Figure BDA0003862628330000026
The corresponding cost C of each node i at the time t λ Required power P D (t) and a factor s i (t)∈[0,1]。C s Representing cost at handover, NS calculationTotal number of switching operations.
In addition, the power flow constraint conditions in the step 2 are as follows:
(1) voltage v of node i Limiting and surviving node only current l ij The limits and limit constraints of MTs active and reactive power are:
Figure BDA0003862628330000031
Figure BDA0003862628330000032
Figure BDA0003862628330000033
Figure BDA0003862628330000034
(2) the rotational tree constraints associated with the radial structure and connectivity of DNs are:
Figure BDA0003862628330000035
Figure BDA0003862628330000036
Figure BDA0003862628330000037
wherein alpha is ij (t)∈{0,1},β ij Is an epsilon of {0,1}, (i, j) ∈ epsilon is two Boolean parameters, epsilon * Representing the set of remaining lines after DNR.
(3) The rule version of the dist _ flow function searches the constraint on the predefined network flow; the relationship constraint between any two connected busbars, and the nonlinear equality constraint relaxation model that makes the dist _ Flow function a convex function, are:
Figure BDA0003862628330000038
Figure BDA0003862628330000039
Figure BDA00038626283300000310
wherein
Figure BDA00038626283300000311
Q-V sag control for MTs run control.
(4) The single commodity flow spinning tree constraint is as follows:
Figure BDA00038626283300000312
|F ij (t)|≤α ij (t)M,(i,j)∈ε * (13)
the ramp constraint for MT, and the run control function for WTs and pv are:
Figure BDA0003862628330000041
Figure BDA0003862628330000042
wherein
Figure BDA0003862628330000043
The ith power output of the previous cycle MT.
Other constraints for nodes containing only the load are:
Figure BDA0003862628330000044
Figure BDA0003862628330000045
the number of switching actions is limited to:
Figure BDA0003862628330000046
the above constraint and network reconstruction problem model is a Mixed Integer Second Order Cone Programming (MISOCP) that can be solved with a commercial solver to solve the globally optimal solution of the proposed problem. The variable needed to be obtained in the above problem is P i (t),Q i (t),s i (t) and alpha ij (t)。
Wherein, the recognition discovery algorithm in the step 3 is as follows:
the information discovery process of the node i is noted as follows:
Figure BDA0003862628330000047
wherein x is i Representing states, including all continuous variables in DN, e.g. schedulable generators
Figure BDA0003862628330000048
And all binary variables in DN, e.g. switch state α ij 。x i ＝[P g,1 ,P g,2 ,P g,3 ,Q g,1 ,Q g,2 ,Q g,3 ,P d,1 ,P d,2 ,P d,3 ,Q d,1 ,Q d,2 ,Q d,3 ，a 12 ，a 23 ，a 13 ]. The first 12 consecutive elements represent the node generators output active and reactive power, as well as load active and reactive power. The last three binary elements represent the state of the circuit monitored by the agent. i =1,2, …, N,
Figure BDA0003862628330000049
Figure BDA00038626283300000410
in the initial state of the agenti,
Figure BDA00038626283300000411
and
Figure BDA00038626283300000412
information found at k and k +1 time slots, respectively, for agent i. a is ij For coefficients exchanging information between adjacent agents and j, 0 if agents and j are connected by a distribution line<a ij <1, otherwise a ij ＝0。
The GAN-based prospective prediction method in the step 4 comprises the following steps:
the generation model GAN based on the deep learning architecture is composed of two substructures, a generator G and a discriminator D. The generator G is responsible for synthesizing data and the discriminator D is responsible for classifying whether the input data is real data or synthesized data. Training GAN is a min-max problem:
Figure BDA0003862628330000051
in the formula theta G And theta D Parameters of the mapping G (-) and D (-) respectively denoted G and D. P rd And P gd Representing a real distribution and a gaussian distribution, respectively. Theta D May measure the difference between the composite probability distribution and the true distribution, and theta G The minimum value of (c) may push the composite probability distribution as close to the true distribution as possible.
