CN-121998288-A - Power distribution network building photovoltaic bearing capacity evaluation method and system

CN121998288ACN 121998288 ACN121998288 ACN 121998288ACN-121998288-A

Abstract

The invention discloses a method and a system for evaluating the photovoltaic bearing capacity of a power distribution network building, which are used for acquiring predicted weather covariant data of the power distribution network connected with building photovoltaic, inputting the predicted weather covariant data into a trained conditional diffusion model, outputting photovoltaic output distribution, defining the Renyi divergence based on the photovoltaic output distribution and a real probability distribution corresponding to the photovoltaic output distribution, establishing a dynamic distribution robust fuzzy set constrained by a dynamic radius based on the Renyi divergence, establishing a two-stage bearing capacity distribution robust optimization model, solving the two-stage bearing capacity distribution robust optimization model by using a column and constraint generation algorithm based on the dynamic distribution robust fuzzy set, obtaining an optimization result, and obtaining the maximum photovoltaic bearing capacity based on the optimization result, thereby effectively improving the accuracy, flexibility and robustness of the photovoltaic bearing capacity evaluation of the power distribution network.

Inventors

DENG BUQING
FANG ZEHUI
GUAN CHENHAO
ZHANG JIALI
LI JIANGHUI

Assignees

国网福建省电力有限公司经济技术研究院
国网福建省电力有限公司

Dates

Publication Date: 20260508
Application Date: 20251211

Claims (10)

