CN-122000897-A - Light storage collaborative robust planning method based on improved Bayesian optimization

CN122000897ACN 122000897 ACN122000897 ACN 122000897ACN-122000897-A

Abstract

The invention discloses an improved Bayesian optimization-based light storage collaborative robust planning method. The method constructs a combined site selection and capacity configuration planning model of a photovoltaic unit and an energy storage unit, introduces a photovoltaic extreme attack scene, realizes specific depiction of operation risk, adopts a heat engine capacity adjustment mechanism based on photovoltaic equivalent load bearing capacity, reduces heat engine redundancy and carbon emission, utilizes random weights based on Dirichlet distribution to process multiple targets, realizes diversified exploration of complex decision space, controllably reduces search space and reduces simulation expenditure by improving Monte Carlo sampling calculation with noise random forest agent, optimizes and identifies the optical storage configuration scheme based on annual simulation results and multi-target performance evaluation, solves the problem that reliability, economy and low carbon are difficult to consider in traditional planning, and realizes efficient robust planning under complex uncertainty conditions.

Inventors

DENG RUILONG
LIU YIFAN
LIU MENGXIANG
XU BOWEN
CHE XIN
CHENG PENG
CHEN JIMING

Assignees

浙江大学

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. An improved Bayesian optimization-based light storage collaborative robust planning method is characterized by comprising the following steps: Acquiring power grid data, establishing a basic operation model of photovoltaic, energy storage, a heat engine and power grid trend, quantifying the effective load bearing capacity of the photovoltaic, reducing the capacity of the heat engine according to the effective load bearing capacity of the photovoltaic, modeling the highest photovoltaic attack from the angles of time and space, fusing the highest photovoltaic attack to the basic operation model, and obtaining a photovoltaic energy storage collaborative addressing and volume-fixing optimization model by taking the electric energy shortage, line loss, investment cost and minimum carbon emission penalty as objective functions; the photovoltaic energy storage collaborative addressing and sizing optimization model uses a double-layer solving structure, the lower layer runs simulation, the upper layer uses a random forest as a proxy model to execute Bayesian optimization, and the installation position and the installation capacity of the photovoltaic energy storage in the power grid are obtained.
2. The improved Bayesian optimization-based light storage collaborative robust planning method according to claim 1, wherein the power grid data comprises load data of each bus of a power grid, line parameters, original heat engine configuration, typical photovoltaic fluctuation parameters of one year and typical load fluctuation parameters, wherein the typical photovoltaic fluctuation parameters of one year are modeled according to daily fluctuation and annual fluctuation of photovoltaic power generation, and the typical load fluctuation parameters of one year are modeled according to load fluctuation data of the last year of the power grid.
3. The improved bayesian optimization-based light and storage collaborative robust planning method according to claim 1, wherein the basic operation model of photovoltaic, energy storage, heat engine and power grid power flow comprises the following constraints: And the capacity and the installation position of the photovoltaic and energy storage units are constrained, so that the thermal engine constraint, the line tide constraint and the bus power balance constraint of physical feasibility are met.
4. The improved bayesian optimization-based optical storage collaborative robustness planning method according to claim 1, wherein the quantification of the payload carrying capacity of the photovoltaic is achieved, and the reduction of the capacity of the heat engine according to the payload carrying capacity of the photovoltaic is specifically as follows: Defining weights proportional to load Calculating the bearing capacity of the photovoltaic effective load by carrying out weighted average on the photovoltaic output value The modeling formula is as follows: is a photovoltaic device Is used for the power rating of the (c), For the moment of time The available proportion of the photovoltaic power generation is that, According to Rated capacity reduction ratio of heat engine The calculation formula is as follows and according to The new capacity of the heat engine is obtained, and the modeling formula is as follows: Wherein the method comprises the steps of Is a heat engine Is used for the power supply of the engine, Is a heat engine Is used for the new power rating of (1), Is a set of heat engines.
5. The improved Bayesian optimization-based light storage collaborative robust planning method of claim 1, wherein the modeling of the highest photovoltaic attack from time and space angles is fused to a basic operation model, specifically comprising the steps of screening the moment in which the load and the photovoltaic output are higher simultaneously in one year in the time dimension, and passing parameters And (3) with Determining unique attack moment by adjustment of (a) In the space dimension, combining node degree and photovoltaic capacity to select key photovoltaic units to form an attack set The strongest attack was modeled as follows: Wherein the method comprises the steps of Is a bus bar At the upper moment Is used for the actual load of the (c) in the (c), Is the middle-front part of annual load Is used as a reference to the value of (a), Middle and front part of available proportion for annual photovoltaic power generation Is used as a reference to the value of (a), For the set of attack moments, Representing photovoltaic Whether or not to be mounted on the bus bar , Is a bus bar Is used for the degree of (a), For the collection of all the bus bars of the power grid, Is a photovoltaic collection.
6. The improved bayesian-optimized-based optical storage collaborative robustness planning method according to claim 1, wherein the objective function is specifically: Wherein, the For the annual energy unsatisfied quantity of the objective function, Is a bus bar At the upper moment Is characterized in that the load shedding amount of the steel plate is equal to the load shedding amount of the steel plate, As a function of the line loss objective, Is a circuit At the moment of time Is used for the active power flow of the (a), Is a circuit Is used for the resistance of the (c), Is a power reference value of the system; As a function of the cost of the object, At the cost of a unit of photovoltaic capacity, Is a photovoltaic device Is used for the power rating of the (c), As a unit cost of the stored power portion, For the unit cost of the stored energy portion, Is an energy storage unit Is used for the energy-rated capacity of the (a), As a penalty factor for the amount of carbon emissions, Is a heat engine Is used for the new power rating of (1), For the set of time instants, For the collection of all the bus bars of the power grid, For the collection of lines, Is a set of heat engines.
7. The improved Bayesian optimization-based light storage collaborative robust planning method according to claim 1, wherein the lower simulation in the double-layer solving structure specifically comprises the steps of obtaining annual operation results based on candidate solution operation simulation selected by each iteration of Bayesian optimization, and further calculating an objective function.
8. The improved bayesian-optimized-based optical storage collaborative robust planning method according to claim 1, wherein a random weight mechanism based on Dirichlet distribution is adopted in the bayesian optimization process, and an average weight is adopted in an initialization stage, specifically: Wherein, the Represents the mth tree in the random forest M, For a uniform distribution over the multidimensional probabilistic simplex, the generated random weight vector is non-negative and sums to 1, Is the normalized objective function.
9. The improved bayesian-optimized-based optical storage collaborative robust planning method according to claim 1, wherein the bayesian optimization iterative process is directly from a random forest proxy model Is sampled in the prediction distribution of (a) Estimating the desired acquisition function, introducing noise in the proxy model To simulate photovoltaic and load uncertainties, in particular: Wherein the random forest is composed of The tree is composed of a plurality of trees, Is the first Tree pair candidate points Is used to determine the predicted value of (c), For the historical optimal target value, For measuring current candidate points Relative to Is a function of the potential improvement in terms of (a), For the next iterative solution.
10. An improved bayesian optimization based light storage collaborative robust planning apparatus comprising a memory and one or more processors, the memory having executable code stored therein, wherein the processor, when executing the executable code, implements an improved bayesian optimization based light storage collaborative robust planning method according to any of claims 1-9.