Compared with the prior art, the invention has the following characteristics: 1) Compared with the existing centralized DNR method, the method does not need a central node to perform centralized processing, can accelerate the processing speed, and enables the system to have the advantage of plug and play. 2) Compared with the N-1 fault mode which can be processed at present, the method provided by the invention can process the N-2 fault condition, so that the DN system has a self-healing function. 3) If predictive data is lacking, it is necessary to generate enough data from history because real-time prediction is difficult. Compared with robust planning and stochastic planning, the invention uses a generative countermeasure network (GAN) to obtain more and more detailed data in the simulation. Furthermore, a look-ahead optimization scheme may establish a long-term plan to achieve better results.
Drawings
FIG. 1 is an IEEE-33 diagram with photovoltaic and wind turbines;
FIG. 2 is a graph of network loss indices for a conventional single-step method and a four-step prediction method;
FIG. 3 is a diagram of a normal operating state;
FIG. 4 is a state diagram of the distribution network under one line fault (9-10);
FIG. 5 is a state diagram of the distribution network under one line fault (20-21);
FIG. 6 is a diagram of the distribution network state under two line faults (18-19 and 10-11);
fig. 7 is a diagram of the distribution network conditions under two line faults (15-16 and 3-4).
Fig. 8 is a schematic structural diagram of a power distribution network self-healing distributed network reconstruction method.
Detailed Description
The invention mainly considers a self-healing distributed network reconstruction method of a power distribution network, and designs a distributed optimization method, which uses the thought of prospective optimization to process the reconstruction problem of the power distribution network with renewable energy penetration, so that the power distribution network can automatically realize economic operation optimization and self-healing under the normal state, N-1 line faults and N-2 line faults.
The invention aims to provide a self-healing distributed network reconstruction method for a power distribution network. The scheme is used for acquiring the whole network topology and node operation information through a distributed consensus protocol. Under the provided power distribution network reconstruction method, the power distribution network has a self-healing function, can realize economic operation in a normal operation state, and can realize automatic reconstruction and self-healing in an N-1 and N-2 line fault state without an additional external trigger signal. In order to cope with the fluctuation of the high proportion of renewable energy, the invention adopts the look-ahead rolling optimization and combines the power generation data generated by the generation countermeasure network and the historical data as the prediction basis.
In order to achieve the purpose, the technical scheme of the invention is as follows: a self-healing distributed network reconstruction method for a power distribution network comprises the following steps:
step 1: constructing a network reconfiguration planning problem (DNR) mathematical model with complex constraints in the power distribution network;
step 2: establishing a power flow constraint condition of the DNR model;
and 3, step 3: designing a communication rule for global information discovery by adopting a consensus algorithm;
and 4, step 4: a prospective prediction method based on GAN is provided;
the network reconfiguration planning problem model in the step 1 specifically comprises the following steps:
Figure BDA0003862628330000061
wherein N is the node number of DN system and uses undirected communication topology
Figure BDA0003862628330000062
Representing DGs, V =1,2, …, m, …, N is a set of nodes of the topology,
Figure BDA0003862628330000063
is a set of edges that make up the topology. (i, j) ∈ denotes a communication edge between node i and node j. Node 0 represents an upstream substation and m represents the number of dispatchable micro gas turbines (MTs), photovoltaics (PV) or Wind Turbines (WTs). For schedulable MTs, set is defined as N MT 。P 0 (t) the amount of electricity purchased from the upstream substation at time t, corresponding to a price of C G 。
Figure BDA0003862628330000064
And
Figure BDA0003862628330000065