1. The method for evaluating the photovoltaic bearing capacity of the power distribution network building is characterized by comprising the following steps of: Acquiring predicted meteorological covariate data of a power distribution network accessed to building photovoltaics; inputting the predicted meteorological covariate data into a trained conditional diffusion model, and outputting photovoltaic output distribution; Defining a renyi divergence based on the photovoltaic output distribution and a true probability distribution corresponding to the photovoltaic output distribution; establishing a dynamic distribution robust fuzzy set constrained by a dynamic radius based on the renyi divergence; establishing a two-stage bearing capacity distribution robust optimization model of the power distribution network; solving the two-stage bearing capacity distribution robust optimization model by using a column and constraint generation algorithm based on the dynamic distribution robust fuzzy set to obtain an optimization result; And obtaining the maximum photovoltaic bearing capacity based on the optimization result.
2. The method for evaluating the building photovoltaic bearing capacity of a power distribution network according to claim 1, further comprising, before obtaining the predicted meteorological covariate data of the power distribution network connected to the building photovoltaic: acquiring historical operation data of a power distribution network connected with building photovoltaics, wherein the historical operation data comprises a distributed photovoltaic historical output sequence of each node and corresponding historical meteorological covariant data thereof; normalizing the historical operation data to obtain processed historical operation data; And training the conditional diffusion model by using the processed historical operation data to obtain a trained conditional diffusion model.
3. The method for evaluating the photovoltaic load-bearing capacity of a power distribution network building according to claim 1, wherein defining the renyi divergence based on the photovoltaic output distribution and the corresponding true probability distribution comprises: Discretizing the photovoltaic output distribution into a reference probability distribution comprising a plurality of typical scenarios; defining the R nyi divergence between the reference probability distribution and the corresponding real probability distribution, specifically: ; in the formula, Representing a reference probability distribution True probability distribution corresponding thereto The dispersion of renyi between them, Representing the order parameter, S representing the total number of typical scenes, p s representing the probability of the true probability distribution in the S-th typical scene, p 0,s representing the probability of the reference probability distribution in the S-th typical scene, and satisfying 。
4. A method of evaluating the building photovoltaic load-carrying capacity of a power distribution network according to claim 3, characterized in that before establishing a dynamic distributed robust fuzzy set constrained by dynamic radii based on said renyi divergence, it further comprises: establishing a loss function of the predicted meteorological covariate data; A dynamic radius is calculated based on the loss function.
5. The method for evaluating the photovoltaic bearing capacity of a power distribution network building according to claim 4, wherein the method for establishing the loss function of the predicted meteorological covariate data is specifically as follows: ; in the formula, Representing predicted meteorological covariate data Is used for the loss function of (a), Representing a positive weighting function of the weight of the sample, Representing all the sets of learnable weights and bias parameters in the neural network, Representing the parameters by The conditional score function of the fit, A state after noise is added at the diffusion time t is shown, Representing a true score function of the score, Representing the transition nuclei of the forward diffusion process, Representing a real photovoltaic output sample; calculating a dynamic radius based on the loss function, specifically: ; in the formula, Representing data of covariates with predicted weather The dynamic radius, r base , which varies in real time, represents a preset base safety radius constant, Representing the sensitivity adjustment coefficient.
6. The method for evaluating the photovoltaic bearing capacity of a power distribution network building according to claim 5, wherein a dynamic distribution robust fuzzy set constrained by a dynamic radius is established based on the renyi divergence, specifically: ; in the formula, Representing a dynamically distributed robust fuzzy set, Representing the S-dimensional non-negative real space.
7. The method for evaluating the photovoltaic bearing capacity of a power distribution network building according to claim 6, wherein establishing a two-stage bearing capacity distribution robust optimization model of the power distribution network comprises: establishing a three-layer objective function aiming at minimizing the sum of the investment cost of the static var compensator, the capacitor switching penalty cost and the expected operation penalty cost under the worst distribution; and establishing a first-stage constraint condition and a second-stage constraint condition corresponding to the three-layer objective function, wherein the first-stage constraint condition comprises an explicit capacity and operation constraint of a group of switched capacitor banks and a capacity planning constraint of a static reactive compensator, and the second-stage constraint condition comprises a two-stage connection constraint, a second-order cone branch power flow model of the power distribution network, a system safe operation and feeder capacity constraint and a building photovoltaic constraint.
8. The method for evaluating the photovoltaic load-bearing capacity of a power distribution network building according to claim 7, wherein three-layer objective functions are established with the aim of minimizing the sum of the investment cost of the static var compensator, the capacitor switching penalty cost and the expected operation penalty cost under the worst distribution, and specifically: ; Where n and Q SVC are the first layer decision variables, n represents the number of capacitor banks put in, Q SVC represents the capacity of the static var compensator in the planned configuration, Representing a set of nodes with static var compensators installed, Represents the investment cost per unit capacity of the static var compensator, Representing the capacity of the static var compensator as planned at node j, T representing the time period, Representing the set of nodes on which the capacitor bank is installed, c sw represents the action penalty factor of a single switch of the capacitor, for preventing unnecessary redundancy inputs, The number of capacitor banks put in at node j at time t, Representing the number of capacitor banks put into the node j at the time t-1, P is a decision variable of the second layer, y s is a decision variable of the third layer, including all voltages and currents of the node, The weight of the network loss is represented by, Represents the total network loss of the system in the s-th typical scene at the time t, Representing the weight of the light rejection penalty, Representing a collection of nodes on which the photovoltaic panels are mounted, And the amount of light discarded by the node j in the s-th typical scene at the time t is shown.
9. The method for evaluating the photovoltaic bearing capacity of a power distribution network building according to claim 1, wherein solving the two-stage bearing capacity distribution robust optimization model based on the dynamic distribution robust fuzzy set by using a column and constraint generation algorithm to obtain an optimization result comprises: Decomposing the two-stage bearing capacity distribution robust optimization model into a main problem and a sub problem, wherein the main problem optimizes a capacitor switching strategy and a static var compensator planning capacity of a first stage, approximates worst operation cost of a second stage by using auxiliary variables, and aims to provide a lower bound solution meeting the constraint of a current known worst scene set, and the sub problem aims to find a new worst probability distribution in the dynamic distribution robust fuzzy set on the premise of giving a first stage decision of the main problem, so that expected operation cost of a system is maximized and upper bound and new cutting plane constraint is provided for the main problem; Initializing a lower bound, an upper bound and iteration times; Solving the main problem to obtain an optimal planning scheme, and updating the lower bound as a main problem objective function value; the optimal planning scheme is transmitted into the sub-problem, the sub-problem is solved, worst probability distribution and sub-problem objective function values are obtained, and the upper bound is updated; And judging whether the difference between the upper bound and the lower bound is smaller than a preset convergence tolerance, if so, stopping iteration, outputting an optimization result, if not, generating a new optimal cutting plane, adding the new optimal cutting plane to the main problem, adding one to the iteration times, and returning to the step of solving the main problem.
10. A power distribution network building photovoltaic load capacity assessment system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of a power distribution network building photovoltaic load capacity assessment method according to any one of claims 1 to 9.

Description

Power distribution network building photovoltaic bearing capacity evaluation method and system Technical Field The invention relates to the technical field of photovoltaic bearing capacity evaluation, in particular to a method and a system for evaluating the photovoltaic bearing capacity of a power distribution network building. Background Building photovoltaics, an important form of distributed photovoltaic power generation, are rapidly spreading in urban distribution networks. Building photovoltaics are photovoltaic power generation systems installed on the roof, the facade and the like of a building, have the advantages of being capable of being consumed in situ, reducing power transmission loss, improving land utilization efficiency and the like, and are key technologies for realizing self-supply of building energy and promoting development of zero-carbon buildings. Building photovoltaics have become a major component of newly added power sources for power distribution networks. However, large-scale access of building photovoltaics brings new challenges to a power distribution network, on one hand, the building photovoltaics output has obvious randomness, volatility and intermittence, is influenced by various factors such as weather conditions, building orientation, shielding and the like, has uncertainty characteristics more complicated than those of a ground centralized photovoltaic power station, and on the other hand, the building photovoltaics are generally accessed to a low-voltage power distribution network, the access points are scattered and have large capacity difference, and higher requirements are put on the voltage quality, the line current carrying capacity, the transformer capacity and the like of the power distribution network. In this context, the evaluation of the photovoltaic carrying capacity of the power distribution network architecture becomes a core problem for planning operation of the power system. Building photovoltaic load capacity refers to the maximum building photovoltaic installed capacity that a power distribution network can accept under the condition that operation constraint conditions such as voltage deviation, line current, transformer capacity, power balance and the like are met. The method for accurately evaluating the bearing capacity of the building photovoltaic has important significance for guiding reasonable layout of the building photovoltaic, optimizing planning and reconstruction schemes of the power distribution network and guaranteeing safe and economic operation of the power grid. Due to the high uncertainty of building photovoltaic output, how to fully consider the uncertainty in the evaluation, not only the robustness and the safety of the evaluation result are ensured, but also the resource waste caused by excessive conservation is avoided, and the method is a technical problem to be solved currently. The existing power distribution network building photovoltaic bearing capacity evaluation method is mainly divided into two types, namely a deterministic method and an uncertainty optimization method. The deterministic method generally selects photovoltaic output data of a plurality of typical days (such as sunny days, cloudy days and rainy days) or typical moments, and determines the maximum accessible capacity of each node under the condition of not violating the operation constraint through tide calculation. The method is simple and convenient to calculate and easy to realize engineering, but is essentially based on sensitivity analysis of limited scenes, and is difficult to comprehensively reflect the random fluctuation characteristic of building photovoltaic output. Since the selection of a typical scene is subjective and cannot cover all possible operating conditions, the evaluation result of the deterministic method often lacks reliability guarantee, and the operating risk in an extreme scene may be underestimated. The uncertainty optimization method is used for processing uncertainty of building photovoltaic output from the perspective of probability theory or robust optimization, and mainly comprises three types of random optimization, traditional robust optimization and distributed robust optimization. The random optimization method assumes that the building photovoltaic output is subjected to known probability distribution, and the bearing capacity is determined by minimizing expected cost or maximizing expected income, but the method has higher accuracy requirement on distribution assumption, and in practical application, because the building photovoltaic is complicated influenced by building microenvironment (such as shielding of surrounding buildings and influence of reflected light), the output distribution is difficult to accurately describe by standard parameter distribution, so that the reliability of random optimization results is limited. The traditional robust optimization method does not rely on probability distribution assumptions