Description

Light storage collaborative robust planning method based on improved Bayesian optimization Technical Field The invention relates to the field of power system planning, in particular to a light storage collaborative robust planning method based on improved Bayesian optimization. Background Photovoltaic power generation costs continue to drop, and large-scale rapid deployment has been achieved in recent years. Statistics show that global photovoltaic total packaging machine capacity increases by approximately two times during 2018 to 2023, with utility scale photovoltaic accounting for the major portion of the newly added packaging capacity. Although the photovoltaic resources are abundant and have obvious emission reduction benefits, the output of the photovoltaic power generation system is greatly influenced by meteorological conditions, and obvious uncertainty exists. Furthermore, as photovoltaic systems are heavily accessed into smart devices, their communication and control links are exposed to more complex network environments, with a consequent increase in potential risk of attack, which may have an impact on system stability. In order to improve the flexibility and reliability of the system, the energy storage system can be used as an important adjusting means, and the energy storage system and the photovoltaic combined planning are beneficial to improving the running performance of the power grid while reducing the carbon emission. However, the existing optical storage collaborative planning method generally assumes that the operation environment is relatively stable, and extreme conditions such as photovoltaic output fluctuation, load uncertainty, network attack and the like are not fully considered, so that the problem of insufficient robustness of the planning scheme in actual operation is caused. On the other hand, the traditional planning optimization method often faces the problems of large calculated amount, low convergence efficiency and the like in a multi-target, strong-coupling and simulation-driven calculation scene, and a planning scheme which takes reliability, economy and safety into consideration is difficult to obtain in a reasonable time. Therefore, there is a need for an optical storage collaborative robust planning method capable of simultaneously processing uncertainty and attack risk and having efficient solving capability, so as to improve the operation safety and planning quality of the system in a complex environment. Disclosure of Invention The invention aims to provide an improved Bayesian optimization-based light storage collaborative robust planning method for solving the problems that the existing light storage collaborative planning method is insufficient in robustness, low in solving efficiency and difficult to simultaneously consider power supply reliability and economy when processing photovoltaic output uncertainty, load fluctuation and potential photovoltaic attack. The invention aims at realizing the technical scheme that the light storage collaborative robust planning method based on improved Bayesian optimization comprises the following steps: Acquiring power grid data, establishing a basic operation model of photovoltaic, energy storage, a heat engine and power grid trend, quantifying the effective load bearing capacity of the photovoltaic, reducing the capacity of the heat engine according to the effective load bearing capacity of the photovoltaic, modeling the highest photovoltaic attack from the angles of time and space, fusing the highest photovoltaic attack to the basic operation model, and obtaining a photovoltaic energy storage collaborative addressing and volume-fixing optimization model by taking the electric energy shortage, line loss, investment cost and minimum carbon emission penalty as objective functions; the photovoltaic energy storage collaborative addressing and sizing optimization model uses a double-layer solving structure, the lower layer runs simulation, the upper layer uses a random forest as a proxy model to execute Bayesian optimization, and the installation position and the installation capacity of the photovoltaic energy storage in the power grid are obtained. Further, the power grid data comprise load data of each bus of the power grid, line parameters, original heat engine configuration, typical photovoltaic fluctuation parameters of one year and typical load fluctuation parameters, wherein the typical photovoltaic fluctuation parameters of one year are modeled according to daily fluctuation and annual fluctuation of photovoltaic power generation, and the typical load fluctuation parameters of one year are modeled according to load fluctuation data of the last year of the power grid. Further, the basic operation model of the photovoltaic, energy storage, heat engine and power grid power flow comprises the following constraints: And the capacity and the installation position of the photovoltaic and energy storage units are constraine