the scheduled power generated by the ith MT and the transmission loss at time t of nodes i and j, respectively, C MT And C Loss Respectively their unit cost. Load shedding
Figure BDA0003862628330000066
The corresponding cost C of each node i at the time t λ Required power P D (t) and a factor s i (t)∈[0,1]。C s Representing the cost at the time of handover, the NS calculates the total number of handover operations.
In addition, the power flow constraint conditions in the step 2 are as follows:
(1) voltage v of node i Limiting and surviving node only current l ij The limits and limits constraints of MTs active and reactive power are:
Figure BDA0003862628330000067
Figure BDA0003862628330000068
Figure BDA0003862628330000069
Figure BDA00038626283300000610
(2) the rotational tree constraint associated with the radial structure and connectivity of DNs is:
Figure BDA0003862628330000071
Figure BDA0003862628330000072
Figure BDA0003862628330000073
wherein alpha is ij (t)∈{0,1},β ij Is an epsilon of {0,1}, (i, j) ∈ epsilon is two Boolean parameters, epsilon * Is represented by DNR [1][21]The set of remaining lines thereafter.
(3) The rule version of the dist _ flow function searches the constraint on the predefined network flow; the relationship constraint between any two connected busbars, and the nonlinear equality constraint relaxation model that makes the dist _ Flow function a convex function, are:
Figure BDA0003862628330000074
Figure BDA0003862628330000075
Figure BDA0003862628330000076
wherein
Figure BDA0003862628330000077
Q-V sag control for MTs run control.
(4) The single commodity flow spinning tree constraint is as follows:
Figure BDA0003862628330000078
|F ij (t)|≤α ij (t)M,(i,j)∈ε * (13)
slope constraint of MT. And WTs and pv the operational control function is:
Figure BDA0003862628330000079
Figure BDA00038626283300000710
wherein
Figure BDA00038626283300000711
The ith power output of the previous cycle MT.
Other constraints for nodes containing only the load are:
Figure BDA00038626283300000712
Figure BDA0003862628330000081
the number of switching actions is limited to:
Figure BDA0003862628330000082
the above constraint and network reconstruction problem model is a Mixed Integer Second Order Cone Programming (MISOCP) that can be solved for a globally optimal solution to the proposed problem using a commercial solver. The variable needed to be obtained in the above problem is P i (t),Q i (t),s i (t) and alpha ij (t)。
Wherein, the recognition discovery algorithm in the step 3 is as follows:
the information discovery process of the node i is recorded as:
Figure BDA0003862628330000083
wherein x i Representing state, including all continuous variables in DN, e.g. schedulable generators
Figure BDA0003862628330000084
And all binary variables in DN, e.g. switch state α ij 。x i ＝[P g,1 ,P g,2 ,P g,3 ,Q g,1 ,Q g,2 ,Q g,3 ,P d,1 ,P d,2 ,P d,3 ,Q d,1 ,Q d,2 ,Q d,3 ，a 12 ，a 23 ，a 13 ]. The first 12 consecutive elements represent the node generators output active and reactive power, as well as load active and reactive power. The last three binary elements represent the state of the circuit monitored by the agent. i =1,2, …, N,
Figure BDA0003862628330000085
Figure BDA0003862628330000086
in the initial state of the agenti,
Figure BDA0003862628330000087
and
Figure BDA0003862628330000088
information found at k and k +1 time slots, respectively, for agent i. a is ij For coefficients exchanging information between adjacent agents and j, 0 if agents and j are connected by a distribution line<a ij <1, otherwise a ij ＝0。
The GAN-based prospective prediction method in the step 4 comprises the following steps:
the generation model GAN based on the deep learning architecture is composed of two substructures, a generator G and a discriminator D. The generator G is responsible for synthesizing data and the discriminator D is responsible for classifying whether the input data is real data or synthesized data. Training GAN is a min-max problem:
Figure BDA0003862628330000089
in the formula [ theta ] G And theta D Parameters of the mapping G (-) and D (-) respectively denoted G and D. P rd And P gd Representing a real distribution and a gaussian distribution, respectively. Theta D May measure the difference between the composite probability distribution and the true distribution, and theta G The minimum value of (c) may push the composite probability distribution as close to the true distribution as possible.
