CN-122026462-A - Multi-level optical storage and charging collaborative planning method based on target cascade and progressive opposite impact

CN122026462ACN 122026462 ACN122026462 ACN 122026462ACN-122026462-A

Abstract

The invention discloses a multi-level optical storage and filling collaborative planning method based on target cascading and progressive hedging. The method comprises the steps of firstly obtaining power distribution network topology and parameters, constructing a typical daily scene set to represent source load uncertainty, secondly establishing a two-stage optimization configuration model of a main system layer and a subsystem layer, taking tie line power as a coupling variable to form a joint optimization problem, finally carrying out double-layer decomposition and solving, carrying out scene decomposition by utilizing a progressive opposite-impact algorithm, carrying out interlayer decomposition by utilizing a target cascade decomposition method, introducing virtual load and virtual power, and realizing coordination between scenes and interlayer by means of penalty function coordination and cross-layer bidirectional feedback iteration. The method can effectively process the uncertainty of the source load, solve the problems of difficult convergence and low efficiency of the traditional algorithm, and realize accurate collaborative planning of the optical storage and filling resources.

Inventors

GUO LINHAI
DONG ZHENGCHENG
TIAN MENG
LUO BINGYANG
WANG YU
LIU XIN

Assignees

武汉理工大学

Dates

Publication Date: 20260512
Application Date: 20260209

Claims (10)