The technical solution of the present invention is further explained with reference to the figures and the specific numerical simulation embodiments.
In this section, we use a modified IEEE 33 bus system to validate our approach. We use the JuMP framework in Julia language to model and solve the forward-looking based MISOCP problem. The data cleaning process, the load generation task and the new energy output are executed in a Python environment. And verifying the power flow scheme by using a Pandapower software package in a Python environment. The simulation computer was configured as I7-7700K 3.6Ghz CPU,8G RAM.
The improved 33-node system is a microgrid that can operate as an island. The single line diagram is shown in fig. 1. The system voltage is 12.66kV,37 branches, 33 nodes and 5 connecting switches.
For the test case, the time series of payload data is taken from ISSDA. We assume that each branch is equipped with a remote switch. The micro gas turbine is installed on the node 4, the node 8 and the node 22, and droop control is mainly adopted. The maximum active power is 2MW, and the maximum reactive power is 1MVar. Then, assuming that the nodes 8, 14 are photovoltaic mounted, the nodes 28, 30 are wind generators mounted. The maximum installed capacity of each photovoltaic wind turbine is 750kVA, and the photovoltaic wind turbine and the wind turbine are controlled by constant power factors and are both 1 (only active power is generated).
To verify the effectiveness of the method, we simulated reactive power optimization under normal operation, N-1 line fault (single line fault) and N-2 line fault (double line fault), respectively. To present the results more clearly, we performed the simulation for 24 hours per day, with an optimal time interval of 1 hour. The prediction step is set to 4 and the number of switching actions is limited to 5.
Firstly, due to the punishment effect of the power failure cost in the objective function, the connection of all buses and substation nodes can be ensured in the topology reconstruction process of normal operation of the power distribution network. Therefore, we mainly study the loss index of the network, as shown in fig. 2. In fig. 2, the solid line represents the conventional one-step optimization method, and the dashed line represents the results of the four-step predictive approach presented herein.
As can be seen from fig. 2, the prediction optimization algorithm proposed herein is compared with the conventional single-step planning algorithm, and it can be seen that the network loss of the method disclosed herein is lower than that of the single-step planning algorithm. This is because we adopt the idea of looking forward to generate prediction data with a limited number of switching actions using historical data, thereby achieving higher optimization performance in the entire optimization time domain.
Then, we analyze the fault scenario. It is assumed that the original DN is running in the initial normal operating state as shown in fig. 3. When line 9-10 fails, the switch for line 9-10 will open. After the output and the 'action' of the MTs are adjusted by the method, the optimization result based on the linear power flow is input into the nonlinear power flow environment based on the pandpower for verification. As can be seen from fig. 4, DN still satisfies the power flow equation, and the voltage of each bus node is within the constraint range. In fig. 4, lines 8 to 11 and lines 11 to 23 represent lines formed by the switches, and the remaining solid lines in the figure are normal operation lines.
When the same N-1 fault occurs on the line 20-21 in the figure 5, the algorithm can reconstruct and recover the power distribution network, and the same effect as that of the figure 4 is achieved.
When rows 18-19 and 10-11 fail, the DN can also be recovered by the method shown in fig. 6, with the same reconfiguration description as in fig. 4.
When 18-19 and 20-21 both failed, the results shown in FIG. 7 were obtained.
It should be noted that the above-mentioned embodiments are not intended to limit the scope of the present invention, and all equivalent modifications and substitutions based on the above-mentioned technical solutions are within the scope of the present invention as defined in the claims.

Claims (5)

1. A self-healing distributed network reconstruction method for a power distribution network is characterized by comprising the following steps:
step 1: constructing a network reconfiguration planning problem (DNR) mathematical model with complex constraints in the power distribution network;
step 2: establishing a power flow constraint condition of the DNR model;
and step 3: designing a communication rule for global information discovery by adopting a consensus algorithm;
and 4, step 4: a prospective prediction method based on GAN is provided.
2. The power distribution network self-healing distributed network reconstruction method according to claim 1, wherein the mathematical model of the network reconstruction planning problem in the step 1) is specifically:
Figure FDA0003862628320000011
wherein N is the node number of DN system and uses undirected communication topology
Figure FDA0003862628320000012
Representing DGs, V =1,2, …, m, …, N is a set of nodes of the topology,
Figure FDA0003862628320000013
to form a set of edges of the topology, (i, j) e epsilon represents the communication edge between node i and node j, node 0 represents an upstream substation, m represents the number of schedulable micro gas turbines (MTs), photovoltaics (PV) or Wind Turbines (WTs), and for schedulable MTs, the set is defined as N MT ，P 0 (t) the amount of electricity purchased from the upstream substation at time t, corresponding to a price of C G ，
Figure FDA0003862628320000014
And
Figure FDA0003862628320000015
the scheduled power generated by the ith MT and the transmission loss at time t at nodes i and j, respectively, C MT And C Loss Respectively, their unit cost, load shedding
Figure FDA0003862628320000016
The corresponding cost C of each node i at the time t λ Required power P D (t) and a factor s i (t)∈[0,1]，C s Representing the cost at the time of handover, the NS calculates the total number of handover operations.
3. A power distribution network self-healing distributed network reconstruction method according to claim 1, wherein the power flow constraint conditions in the step 2) are as follows:
(1) voltage v of node i Limiting and survival node only current/ ij The limits and limit constraints of MTs active and reactive power are:
Figure FDA0003862628320000017
Figure FDA0003862628320000018
Figure FDA0003862628320000019
Figure FDA00038626283200000110
(2) the rotational tree constraints associated with the radial structure and connectivity of DNs are:
Figure FDA0003862628320000021
Figure FDA0003862628320000022
Figure FDA0003862628320000023
wherein alpha is ij (t)∈{0,1},β ij Is an epsilon of {0,1}, (i, j) ∈ epsilon is two Boolean parameters, epsilon * Represents the set of remaining lines after DNR;
(3) rule version of dist _ flow function, search constraints on predefined network flows; the relationship constraint between any two connected busbars, and the nonlinear equality constraint relaxation model that makes the dist _ Flow function a convex function, are:
Figure FDA0003862628320000024
Figure FDA0003862628320000025
Figure FDA0003862628320000026
wherein
Figure FDA0003862628320000027
Q-V sag control for MTs run control;
(4) the single commodity flow spinning tree constraint is as follows:
Figure FDA0003862628320000028
|F ij (t)|≤α ij (t)M,(i,j)∈ε * (13)
the ramp constraint for MT, and the run control function for WTs and pv are:
Figure FDA0003862628320000029
Figure FDA00038626283200000210
wherein
Figure FDA00038626283200000211
The ith power output for the previous cycle MT;
other constraints for nodes containing only the load are:
Figure FDA0003862628320000031
Figure FDA0003862628320000032
the number of switching actions is limited to:
Figure FDA0003862628320000033
if the state of the line changes from time t-1 to t, λ ij (t) becomes 1
The above constraint and network reconstruction problem model is a Mixed Integer Second Order Cone Programming (MISOCP), and a commercial solver can be used to solve the global optimal solution of the proposed problem, in which the variable to be obtained is P i (t),Q i (t),s i (t) and alpha ij (t)。
4. A power distribution network self-healing distributed network reconfiguration method according to claim 1, wherein the recognition discovery algorithm in step 3) is:
the information discovery process of the node i is recorded as:
Figure FDA0003862628320000034
wherein x i Representing states, including all continuous variables in DN, generator dispatchable
Figure FDA0003862628320000035
And all binary variables in DN, e.g. switch state α ij ，x i ＝[P g,1 ,P g,2 ,P g,3 ,Q g,1 ,Q g,2 ,Q g,3 ,P d,1 ,P d,2 ,P d,3 ,Q d,1 ,Q d,2 ,Q d,3 ，a 12 ，a 23 ，a 13 ]The first 12 consecutive elements represent the node generators output real and reactive power, and the load real and reactive power, the last three binary elements represent the circuit state monitored by the agent,
Figure FDA0003862628320000036
in the initial state of the agenti,
Figure FDA0003862628320000037
and
Figure FDA0003862628320000038
information found at k and k +1 time slots, a, respectively, for agent i ij For coefficients exchanging information between adjacent agents i and j, 0 if agents i and j are connected by a distribution line<a ij <1, otherwise a ij ＝0。
5. The power distribution network self-healing distributed network reconstruction method according to claim 1, wherein the GAN-based prospective prediction method in the step 4) is as follows:
the generation model GAN based on the deep learning architecture consists of two substructures, namely a generator G and a discriminator D, wherein the generator G is responsible for synthesizing data, the discriminator D is responsible for classifying whether input data is real data or synthesized data, and the GAN is trained to be a minimum-maximum problem:
Figure FDA0003862628320000039
in the formula [ theta ] G And theta D Parameters of the mappings G (-) and D (-) respectively denoted G and D, P rd And P gd Respectively representing a real distribution and a Gaussian distribution, [ theta ] D May measure the difference between the composite probability distribution and the true distribution, and theta G The minimum value of (c) may push the composite probability distribution as close to the true distribution as possible.
CN202211168721.4A 2022-09-24 2022-09-24 Self-healing distributed network reconstruction method for power distribution network Pending CN115495986A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211168721.4A CN115495986A (en) 2022-09-24 2022-09-24 Self-healing distributed network reconstruction method for power distribution network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211168721.4A CN115495986A (en) 2022-09-24 2022-09-24 Self-healing distributed network reconstruction method for power distribution network