1. The multi-level optical storage and inflation collaborative planning method based on target cascade and progressive opposite flushing is characterized by comprising the following steps of: Firstly, obtaining topological structures and system parameters of a power distribution network, constructing a source load scene set, and screening to obtain a plurality of typical daily scenes, wherein the plurality of typical day show scenes form a multi-scene set for representing uncertainty of source load output; Step two: the method comprises the steps of establishing a two-stage optimizing configuration model of a main system layer, namely taking an energy storage device of a power distribution network of the main system layer and an electric vehicle charging facility as objects, and constructing an optimizing model comprising a first stage and a second stage, wherein the first stage is a configuration planning stage, the site selection and the constant volume of equipment are determined, and the configuration variable of the first stage is irrelevant to scenes; establishing a subsystem layer two-stage optimization configuration model, namely taking distributed photovoltaics of a subsystem layer power distribution network as objects, and adopting a two-stage structure identical to a main system layer to construct the optimization model; establishing an interlayer coupling relation, namely taking the power of a connecting line between a main system layer and a subsystem layer as a coupling variable to form an interlayer joint optimization problem; step three, carrying out double-layer decomposition solving on the interlayer joint optimization problem: Decomposing the first layer into scene decomposition, and decomposing the multi-scene optimization problem into a plurality of single-scene sub-problems by utilizing a progressive hedging algorithm; Decomposing the second layer into interlayer decomposition, under each single scene, equivalent the power of the connecting line to be virtual load from the angle of a main system layer, equivalent the power of the connecting line to be virtual power source from the angle of a subsystem layer, and decoupling an interlayer joint optimization problem under the single scene to be a main system layer sub-problem and a subsystem layer sub-problem by using a target cascade decomposition method; And adopting a penalty function to coordinate consistency between the virtual load and the virtual power supply, carrying out iterative solution through cross-layer bidirectional feedback, and enabling configuration variables of all single-scene sub-problems to be consistent through inter-scene coordination until convergence conditions are met, so as to obtain an optimal configuration scheme of the optical storage and charging resources.
2. The method of claim 1, wherein in the two-stage optimization configuration model, the first stage and the second stage are matched in the following manner: The configuration variables of the first stage are determined once in the initial stage of planning, remain unchanged in the whole planning period, and provide equipment configuration boundaries for the operation scheduling of the second stage; Under the constraint of the configuration variables of the first stage, the operation variables of the second stage are respectively optimized for each scene in the multi-scene set so as to adapt to the source load output conditions under different scenes; The optimization target is the sum of the first-stage configuration target and the second-stage multi-scene expected operation target is minimized, so that the comprehensive operation performance of the configuration scheme in all possible scenes is optimal.
3. The method of claim 1, wherein the main system layer distribution network is a 10kV side feeder, the subsystem layer distribution network is a 0.4kV side distribution network, and the electric vehicle charging facility comprises a slow charging pile and a fast charging pile.
4. The method according to claim 1, wherein the process of constructing the source load scene set and screening in the first step specifically includes: a multidimensional random sampling technology is applied, and a large-scale source-load combination is generated by comprehensively considering a time dimension and a space dimension; A gradual rejecting strategy is applied, scene combinations which do not meet the physical constraint requirements or have high repeatability are gradually removed according to a preset screening rule, and a plurality of representative typical daily scenes are reserved; each typical day scene is assigned a corresponding probability weight that is used to calculate the second-stage multi-scene expected operating objective.
5. The method of claim 1, wherein in the two-stage optimization configuration model of the main system layer: The configuration variables of the first stage comprise site selection variables of the energy storage device, active capacity and reactive capacity of the energy storage device, site selection variables of various types of charging facilities and installation quantity of various types of charging facilities, wherein the site selection variables are 0-1 discrete variables and are used for representing whether corresponding nodes are provided with corresponding equipment or not; The operation variables of the second stage comprise the charging power of the energy storage device at each moment in each scene, the discharging power of the energy storage device at each moment in each scene, the charging state mark of the energy storage device at each moment in each scene, the discharging state mark of the energy storage device at each moment in each scene and the power distribution of the charging facilities at each moment in each scene; the second stage is operated with the aim of minimizing system grid loss and reducing carbon emissions.
6. The method of claim 1, wherein in the subsystem layer two-stage optimization configuration model: The configuration variables of the first stage comprise site selection variables of the distributed photovoltaic and rated capacity of the distributed photovoltaic, wherein the site selection variables are 0-1 discrete variables used for representing whether the distributed photovoltaic is installed on a corresponding node or not; the operation variables of the second stage comprise the active power output of the distributed photovoltaic at each moment in each scene and the reactive power output of the distributed photovoltaic at each moment in each scene; The second stage is operated with the aim of minimizing system network loss and improving node voltage quality.
7. The method according to claim 1, wherein the main system layer two-stage optimization configuration model includes a power flow constraint, an energy storage device operation constraint, a carbon emission constraint and a charging facility constraint, and the constraints are respectively satisfied in each scenario, and specifically include: The method comprises the steps of power flow constraint, namely, meeting active power balance constraint and reactive power balance constraint of a power distribution system at each moment of each scene, and branch current non-out-of-limit constraint and node voltage amplitude constraint, wherein when a traditional generator exists in the power distribution system, the active output and reactive output of the traditional generator are required to be within rated value ranges; the energy storage device is restricted in operation, namely the state of charge restriction prescribes that the state of charge under any scene and time is required to be between an upper limit and a lower limit and is consistent at the beginning and the end of a dispatching cycle, and the influence of charge and discharge efficiency is required to be considered in updating calculation of the state of charge; the full life cycle constraint is that the constraint is set based on the relation among the total circulation times of the energy storage device in the full life cycle, the effective annual operation days and the upper limit of the daily charge and discharge times, so that the charge and discharge times in the planning period are ensured not to exceed the total allowed times in the full life cycle; Setting the actual carbon emission amount of the area not exceeding the carbon emission quota, wherein the actual carbon emission amount is calculated by the carbon emission coefficient of the unit electric quantity and the net purchase electric quantity of the system; The charging facility constraint is to set the upper and lower limit constraint of the configuration quantity of each type of charging facility, the upper limit of the total quantity of resources configured by the main system layer and the subsystem layer, and the charging load distribution constraint of the electric automobile, The electric automobile charging load distribution constraint specifically comprises: calculating the charging load of each charging facility based on a probability model of the electric automobile selecting the charging facility; In the probability model, the probability of selecting a certain charging facility by the electric automobile is inversely related to the distance between the charging facility and the nearest access point, and is positively related to the number of vehicles to be charged at the access point; the charging load of each charging facility is equal to the product of the number of electric vehicles selecting the facility and the charging power of the single vehicle, and does not exceed the product of the rated power and the installation number of the facility.
8. The method according to claim 1, wherein the subsystem layer two-stage optimization configuration model includes a power balance constraint, a capacity constraint, a node voltage constraint and a demand response constraint, and the constraints are respectively satisfied in each scenario, and specifically include: the power balance constraint is that the active power balance and the reactive power balance of the power distribution system at each moment in each scene are met; the capacity constraint is to limit the maximum total capacity of the accessible distributed photovoltaic under the distribution system and the maximum capacity of the installable distributed photovoltaic at each access point; node voltage constraint, namely setting an allowable deviation range of the voltage amplitude of each node and the rated voltage; And (3) a demand response constraint, namely introducing an adjustable capacity coefficient of the load, defining an accumulated time upper limit of the participation response, a single response minimum duration and an adjacent response minimum interval time, and calculating the power adjustment quantity after the participation response based on the adjustable capacity coefficient.
9. The method according to claim 1, wherein a link power constraint is set in the interlayer decomposition process, specifically: the active power of the connecting line is required to be within the upper limit and the lower limit of the transmission capacity; The reactive power of the connecting line is required to be within the upper limit and the lower limit of the transmission capacity; The penalty function is an augmented Lagrangian penalty function, and a primary term multiplier and a secondary term penalty function aiming at the deviation between the virtual load variable and the virtual power variable are added in a main system layer optimization target and a subsystem layer optimization target; The cross-layer bidirectional feedback iterative solving process sets double convergence criteria, wherein the first convergence criteria are that the absolute value of the deviation between the virtual load variable calculated by the main system layer and the virtual power variable calculated by the subsystem layer is smaller than or equal to a first threshold value, and the second convergence criteria are that the absolute value of the variation of the configuration variable between two adjacent iterations is smaller than or equal to a second threshold value.
10. The method according to claim 9, wherein the specific process of cross-layer bi-directional feedback iterative solution is: After solving the self-optimization problem, the main system layer transmits the optimization value of the virtual load to the subsystem layer as a boundary condition; the subsystem layer considers the minimum deviation between the virtual power supply variable and the boundary condition transmitted by the main system layer when solving the self-optimization problem, and feeds back the optimization value of the virtual power supply to the main system layer after the solution is completed; updating a first term multiplier and a second term multiplier of the extended Lagrangian penalty function according to the feedback result of the previous iteration, and gradually adjusting configuration variables and operation variables until the first convergence criterion and the second convergence criterion are simultaneously met; After the interlayer iteration convergence of each single field Jing Zi problem is completed, the configuration variables of each single scene are coordinated to be consistent through the scene interval penalty function of the progressive hedging algorithm, and the optimal configuration scheme of the optical storage and filling resources is output.

Description

Multi-level optical storage and charging collaborative planning method based on target cascade and progressive opposite impact Technical Field The invention belongs to the technical field of power system distributed resource collaborative planning, and particularly relates to a multi-level optical storage and filling collaborative planning method based on target cascading and progressive hedging. Background With the increasing aggravation of global energy crisis and the increasing serious of environmental pollution, the permeability of novel loads such as distributed power sources, energy storage devices and electric vehicles represented by photovoltaic power generation in a power distribution network is continuously improved. The light storage and charging integrated system is used as an important form for eliminating new energy and supporting flexible operation of a power grid, and has become an important direction for the development of a power system. However, the access of large-scale distributed resources makes the distribution network transform from a traditional unidirectional radial network to an active distribution network with source-load interaction, and the planning and operation of the distribution network face serious challenges. In the existing optical storage and filling collaborative planning technology, two modeling solution ideas mainly exist in a centralized mode and a distributed mode. The centralized planning method is to build a unified mathematical model to solve all levels (such as 10kV feeder lines and 0.4kV distribution network) and all resources (photovoltaic, energy storage and charging piles) in the distribution network as a whole. However, the method has obvious limitations that firstly, along with the expansion of the power distribution network scale and the rapid increase of the number of access devices, the optimization variable grows exponentially, so that the calculation scale is huge, the dimension disaster is easy to occur, and the optimal solution is difficult to obtain in a reasonable time. Secondly, centralized modeling often ignores the actual situation that different voltage classes of the power distribution network possibly belong to different benefit subjects or management departments, cannot embody the management requirements of independent operation and collaborative optimization of each level, and is difficult to protect privacy data of each subject. In addition, when uncertainty of source load (photovoltaic output, load demand) is considered, if a multi-scenario technology is adopted, the scale of the centralized model is further expanded, so that solving is extremely difficult. In order to solve the defects of the centralized method, a distributed planning method is adopted in part of the prior art. Such methods typically utilize intelligent algorithms (e.g., particle swarm algorithms, differential evolution algorithms, etc.) to perform hierarchical solutions to the model. However, when the problem of complex nonlinear programming is solved, the distributed programming method based on the heuristic intelligent algorithm often lacks strict mathematical convergence proof, the algorithm has low convergence speed and is very easy to fall into a locally optimal solution, so that the finally generated equipment configuration scheme (such as a locating and volume fixing result) has poor economical efficiency or cannot meet the safety constraint of a power grid. Meanwhile, when the prior hierarchical planning method processes interlayer coupling variables (such as tie line power), an effective coordination mechanism is often lacked, and Nash equilibrium or global optimal coordination of a main system layer and a subsystem layer under multiple scenes is difficult to realize. Disclosure of Invention The invention aims to solve the defects of the background technology, and provides a multi-level optical storage and inflation collaborative planning method based on target cascade and progressive opposite flushing, which effectively processes source load uncertainty on the premise of ensuring benefits and privacy of different levels of main bodies, solves the problems of difficult convergence and low calculation efficiency of the traditional algorithm, and realizes accurate collaborative planning of multi-level optical storage and inflation resources. The technical scheme adopted by the invention is that the multi-level optical storage and inflation collaborative planning method based on target cascade and progressive opposite flushing comprises the following steps: Firstly, obtaining topological structures and system parameters of a power distribution network, constructing a source load scene set, and screening to obtain a plurality of typical daily scenes, wherein the plurality of typical day show scenes form a multi-scene set for representing uncertainty of source load output; Step two: the method comprises the steps of establishing a two-stage optimizing configuratio