Publications (1)

Publication Number Publication Date
CN115495986A true CN115495986A (en) 2022-12-20

Family

ID=84469889

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211168721.4A Pending CN115495986A (en) 2022-09-24 2022-09-24 Self-healing distributed network reconstruction method for power distribution network

Country Status (1)

Country Link
CN (1) CN115495986A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117791597A (en) * 2024-02-23 2024-03-29 广东电网有限责任公司广州供电局 Power distribution network fault self-healing method and system based on machine learning

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117791597A (en) * 2024-02-23 2024-03-29 广东电网有限责任公司广州供电局 Power distribution network fault self-healing method and system based on machine learning
CN117791597B (en) * 2024-02-23 2024-05-28 广东电网有限责任公司广州供电局 Power distribution network fault self-healing method and system based on machine learning

Similar Documents

Publication Publication Date Title
Ding et al. Operational reliability evaluation of restructured power systems with wind power penetration utilizing reliability network equivalent and time-sequential simulation approaches
Zou et al. Distribution system restoration with renewable resources for reliability improvement under system uncertainties
Zhao et al. Resilient unit commitment for day-ahead market considering probabilistic impacts of hurricanes
Zhao et al. Distributed risk-limiting load restoration for wind power penetrated bulk system
Liu et al. Availability assessment based case-sensitive power system restoration strategy
Jalilpoor et al. Resilient energy management incorporating energy storage system and network reconfiguration: A framework of cyber‐physical system
Liu et al. Bi-level coordinated power system restoration model considering the support of multiple flexible resources
CN115495986A (en) Self-healing distributed network reconstruction method for power distribution network
CN113344283B (en) Energy internet new energy consumption capability assessment method based on edge intelligence
Pakdel Intelligent instability detection for islanding prediction
Memari et al. Reliability evaluation of active distribution networks based on scenario reduction method using PSO algorithm
Liu et al. A resilience enhancement scheme of cyber-physical power system for extreme natural disasters
Xia et al. Fuzzy logic based network reconfiguration strategy during power system restoration
Niu et al. Power supply reliability evaluation method for DC distribution network with power energy router
Vyas et al. Unsupervised learning in islanding studies: Applicability study for predictive detection in high solar PV penetration distribution feeders
CN111106622B (en) Active power distribution network power supply recovery method based on RMPC
Qi Operation ControlMethod of Relay Protection in Flexible DC Distribution Network Compatible with Distributed Power Supply.
Hatziargyriou et al. On-line dynamic security assessment of power systems in large islands with high wind power penetration
Ndlela et al. Reliability and security analysis of the southern Africa power pool regional grid
Singh et al. Classification of Faults in a Distributed Generator Connected Power System Using Artificial Neural Network
Salyani et al. An Analytical Framework for Evaluating the Impact of Distribution-Level LVRT Response on Transmission System Security
Wang et al. Reliability evaluation method for distribution network with distributed generations considering feeder fault recovery and network reconfiguration
Yuan et al. A review on decision-making methods for restoration control of transmission and distribution systems
Hou et al. A Tri-Level Typhoon-DAD Robust Optimization Framework to Enhance Distribution Network Resilience
Qi et al. Research on optimization of wind power system based on reliability evaluation and modeling

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination