Active power distribution network distributed robust optimization scheduling method, system and device considering net-load-storage coordination and medium

Abstract

The application discloses a method, system, device and medium for active power distribution network distributed robust optimization scheduling considering net-load-storage coordination, the method comprising: obtaining operation data of the active power distribution network, constructing a target function and a constraint condition according to the operation cost, and establishing a deterministic scheduling model based on the target function and the constraint condition; performing marginal distribution fitting on the output of each distributed power supply, and depicting the nonlinear correlation between the outputs of multiple power supplies; generating an initial scenario set through hierarchical sampling and scenario reduction, and constructing a fuzzy set of scenario probability distribution according to the initial scenario set; based on the deterministic scheduling model and the fuzzy set, reconstructing the deterministic scheduling model into a multi-stage distributed robust optimization model, solving the multi-stage distributed robust optimization model by using a decomposition iteration algorithm, obtaining a day-ahead optimization scheduling scheme of the active power distribution network, and realizing the optimization scheduling of the active power distribution network, which has good engineering practicability and expansibility.

Description

Technical Field

[0001] This invention relates to the field of active distribution network optimization scheduling technology with distributed power sources, and particularly to a method, system, equipment and medium for active distribution network distributed robust optimization scheduling that takes into account grid-load-storage coordination. Background Technology

[0002] With the rapid development of new energy power generation and power electronic converter technology, the proportion of distributed power sources, represented by wind power, photovoltaics, and energy storage, in medium and low voltage distribution networks has been increasing year by year. Taking distributed photovoltaics as an example, by the end of 2025, the cumulative installed capacity of distributed photovoltaic power generation had exceeded 493.3 GW, accounting for approximately 51.6% of the total installed photovoltaic capacity. This trend has driven the continuous development of medium and low voltage distribution network technology, and active distribution networks have become an inevitable trend in the development of new distribution networks. However, facing the high proportion of distributed wind and photovoltaic access, their power output has the characteristics of uncertainty and "anti-peak" mode, and the problems of new energy absorption and voltage limit exceedance occur frequently. How to make full use of various technical means within the system to smooth out random changes in source load and ensure the safe and stable operation of the system while meeting various dispatching needs requires in-depth research. Static network reconfiguration, conventional power load demand response, and battery energy storage participating in distribution network optimization dispatching have the ability to reduce grid dispatching costs and improve the absorption capacity of new energy. Data-driven distributed robust optimization is an effective means to deal with the uncertainty of new energy and is worthy of in-depth research and application in the field of active distribution networks.

[0003] However, existing research mainly utilizes static network reconfiguration, single energy storage, or demand response to participate in distribution network optimization scheduling, which suffers from insufficient utilization of flexibility resources and reduced user satisfaction. It fails to comprehensively explore multi-side technical regulation methods such as dynamic network reconfiguration on the grid side of active distribution networks, energy storage-side consideration of the characteristics of mobile energy storage in electric vehicles, and demand-side consideration of real-time electricity to participate in grid optimization scheduling. Furthermore, data-driven distributed robustness in the face of high-penetration distributed power generation access to active distribution networks remains unexplored and urgently requires further research. Therefore, this invention provides a distributed robust optimization scheduling method for active distribution networks that considers grid-load-storage coordination. Summary of the Invention

[0004] In view of the above-mentioned existing problems, the present invention provides a distributed robust optimization scheduling method, system, equipment and medium for active distribution networks that takes into account grid-load-storage coordination, and solves the problems of insufficient flexibility of traditional regulation methods, insufficient exploitation of the mobile energy storage potential of electric vehicles, and difficulty in balancing robustness and economy in uncertainty handling methods in existing active distribution network optimization scheduling technologies.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination, including: Obtain operational data of the active distribution network, construct an objective function and constraints based on the operational costs, and establish a deterministic scheduling model based on the objective function and constraints; Marginal distribution fitting is performed on the output of each distributed power source, and the nonlinear correlation between the outputs of multiple power sources is characterized. An initial scene set is generated through hierarchical sampling and scene reduction. A fuzzy set of scene probability distribution is constructed based on the initial scene set. Based on the deterministic scheduling model and fuzzy sets, the deterministic scheduling model is reconstructed into a multi-stage sub-Blubar optimization model. The multi-stage sub-Blubar optimization model is solved using a decomposition iterative algorithm to obtain the day-ahead optimal scheduling scheme for the active distribution network, thereby realizing the optimal scheduling of the active distribution network.

[0006] As a preferred embodiment of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination described in this invention, the deterministic scheduling model established based on the objective function and constraints includes: The objective function is to minimize the total scheduling cost of the active distribution network. The constraints must include at least grid-side regulation constraints, load-side regulation constraints, energy storage-side regulation constraints, distributed generation operation constraints, and grid power flow and security constraints. A deterministic scheduling model is constructed based on the joint constraints of the objective function.

[0007] As a preferred embodiment of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination described in this invention, the marginal distribution fitting of the output of each distributed power source includes: Obtain the output data of each distributed power source during its historical operating cycle, and construct a historical output sample set for each power source. For each distributed power source, the output probability density function of each distributed power source is fitted using nonparametric estimation based on the historical output sample set; The output probability density function of each distributed power source is used as the marginal probability density function of the corresponding distributed power source.

[0008] The beneficial effect of this preferred technical solution is that it can accurately reproduce the true random distribution characteristics of each power source output without the need for a preset distribution pattern.

[0009] As a preferred embodiment of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination described in this invention, the nonlinear correlation between the outputs of multiple power sources includes: Perform a probability integral transformation on the historical output sequence of each distributed power source to convert each historical output sequence into a transformed random variable that follows a standard uniform distribution; Based on the transformed random variables, a joint distribution model among the outputs of multiple distributed power sources is constructed using a multivariate correlation distribution model. By capturing the nonlinear correlation of the output of multiple distributed power sources through a joint distribution model, a multidimensional joint probability distribution is generated.

[0010] The beneficial effect of this preferred technical solution is that it accurately captures the nonlinear coupling relationship between wind and solar power output, thus conforming to the real spatiotemporal correlation characteristics.

[0011] As a preferred embodiment of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination described in this invention, the method includes: generating an initial scenario set through hierarchical sampling and scenario reduction, and constructing a fuzzy set of scenario probability distributions based on the initial scenario set, comprising: Based on the multidimensional joint probability distribution, the Latin hypercube stratified sampling method is used to generate an initial scenario set that takes into account the output correlation of multiple distributed power sources. The initial scene set is reduced to obtain the reduced typical scene set and the initial probability distribution corresponding to each typical scene. Using the initial probability distribution as the center, a fuzzy set of the scene probability distribution is constructed by utilizing the comprehensive norm constraint.

[0012] As a preferred embodiment of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination described in this invention, wherein: the deterministic scheduling model is reconstructed into a multi-stage sub-robust optimization model including: The deterministic scheduling model is reconstructed into a two-stage sub-Brussels bar optimization model; The first phase determines the day-ahead scheduling plan based on the predicted output information and load forecast information of distributed power sources. The second phase is based on the day-ahead scheduling plan determined in the first phase. The worst-case probability distribution scenario is obtained in the fuzzy set. Under the worst-case probability distribution scenario, the demand response load and the power curtailment of new energy are adjusted in real time. The objective function of the two-stage sub-Bruker optimization model is to minimize the sum of the decision cost in the first stage and the expected cost in the second stage under the worst-case probability distribution. The feasible region of the decision variables in the first stage consists of constraints that are only related to the first stage, while the feasible region of the decision variables in the second stage consists of second-stage constraints that are related to the variables in the first stage and the scenario.

[0013] The beneficial effect of this preferred technical solution is that by separating day-ahead and real-time decision-making, it balances scheduling economy and robustness in extreme scenarios.

[0014] As a preferred embodiment of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination described in this invention, the multi-stage distributed robust optimization model is solved using a decomposition iterative algorithm to obtain the day-ahead optimal scheduling scheme for the active distribution network, including: The two-stage bibar optimization model is decoupled from the main problem and sub-problems by using a column and constraint generation algorithm for iterative solution. The main problem is a single-layer optimization problem given a finite number of adverse probability distributions; The subproblem is to find the worst-case scenario probability distribution within the fuzzy set based on the first-stage decision variables obtained from the main problem; The main problem and subproblems are solved iteratively until the upper and lower bounds meet the convergence accuracy. The first-stage decision variables are then output as the day-ahead optimization scheduling scheme for the active distribution network.

[0015] Secondly, this invention provides a system for a distributed robust optimization scheduling method for active distribution networks that considers grid-load-storage coordination, including: The model building module is used to acquire the operating data of the active distribution network, construct the objective function and constraints based on the operating cost, and establish a deterministic scheduling model based on the objective function and constraints. The fuzzy set construction module is used to fit the marginal distribution of the output of each distributed power source and characterize the nonlinear correlation between the outputs of multiple power sources. It generates an initial scene set through hierarchical sampling and scene reduction, and constructs a fuzzy set of scene probability distribution based on the initial scene set. The model solving module is used to reconstruct the deterministic scheduling model into a multi-stage sub-Blubar optimization model based on the deterministic scheduling model and fuzzy sets, and solve the multi-stage sub-Blubar optimization model using a decomposition iterative algorithm to obtain the day-ahead optimal scheduling scheme of the active distribution network, thereby realizing the optimal scheduling of the active distribution network.

[0016] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, characterized in that the processor executes the computer program to implement the steps of the active distribution network distributed robust optimization scheduling method that considers grid-load-storage coordination.

[0017] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination.

[0018] Compared with existing technologies, the beneficial effects of this invention are as follows: Starting from dynamic network reconfiguration, electric vehicle mobile energy storage, and real-time electricity price demand response to achieve multi-side coordinated regulation, this invention constructs an active distribution network distributed robust optimization scheduling model that considers grid-load-storage coordination with the goal of economic efficiency. It combines data-driven scenario generation and comprehensive norm fuzzy sets to accurately address the uncertainty of wind and solar power output. Through multi-side coordination of dynamic network reconfiguration, electric vehicle mobile energy storage, and real-time electricity price demand response, this invention can significantly reduce system scheduling costs and renewable energy curtailment rates, effectively reduce node voltage deviations, improve user satisfaction, and achieve synergistic optimization of grid economy, security, and user comfort. The scenario generation method based on kernel density estimation-Copula-LHS in this invention can accurately characterize the nonlinear correlation of high-dimensional wind and solar power output. Combined with distributed robust optimization using comprehensive norm fuzzy sets, it effectively alleviates the shortcomings of traditional robust optimization, such as over-conservatism and dependence on stochastic programming distributions. This invention is adaptable to different electric vehicle access scales, and its scheduling performance and computational efficiency are superior to deterministic optimization, stochastic optimization, and traditional robust optimization, demonstrating good engineering practicality and scalability. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic diagram of the overall process logic of an active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination provided in an embodiment of the present invention; Figure 2 A schematic diagram of an improved IEEE 33-node active distribution network topology, which takes into account grid-load-storage coordination, as provided in an embodiment of the present invention. Figure 3 A schematic diagram of the predicted output of distributed power sources in an active distribution network, which takes into account grid-load-storage coordination, as an embodiment of the present invention. Figure 4 A schematic diagram of the conventional electrical load and electric vehicle load of an active distribution network, which takes into account grid-load-storage coordination, as provided in an embodiment of the present invention; Figure 5 A thermal diagram of Kendall correlation coefficients under different scenario generation methods for an active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination, as provided in an embodiment of the present invention. Figure 6A schematic diagram of the predicted output of distributed power sources in an active distribution network, which takes into account grid-load-storage coordination, as an embodiment of the present invention. Figure 7 A time-of-use pricing and real-time pricing based on net load status are illustrated in an embodiment of the present invention for an active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination. Figure 8 A schematic diagram of load offset for Scheme 4 of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination, provided in an embodiment of the present invention; Figure 9 A schematic diagram of the power balance situation of Scheme 4 of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination, provided in an embodiment of the present invention; Figure 10 This is a schematic diagram illustrating the optimization results of an active distribution network under different numbers of electric vehicles, based on an embodiment of the present invention, which considers grid-load-storage coordination in a distributed robust optimization scheduling method for active distribution networks. Detailed Implementation

[0021] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0022] Example 1, referring to Figure 1 As an embodiment of the present invention, a robust distributed optimization scheduling method for active distribution networks considering grid-load-storage coordination is provided, comprising: S100: Acquire operational data of the active distribution network, construct objective functions and constraints based on operational costs, and establish a deterministic scheduling model based on the objective functions and constraints; S200: The marginal distribution of the output of each distributed power source is fitted, and the nonlinear correlation between the outputs of multiple power sources is characterized. An initial scene set is generated through hierarchical sampling and scene reduction. A fuzzy set of scene probability distribution is constructed based on the initial scene set. In one optional embodiment, stratified sampling can be quasi-Monte Carlo stratified sampling, which divides the stratified intervals of equal probability according to the edge distribution of the output of each wind and solar unit; generates a low-difference quasi-random sequence to replace pure random numbers to avoid sample clustering; matches quasi-random values ​​according to the stratified intervals to obtain stratified sampling points for each unit; and couples the sampling points of each unit with a multivariate correlation distribution model to form a complete wind and solar random scenario. In another alternative embodiment, stratified sampling can also be orthogonal stratified sampling, where the output range of each distributed power source is equally divided into strata, and the number of strata is determined; an orthogonal sampling matrix is ​​constructed to ensure that only a small number of samples are drawn for each stratum and each variable; sampling points are selected from each stratum, and multi-unit stratified samples are orthogonally combined; multivariate correlation constraints are introduced to correct the sample combination and generate output scenarios with correlation. In this embodiment of the invention, stratified sampling includes the Latin hypercube stratified sampling method; Specifically, the output probability interval of each wind and solar turbine is determined based on the high-dimensional joint distribution of Copula; the output distribution of each turbine is evenly layered, and only one sample point is extracted from each layer; the layered sample numbers of each turbine are randomly shuffled to complete the sample pairing between variables; a massive number of original random scenes are generated, and then the number of scenes is reduced by the probability distance criterion to retain the original distribution characteristics.

[0023] S300: Based on the deterministic scheduling model and fuzzy sets, the deterministic scheduling model is reconstructed into a multi-stage sub-Blubar optimization model. The multi-stage sub-Blubar optimization model is solved using a decomposition iterative algorithm to obtain the day-ahead optimal scheduling scheme of the active distribution network, thereby realizing the optimal scheduling of the active distribution network.

[0024] In one optional embodiment, the decomposition iterative algorithm can be the Benders cut plane algorithm, which decomposes the multi-stage distributed robust optimization into a main problem and sub-problems; the main problem solves the basic day-ahead scheduling scheme and outputs decision variables; the sub-problems search for the worst-case scenario in the fuzzy set, solve the real-time correction cost, and generate Benders cut constraints to feed back to the main problem; the main problem is re-optimized with the addition of cut constraints, and the process is iterated until the difference between the objectives of the main problem and the sub-problems is less than the convergence threshold, and the optimal scheduling scheme is output. In another alternative embodiment, the decomposition iterative algorithm can also be the alternating direction multiplier method, which introduces consistency variables to decompose the global scheduling model, decouples the network, load, storage, and uncertainty-related sub-problems for parallel computation; after each sub-problem is solved individually, variable information is exchanged, and multipliers and consistency constraints are updated; each sub-problem is solved alternately and the global variables are updated repeatedly, and the global optimal scheduling solution is obtained after the convergence condition is met.

[0025] In this embodiment of the invention, the decomposition iterative algorithm includes the column sum constraint generation (CCG) algorithm; Specifically, the original two-layer max-min model is split into a main problem and sub-problems. The main problem solves first-order decisions such as day-ahead network reconstruction and energy storage planning. The sub-problems solve the inner-layer optimization of each scenario in parallel, searching for the worst-case probability distribution within the fuzzy set. The scenario constraints and auxiliary variables generated by the sub-problems are added to the main problem for iterative solution, repeating the loop until convergence, and outputting a complete scheduling scheme.

[0026] It should be noted that this invention constructs a grid-load-storage collaborative deterministic scheduling model, then accurately characterizes the high-dimensional stochastic correlation between wind and solar power through multivariate correlation distribution, builds a probabilistic fuzzy set, and solves the problem through multi-stage distributed robust optimization iteration. This approach fully leverages the potential of network reconfiguration, electric vehicle energy storage, and real-time electricity price adjustment, reduces scheduling costs and curtailment rates, improves voltage quality, and balances scheduling robustness and economy, with computational efficiency superior to traditional optimization methods.

[0027] In this embodiment of the invention, step S100 includes the following sub-steps A1-A3; In A1: The objective function is to minimize the total scheduling cost of the active distribution network; In A2: the constraints must include at least grid-side regulation constraints, load-side regulation constraints, energy storage-side regulation constraints, distributed generation operation constraints, and grid power flow and security constraints; In A3: Construct a deterministic scheduling model based on the joint constraints of the objective function.

[0028] In this embodiment of the invention, for an active power distribution system, the invention coordinates dynamic network reconfiguration on the grid side, demand response guided by electricity prices on the load side, and energy storage characteristics of electric vehicles on the energy storage side, with the goal of reducing the active power distribution network dispatch cost, to meet various dispatch requirements such as economy, stability, and user comfort.

[0029] The established objective function is expressed as: in, The cost of purchasing electricity from the upstream power grid for the distribution network. For switch operating costs, To reduce costs for new energy sources, For network loss costs, The cost of charging electric vehicles, and These are the sets of nodes that purchase electricity from the upper-level power grid, wind power, photovoltaic power, energy storage, AC power, and electric vehicle connections. The price of electricity purchased from the grid for distribution tiles, exist Time period The amount of electricity a node purchases from its upstream power grid. , and They are respectively in Time period Operating power of nodal wind power, photovoltaic power, and energy storage. and For the unit curtailment cost of wind and solar power, and Wind power and solar power respectively Time period The power curtailment of nodes, For the exchange area branch road The resistance, For the exchange area branch road exist Current during a period of time For time-of-use electricity pricing, In response to the real-time electricity price for renewable energy output, and To charge electric vehicles and absorb grid and wind and solar power.

[0030] Other active distribution network dispatching requirements include node voltage deviation, expressed as: in, Number the nodes. This refers to the number of nodes in the distribution network. For time period, The total number of scheduling periods is 24 hours. for Time period nodes voltage, This refers to the rated voltage of the distribution network node.

[0031] User comfort includes the fact that demand response can guide users to respond to renewable energy output, but after users participate in demand response, their own energy consumption curves change, reducing user satisfaction. To promote long-term relationships, user comfort is characterized by user electricity satisfaction and economic efficiency, expressed as: in, and , For demand response before and after Time period The electrical load of the node and the power of the electric vehicle, The comfort factor is set to 0.8. This is the economic efficiency coefficient, with a value of 0.2.

[0032] Constraints include grid-side regulation constraints. Network reconfiguration, as an important means of distribution network power flow optimization, optimizes network power distribution by changing the distribution network topology through switching. Ensuring a radial distribution network topology is a prerequisite for network reconfiguration; therefore, active distribution network reconfiguration must adhere to the following constraints: in, for Time-of-day branch The direction of the power flow, equal to 1 indicates a node. It is a node The parent node, for Time-of-day branch The connectivity status is indicated by 0 for disconnection and 1 for connectivity. This represents the total number of nodes in the distribution network. This represents the number of nodes, including substations. For the set of all nodes, For the set of substation nodes, This represents the maximum number of times all switches in the distribution network can operate during each time period.

[0033] Load-side regulation constraints include the fact that with the integration of distributed photovoltaic power and electric vehicles, new renewable energy output and peak loads are generated, making it difficult for traditional price-based demand response to achieve real-time dynamic matching of source and load, resulting in a significant decrease in smooth regulation capability. This invention proposes a dynamic pricing mechanism based on real-time net load status adjustment and matching of renewable energy generation with load. When the net load is less than zero, the price is appropriately reduced. This price signal guides load participation in demand response, promoting peak-shifting energy use, optimizing grid flow, and reducing charging costs. This can be expressed as: in, It is a standard time-of-use electricity price; For dynamic real-time electricity pricing; Expected contribution to new energy forecasting; This represents the allowable fluctuation ratio of load response power. The upper and lower limits allowed for dynamic electricity pricing; This is the demand response coefficient of the load to changes in electricity prices, i.e., the price elasticity of electricity demand. According to economic theory, user electricity consumption behavior is related to the electricity price at the current time. The larger the value, the greater the impact of electricity prices on users' electricity consumption behavior. Its positive or negative sign represents the positive or negative correlation between price changes and demand changes.

[0034] Energy storage-side regulation constraints include, to ensure highly self-consistent operation of the AC / DC distribution network, considering not only single-battery energy storage on the energy storage side but also large-scale distributed mobile energy storage resources (electric vehicles), utilizing electric vehicle V2G technology to achieve bidirectional power regulation, effectively serving the energy flow of the distribution network, and improving node voltage fluctuations and network losses. The energy storage-side battery energy storage-electric vehicle virtual energy storage model is expressed as follows: in, State of charge (SOC) for battery energy storage or virtual energy storage in electric vehicles. and These refer to the charging and discharging efficiencies of battery energy storage. and This refers to the charging and discharging states of energy storage. and These represent the upper limits of the charging and discharging power of energy storage.

[0035] Distributed power generation operation constraints include, within the active distribution network, the active-reactive power constraints of photovoltaic power connected to the grid via inverters need to be constructed by combining inverter capacity characteristics and photovoltaic power prediction, as follows: in, and These are the upper and lower limits of the capacity for distributed photovoltaic power. For nodes Distributed photovoltaic grid-connected capacity, and These represent the active and reactive power outputs of distributed photovoltaic systems, respectively. This provides information on active power prediction for distributed photovoltaic systems. and These represent the upper and lower limits of reactive power output from photovoltaic systems. The grid connection and output constraints for distributed wind turbines are similar to those for photovoltaic systems.

[0036] Power flow in the power grid includes the second-order cone relaxation power flow constraint of the AC network, which is expressed as: in, and They are time points From node Flow to Node Active and reactive power, and The lines are respectively The resistance, reactance, and square of the current, and They are time points From node The active and reactive input power, and They are nodes and nodes The square of the voltage, They are nodes The active power output or load of the main grid, wind power, photovoltaic power, energy storage, electrical load, and electric vehicles. They are nodes The reactive power output or load of the main grid, wind power, photovoltaic, energy storage, electrical load, and electric vehicles.

[0037] Safety constraints, including node voltage, branch current, and line capacity constraints, are expressed as follows: in, and They are nodes The upper and lower limits of the voltage amplitude at that location. For the line Upper limit of current amplitude For the line Maximum transmitted active power.

[0038] In this embodiment of the invention, step S200 includes the following sub-steps B1-B3; In B1: Obtain the output data of each distributed power source during the historical operating cycle, and construct the historical output sample set corresponding to each power source; In B2: For each distributed power source, the output probability density function of each distributed power source is fitted using nonparametric estimation based on the historical output sample set; In B3: The output probability density function of each distributed power source is used as the marginal probability density function of the corresponding distributed power source.

[0039] In one optional embodiment, the nonparametric estimation can be a local polynomial estimation method, which involves selecting all historical power output samples of the distributed power source and setting a local neighborhood bandwidth; taking each power output estimation point as the center to extract local samples and constructing a low-order polynomial to fit the local sample distribution; solving the polynomial coefficients by weighted least squares to obtain the local probability density of that point; and traversing all estimation points and stitching together all local fitting results to form a complete power output probability density curve. In another alternative embodiment, the nonparametric estimation can also be a nearest neighbor estimation. Given a fixed number of nearest neighbor samples K, the historical data set of wind and solar power output is read; for any power output estimate, the K nearest samples to the value are selected from the historical samples; the width of the data interval covered by the K nearest neighbor samples is calculated; the probability density of the current point is characterized by the ratio of the number of nearest neighbors to the interval width, and the complete density distribution is calculated point by point. In this embodiment of the invention, nonparametric estimation includes nonparametric kernel density estimation; Specifically, considering the randomness and volatility of distributed wind and solar power, a nonparametric kernel density estimation method that does not require a pre-defined distribution form is adopted. The density function is directly estimated based on historical power output samples, expressed as: in, For the output estimation point of distributed power sources, For the first A historical example of contribution, For kernel function, For the number of historical samples, This represents the bandwidth coefficient.

[0040] In this embodiment of the invention, after completing steps B1-B3, step S200 also includes steps B4-B6. In B4: Perform a probability integral transformation on the historical output sequence of each distributed power source to convert each historical output sequence into a transformed random variable that follows a standard uniform distribution; In B5: Based on the transformed random variables, a joint distribution model among the outputs of multiple distributed power sources is constructed using a multivariate correlation distribution model; In B6: The nonlinear correlation of the output of multiple distributed power sources is captured by the joint distribution model, and a multidimensional joint probability distribution is generated.

[0041] In one optional embodiment, the multivariate correlation distribution model can be a Gaussian Copula model. Marginal distribution fitting is performed on the historical power output data of each distributed wind and solar unit, transforming the original power output data into uniform variables in the range of 0 to 1. The linear correlation coefficients between the variables after uniform transformation of all wind and solar units are calculated, generating a correlation coefficient matrix. A high-dimensional Gaussian Copula correlation model is constructed based on the correlation coefficient matrix to characterize the synchronous change trend of power output among each unit. Combining the marginal distributions of each unit, a high-dimensional joint distribution of the overall wind and solar power output of multiple units is obtained. In an optional embodiment, the multivariate correlation distribution model can also be an elliptic t-Copula model, using nonparametric methods to fit the edge distribution of single-unit wind and solar power output, completing the homogenization transformation of the original power output data; setting degree-of-freedom parameters, statistically analyzing the correlation of the transformed samples of each unit, and constructing a high-dimensional t-Copula correlation structure; using... The distribution characteristics depict the strong tail correlation features under extreme fluctuations in wind and solar power output; by combining the marginal distribution with the t-Copula correlation structure, a multi-unit joint power output distribution that takes into account both normal and extreme operating conditions is generated. In this embodiment of the invention, the multivariate association distribution model includes the Copula model; Specifically, the historical output sequence of each distributed power source is represented by a probability integral transform as follows: in, For the number of uncertain variables, For the first The marginal cumulative distribution function of each variable. Let be the transformed random variable, which follows the... The standard uniform distribution; The joint density function of multidimensional uncertain variables based on Copula can be expressed as: in, For the first An uncertain variable, For the first The marginal probability density function of each variable. For the edge The corresponding bivariate Copula probability density function, and Variables and Given a set of conditions The conditional probability integral transform value is used to characterize the correlation between different variables.

[0042] It should be noted that, considering the spatial distribution characteristics of wind and solar power stations in the distribution network, which usually exhibit an approximately symmetrical mesh interconnection relationship, the traditional multivariate Copula model is difficult to characterize the nonlinear and asymmetric correlation between the outputs of multiple distributed power sources. The use of the Teng Copula model to construct a high-dimensional joint output distribution model of multiple distributed wind and solar power sources can better characterize the nonlinear correlation between the outputs of multiple distributed power sources.

[0043] In this embodiment of the invention, after completing steps B4-B6, step S200 also includes steps B7-B9; In B7: Based on the multidimensional joint probability distribution, the Latin hypercube stratified sampling method is used to generate an initial scenario set that takes into account the output correlation of multiple distributed power sources; In B8: The initial scene set is reduced to obtain the reduced typical scene set and the initial probability distribution corresponding to each typical scene; In B9: Using the initial probability distribution as the center, a fuzzy set of the scene probability distribution is constructed using the comprehensive norm constraint.

[0044] In this embodiment of the invention, based on the constructed high-dimensional joint distribution model, the Latin Hypercube (LHS) is used to generate random scenarios that consider the output correlation of multidimensional distributed power sources. LHS avoids sample clustering through hierarchical sampling, covering the entire distribution range of variables with fewer samples. To address the computational burden caused by an excessive number of sampled scenarios, a scenario reduction method based on probabilistic distance is adopted, reducing the size of the scenario set while preserving the probability distribution characteristics of the original scenarios to the greatest extent possible.

[0045] However, considering the uncertainty of the discrete values ​​in each reduction scenario, to ensure that the scenario probability values ​​fluctuate within a reasonable range, a comprehensive norm constraint centered on the aforementioned initial probability distribution is constructed to restrict the probability distribution of uncertain scenarios, making it closer to the real scenario. The comprehensive norm constraint includes the 1-norm and the ∞-norm, as shown below: in, As a probability measure, and These represent the allowable deviation values ​​of the probability distribution under the comprehensive norm constraint. For discrete scene probability distribution, Predict probability distributions for discrete scenarios.

[0046] The allowable deviation of the probability distribution under the comprehensive norm constraint depends on the confidence level that the probability distribution must satisfy. The transformation process is shown below: in, and These represent the confidence levels of uncertainty probability corresponding to the comprehensive norm constraint.

[0047] In summary, the confidence sets for each discrete scenario based on the probability distribution are shown below: It should be noted that by using LHS stratified sampling combined with probabilistic distance scene reduction to reduce computational load, and then combining binorm constraints to define probability confidence intervals, the system balances the realism of scene distribution with model robustness, and the scheduling economy with security.

[0048] In this embodiment of the invention, step S300 includes the following sub-steps C1-C5; In C1: The deterministic scheduling model is reconstructed into a two-stage sub-Brussels bar optimization model; In C2: The first phase determines the day-ahead scheduling plan based on the predicted output information and load forecast information of distributed power sources; In C3: The second phase is based on the day-ahead scheduling plan determined in the first phase. The worst probability distribution scenario is obtained in the fuzzy set. Under the worst probability distribution scenario, the demand response load and the power curtailment of new energy are adjusted in real time. In C4: The objective function of the two-stage sub-Bruker optimization model is to minimize the sum of the decision cost in the first stage and the expected cost in the second stage under the worst-case probability distribution; In C5: the feasible region of the first-stage decision variables consists of constraints that are only related to the first stage, and the feasible region of the second-stage decision variables consists of second-stage constraints that are related to the first-stage variables and the scenario.

[0049] In this embodiment of the invention, the deterministic scheduling model is reconstructed into a two-stage distributed robust scheduling model. The first stage relies on wind and solar power and load forecasting information to determine the day-ahead decision variables of the main distribution network interaction power and energy storage units. The second stage adjusts the distribution network scheduling plan based on the decision variables of the first stage according to the actual wind and solar power processing. The specific model is shown below: in, For the first stage variables, For the second stage variables, For the k-th scene, These are the corresponding cost coefficients. For the first The probability of each scenario occurring The number of discrete scenes.

[0050] Given a set Time optimization variables feasible domain, Constraints that apply only to the first phase. This is the second phase of constraints.

[0051] In this embodiment of the invention, after completing steps C1-C5, step S300 further includes steps C6-C9; In C6: The column and constraint generation algorithm is used to decouple the two-stage bibar optimization model from the main problem and sub-problems for iterative solution; In C7: the main problem is a single-layer optimization problem given a finite number of adverse probability distributions; In C8: The subproblem is to find the worst-case scenario probability distribution within the fuzzy set based on the first-stage decision variables obtained from the main problem; In C9: Iteratively solve the main problem and subproblems until the upper and lower bounds meet the convergence accuracy, and output the first-stage decision variables as the day-ahead optimization scheduling scheme for the active distribution network.

[0052] In this embodiment of the invention, the main problem is expressed as: The subproblem is represented as: in, Indicates the number of iterations. This represents the total cost for the first phase, i.e., the planned cost for the current day. As an auxiliary variable, For the first In the next iteration, the scene The probability value, For the first In the next iteration, the scene The second-stage optimal decision variables For the first In the next iteration, the scene The random parameters.

[0053] The meaning of subproblems is given by the main problem. Based on this, we find the worst-case scenario probability distribution within the confidence interval to provide iterative computation for the main problem and solve for the lower bound of the model. Although the subproblem is a difficult-to-solve max-min two-level optimization problem, the min problems under each scenario are independent. We can use parallel computation to solve the inner min problem simultaneously. Therefore, the subproblems can be solved sequentially, from inner to outer, as shown below: in, For a given and scene The optimal value of the inner minimization problem. This represents the probability value of the scenario.

[0054] It should be noted that two-phase partial Bruker optimization is typically based on the idea of ​​zero-sum game and uses the column sum constraint generation (CCG) algorithm or the cutting plane algorithm to solve the problem, decoupling the original problem into a main problem and subproblems for iterative processing. Compared with traditional solution algorithms, the CCG algorithm continuously introduces auxiliary variables and constraints related to the subproblems during the process of solving the main problem, thus accelerating the convergence speed.

[0055] Example 2, refer to Table 1, Figures 2-6 Based on the above embodiments, this embodiment provides an application simulation test of an active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination, in order to verify the feasibility and advantages of the present invention.

[0056] by Figure 2 The improved IEEE 33-node model shown is used as the test object to verify the effectiveness of the proposed model and solution algorithm. This active distribution network can purchase electricity from the upstream grid, i.e., the main grid. The demand side includes conventional electrical loads and electric vehicle charging loads. Typical daily predicted output data for distributed wind power (WT1, WT2, WT3) and photovoltaic (PV1, PV2, PV3, PV4) are as follows: Figure 3 As shown. Conventional electrical load and electric vehicle charging load are as follows: Figure 4 As shown.

[0057] To accurately quantify the stochastic spatiotemporal coupling uncertainty of distributed wind and solar power output within an active distribution network, and to avoid the biases of traditional parametric distribution assumptions and the overly conservative nature of single robust optimization, this paper proposes a method based on nonparametric kernel density estimation, high-dimensional correlation modeling using the Copula framework, and Latin hypercube sampling and scene reduction techniques. This method generates wind and solar power output scenarios and constructs data-driven fuzzy sets of scenarios. The superiority of the proposed method is verified by comparing different scenario generation methods. Details are as follows: Method 1: The scene generation method proposed in this invention.

[0058] Method 2: Beta and Weibull distribution fitting of photovoltaic and wind power output + Gaussian Copula correlation modeling + Monte Carlo sampling; Method 3: KDE edge fitting + R-vine Copula correlation modeling + Monte Carlo sampling; Method 4: KDE edge fitting + C-Vine Copula correlation modeling + LHS sampling.

[0059] To accurately quantify the superiority of this invention, the KS test statistic is used. Kendall correlation matrix mean error , Time series fitting error The evaluation indicators are expressed as follows: in, These are the cumulative distribution functions obtained from measured historical data and scene generation methods, respectively. and These represent the output values ​​of distributed photovoltaic (PV) and wind power, respectively. This represents the total number of distributed stations. These are the measured historical data and the scene generation method mid-terms. and station Kendall rank correlation coefficient between them For scheduling periods, The number of typical scenes generated. For the first In each generated scene Time of day station The output value, For actual measurement of typical day and night Time of day station The actual output value.

[0060] Table 1 shows the evaluation metrics obtained based on different scenario generation methods. The distributed power output data generated by the scenario generation method proposed in this invention is as follows: Figure 5 As shown.

[0061] Table 1 Comparison of Statistical Fitting Accuracy of Different Scene Generation Methods

[0062] As shown in Table 1, the scene generation method proposed in this invention has the smallest KS test statistic, the average error of the Kendall correlation matrix, and the time series fitting error. This indicates that while overcoming the shortcomings of the parameterized distributed hypothesis bias, the method of this invention has higher accuracy in characterizing the nonlinearity and asymmetric spatiotemporal correlation of high-dimensional wind and solar power output, higher scene time series restoration accuracy, and a closer approximation of the probability distribution. The actual situation.

[0063] To further and more intuitively verify the ability of each method to characterize the correlation of output from multiple distributed power sources, a Kendall correlation coefficient heatmap of historical data and the scenarios generated by each method was plotted.

[0064] like Figure 6As shown, by comparing the heatmaps obtained from data generated in different scenarios, it can be found that the scenario generation method proposed in this invention has the highest accuracy in characterizing correlations, with the correlation coefficients deviating from the original data by less than 0.01, fully reproducing the full-dimensional correlation characteristics of wind and solar power output. In contrast, the scenario generation method based on the traditional Gaussian Copula has the most significant deviation, not only greatly underestimating the positive correlation strength between photovoltaics but also weakening the negative correlation between wind and solar power output, failing to characterize the nonlinear and asymmetric correlation characteristics of wind and solar power output. The accuracy of the Copula-based scenario generation method is between the two. Although the overall trend is consistent with the original data, it is difficult to adapt to the network correlation structure of wind and solar power output due to the limitations of the Copula structure, and the correlation coefficients still have some deviation.

[0065] Example 3, refer to Tables 2-5, Figures 7-10 To verify the advantages of electric vehicle virtual energy storage, real-time electricity pricing, and dynamic network reconfiguration in terms of active distribution network operating costs, renewable energy consumption, voltage deviation, and user satisfaction, four different dispatch schemes were adopted, denoted as Scheme 1 to Scheme 4, as shown below: Option 1: Active distribution network operation considering only single battery energy storage, traditional load demand response, and static network reconfiguration; Option 2: Based on Option 1, consider the participation of electric vehicle mobile energy storage in the active distribution network operation; Option 3: Based on Option 2, consider the participation of real-time electricity price-guided load demand response in active distribution network operation; Option 4: Based on Option 3, consider dynamic network reconfiguration to participate in active distribution network operation.

[0066] Scheme 1 only includes battery energy storage, traditional demand response, and static network reconfiguration functions, where demand response can be equated with energy storage in a broad sense. During periods of high renewable energy generation and low load, energy is charged. Conversely, during periods of low renewable energy generation and high load, energy is discharged to meet real-time load balancing, thus playing a crucial role in renewable energy consumption and reducing the operating costs of active distribution networks. However, with the emerging load of electric vehicles connecting to active distribution networks, their mobile energy storage characteristics cannot be ignored. Simultaneously, facing the high penetration rate of distributed energy access, traditional static electricity pricing cannot smoothly adjust the load curve. Scheme 3, with its real-time electricity price based on net load status, can further encourage electric vehicles to charge during periods of low net load, and static network reconfiguration cannot provide a flexible topology for highly uncertain distributed power sources. The results of active distribution network optimization scheduling for different schemes are shown in Table 2.

[0067] Table 2 Results of Active Distribution Network Optimization Scheduling under Different Schemes

[0068] Table 2 shows the scheduling results of schemes 1 to 4, including total cost, electricity purchase cost, renewable energy curtailment cost, network loss cost, switching operation cost, and electric vehicle charging cost. Compared with scheme 1, scheme 2 considers the energy storage characteristics of electric vehicles, reducing the total cost and renewable energy curtailment rate by 23.51% and 0.89%, respectively. This indicates that utilizing the energy storage characteristics of electric vehicles has a positive impact on improving grid economy and renewable energy absorption, but user satisfaction actually decreased by 10.49%. Compared with scheme 2, scheme 3, which introduces real-time electricity pricing, has a highly significant effect on improving user satisfaction because it considers the economic benefits for users, allowing them to profit from discharging electricity, thereby increasing user satisfaction. It also achieves temporal equalization of grid load and distributed power sources, optimizing power transmission and power in the network. Scheme 4, which combines multi-side and multi-regulation technologies, reduces network loss cost and voltage deviation by 17.32% and 35.56%, respectively, compared with scheme 1, demonstrating the effectiveness of the proposed optimized scheduling method in meeting various scheduling needs such as distribution network economy, greenness, stability, and user comfort.

[0069] Figure 7 The charging price for electric vehicles, taking into account net load conditions, is compared to the regular time-of-use electricity price. Figure 8 The display shows the offset of different electrical loads. The load offset is the difference between the actual load after scheduling and the initial load. A value greater than zero indicates that the load has been moved upward or increased charging, while a value less than zero indicates that the load has been moved downward or discharged.

[0070] From Figure 7 and Figure 8 It can be seen that during the periods of 8:00-11:00 and 18:00-21:00 when the net load is greater than zero, the real-time electricity price and the time-of-use electricity price are basically the same. Among them, Scheme 4, based on the charging and discharging characteristics of electric vehicle energy storage, can discharge during the peak electricity price period of 19:00-21:00, reducing the charging cost of electric vehicles. During the period of 12:00-16:00 when the net load is less than zero, the real-time electricity price is lower than the time-of-use electricity price, and electric vehicles tend to increase their charging volume during this period. Finally, from the net load curves before and after scheduling, it can be seen that the scheduling method of this invention has a significant impact on reducing the peak-valley difference of net load, reducing it by 19.32%.

[0071] like Figure 9 The diagram illustrates the power balance of the distribution network under Scheme 4. In this diagram, a value greater than 0 for electric vehicles and energy storage represents discharging, and a value less than 0 represents charging. (Combined) Figure 7It can be seen that during the periods of high distributed power generation (1:00-4:00 AM and 12:00-4:00 PM), when the net load is underestimated due to large-scale distributed power generation, the electricity load demand is mainly met by wind and solar power generation. The surplus renewable energy is consumed by electric vehicles with energy storage characteristics and energy storage devices. During the peak load period (7:00-9:00 PM), electric vehicles discharge in conjunction with energy storage devices, and the shortfall is supplemented by electricity purchased from the upstream grid, reducing the grid's electricity purchase cost. However, at the end of the dispatch period, in order to ensure the continuity of the next dispatch period, the distribution network still needs to purchase some electricity from the upstream grid to meet the SOC level of the energy storage devices.

[0072] With the continuous growth in the number of electric vehicles, based on the scheduling plan of Scheme 4 (where the number of electric vehicles is 400), variables ranging from 200 to 600 are set to compare the distribution network optimization scheduling results under different electric vehicle penetration rates. Figure 10 As shown.

[0073] Depend on Figure 10 It is evident that the total cost of power grid dispatching increases with the continuous increase in the number of electric vehicles. This is because, on the one hand, the demand for power grid load is constantly growing. Besides the limited output of distributed power sources, the distribution network must purchase electricity from the upper-level grid to meet the power supply and demand balance, leading to a continuous increase in electricity purchase costs. The increase in the number of vehicles also leads to increased charging costs. However, relying on the energy storage characteristics of electric vehicles, sufficient adjustment space can be provided for the grid to absorb new energy sources, reducing the new energy curtailment rate from 9.28% to 5.08%. At the same time, the dispatchability of electric vehicles optimizes grid power flow and reduces node voltage deviation. However, when the number of vehicles increases from 400 to 600, excessive load access leads to line overload, and node voltages are generally low. Therefore, the grid voltage deviation exhibits a V-shaped change with the increase in the number of electric vehicles.

[0074] The proposed data-driven distributed robust optimization model is based on a flexible and adjustable framework. To verify the impact of different data-driven DRO parameters on the distribution network optimization scheduling results, the uncertainty probability confidence levels corresponding to the comprehensive norm constraint were set to 0.5 and 0.99, respectively. The effects of the number of Latin hypercube samples and the number of reduced typical scenarios on the robustness of the DRO model were studied, and the results are shown in Tables 3 and 4.

[0075] Table 3. Distribution network optimization scheduling results under different Latin hypercube sampling sizes

[0076] Table 4. Distribution network optimization scheduling results under different typical scenario numbers

[0077] As shown in Tables 3 and 4, the number of sampled samples and the number of typical scenarios jointly influence the scheduling results of the data-driven DRO model. On the one hand, with the increase in the number of sampled samples, the allowable deviation of the distributed power generation output probability distribution gradually decreases, the fuzzy set of scenarios more closely matches the real distribution, the excessive conservatism of the model is alleviated, the economic efficiency of the scheduling scheme increases significantly, the adaptability to wind and solar power output fluctuations is stronger, the node voltage deviation gradually decreases, and user satisfaction increases slightly. On the other hand, with the increase in the number of typical scenarios, the probability of extreme data in the original large number of samples becoming a typical scenario increases. Therefore, in order to cover many extreme typical scenarios, the scheduling cost of the grid scheduling scheme increases with the increase in the number of typical scenarios, the node voltage deviation also increases slightly, and user satisfaction decreases slightly. In summary, increasing the number of sampled samples can effectively alleviate the excessive conservatism of the DRO model and achieve simultaneous improvement in economic efficiency and power quality; while increasing the number of typical scenarios can improve the completeness of scenario characterization, it will lead to increased conservatism and decreased economic efficiency in the scheduling scheme. In practical applications, it is necessary to select an appropriate parameter combination based on computational efficiency and scheduling performance.

[0078] The data-driven DRO model that considers the multidimensional output correlation of distributed power sources is compared with the data-driven DRO model that does not consider the output correlation of distributed power sources, the traditional deterministic optimization model, the two-stage robust optimization model, and the stochastic optimization model. The results of this comparative analysis are detailed in Table 5.

[0079] Table 5. Results of Active Distribution Network Dispatch under Different Optimization Methods

[0080] Table 5 shows that the deterministic optimization DO model, which schedules based solely on distributed generation (DG) prediction information, offers the best economic benefits but lacks robustness due to neglecting wind and solar uncertainties, resulting in significant operational risks. The robust optimization model considers the most extreme DG output scenarios, exhibiting the strongest robustness, but its overly conservative scheduling reduces economic efficiency, while extreme DG output leads to the largest voltage deviation at grid nodes. The stochastic optimization model, considering scheduling results under all possible scenarios, offers higher economic benefits despite its poor robustness, but its economic efficiency significantly decreases in severe scenarios. The data-driven DRO model combines the advantages of RO and SO models, employing a comprehensive norm constraint to construct a fuzzy set of scenario probability distributions, achieving a balance between economic efficiency and robustness in day-ahead grid scheduling. This invention's method describes the correlation of high-dimensional DG output using Copula theory, constructing an uncertainty fuzzy set that better reflects the actual distribution, thus considering the high-dimensional DG output correlations neglected by traditional DRO models and overcoming the shortcomings of traditional optimization methods. Furthermore, since the data-driven DRO model does not require dual processing in its subproblem handling compared to the RO model, its parallel computing method results in a shorter computation time. The SO model, on the other hand, needs to consider scheduling results under all possible scenarios, and its computation time is the longest. In summary, the data-driven DRO proposed in this invention is an ideal method for handling the output uncertainty of high-dimensional distributed power sources.

[0081] Example 4 illustrates the schematic scheme of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination in this embodiment. It should be noted that the technical solution of this active distribution network distributed robust optimization scheduling method system considering grid-load-storage coordination belongs to the same concept as the above-described active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination. Details not described in detail in this embodiment can be found in the description of the above-described active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination.

[0082] This embodiment of the active distribution network distributed robust optimization scheduling method system that considers grid-load-storage coordination includes: The model building module is used to acquire the operating data of the active distribution network, construct the objective function and constraints based on the operating cost, and establish a deterministic scheduling model based on the objective function and constraints. The fuzzy set construction module is used to fit the marginal distribution of the output of each distributed power source and characterize the nonlinear correlation between the outputs of multiple power sources. It generates an initial scene set through hierarchical sampling and scene reduction, and constructs a fuzzy set of scene probability distribution based on the initial scene set. The model solving module is used to reconstruct the deterministic scheduling model into a multi-stage sub-Blubar optimization model based on the deterministic scheduling model and fuzzy sets, and solve the multi-stage sub-Blubar optimization model using a decomposition iterative algorithm to obtain the day-ahead optimal scheduling scheme of the active distribution network, thereby realizing the optimal scheduling of the active distribution network.

[0083] This embodiment also provides a computer device applicable to the active distribution network distributed robust optimization scheduling method that considers grid-load-storage coordination, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement the active distribution network distributed robust optimization scheduling method that considers grid-load-storage coordination, as proposed in the above embodiments.

[0084] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination as proposed in the above embodiments.

[0085] The storage medium proposed in this embodiment and the active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0086] Based on the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computing device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0087] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A robust distributed optimization scheduling method for active distribution networks considering grid-load-storage coordination, characterized in that, include: Obtain operational data of the active distribution network, construct an objective function and constraints based on the operational costs, and establish a deterministic scheduling model based on the objective function and constraints; Marginal distribution fitting is performed on the output of each distributed power source, and the nonlinear correlation between the outputs of multiple power sources is characterized. An initial scene set is generated through hierarchical sampling and scene reduction. A fuzzy set of scene probability distribution is constructed based on the initial scene set. Based on the deterministic scheduling model and fuzzy sets, the deterministic scheduling model is reconstructed into a multi-stage sub-Blubar optimization model. The multi-stage sub-Blubar optimization model is solved using a decomposition iterative algorithm to obtain the day-ahead optimal scheduling scheme for the active distribution network, thereby realizing the optimal scheduling of the active distribution network.

2. The active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination as described in claim 1, characterized in that, The deterministic scheduling model established based on the objective function and constraints includes: The objective function is to minimize the total scheduling cost of the active distribution network. The constraints must include at least grid-side regulation constraints, load-side regulation constraints, energy storage-side regulation constraints, distributed generation operation constraints, and grid power flow and security constraints. A deterministic scheduling model is constructed based on the joint constraints of the objective function.

3. The active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination as described in claim 1 or 2, characterized in that, Marginal distribution fitting for the output of each distributed power source includes: Obtain the output data of each distributed power source during its historical operating cycle, and construct a historical output sample set for each power source. For each distributed power source, the output probability density function of each distributed power source is fitted using nonparametric estimation based on the historical output sample set; The output probability density function of each distributed power source is used as the marginal probability density function of the corresponding distributed power source.

4. The active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination as described in claim 3, characterized in that, Characterizing the nonlinear correlation between the outputs of multiple power sources includes: Perform a probability integral transformation on the historical output sequence of each distributed power source to convert each historical output sequence into a transformed random variable that follows a standard uniform distribution; Based on the transformed random variables, a joint distribution model among the outputs of multiple distributed power sources is constructed using a multivariate correlation distribution model. By capturing the nonlinear correlation of the output of multiple distributed power sources through a joint distribution model, a multidimensional joint probability distribution is generated.

5. The active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination as described in claim 4, characterized in that, An initial scene set is generated through stratified sampling and scene reduction. A fuzzy set of scene probability distributions is then constructed based on this initial scene set, including: Based on the multidimensional joint probability distribution, the Latin hypercube stratified sampling method is used to generate an initial scenario set that takes into account the output correlation of multiple distributed power sources. The initial scene set is reduced to obtain the reduced typical scene set and the initial probability distribution corresponding to each typical scene. Using the initial probability distribution as the center, a fuzzy set of the scene probability distribution is constructed by utilizing the comprehensive norm constraint.

6. The active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination as described in claim 5, characterized in that, The deterministic scheduling model is reconstructed into a multi-stage sub-Brussels bar optimization model, including: The deterministic scheduling model is reconstructed into a two-stage sub-Brussels bar optimization model; The first phase determines the day-ahead scheduling plan based on the predicted output information and load forecast information of distributed power sources. The second phase is based on the day-ahead scheduling plan determined in the first phase. The worst-case probability distribution scenario is obtained in the fuzzy set. Under the worst-case probability distribution scenario, the demand response load and the power curtailment of new energy are adjusted in real time. The objective function of the two-stage sub-Bruker optimization model is to minimize the sum of the decision cost in the first stage and the expected cost in the second stage under the worst-case probability distribution. The feasible region of the decision variables in the first stage consists of constraints that are only related to the first stage, while the feasible region of the decision variables in the second stage consists of second-stage constraints that are related to the variables in the first stage and the scenario.

7. The active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination as described in claim 6, characterized in that, Solving the multi-stage sub-Bruker optimization model using a decomposition iterative algorithm yields the following day-ahead optimal scheduling schemes for the active distribution network: The two-stage bibar optimization model is decoupled from the main problem and sub-problems by using a column and constraint generation algorithm for iterative solution. The main problem is a single-layer optimization problem given a finite number of adverse probability distributions; The subproblem is to find the worst-case scenario probability distribution within the fuzzy set based on the first-stage decision variables obtained from the main problem; The main problem and subproblems are solved iteratively until the upper and lower bounds meet the convergence accuracy. The first-stage decision variables are then output as the day-ahead optimization scheduling scheme for the active distribution network.

8. A distributed robust optimization scheduling method system for active distribution networks considering grid-load-storage coordination, comprising applying the distributed robust optimization scheduling method for active distribution networks considering grid-load-storage coordination as described in any one of claims 1-7, characterized in that, include: The model building module is used to acquire the operating data of the active distribution network, construct the objective function and constraints based on the operating cost, and establish a deterministic scheduling model based on the objective function and constraints. The fuzzy set construction module is used to fit the marginal distribution of the output of each distributed power source and characterize the nonlinear correlation between the outputs of multiple power sources. It generates an initial scene set through hierarchical sampling and scene reduction, and constructs a fuzzy set of scene probability distribution based on the initial scene set. The model solving module is used to reconstruct the deterministic scheduling model into a multi-stage sub-Blubar optimization model based on the deterministic scheduling model and fuzzy sets, and solve the multi-stage sub-Blubar optimization model using a decomposition iterative algorithm to obtain the day-ahead optimal scheduling scheme of the active distribution network, thereby realizing the optimal scheduling of the active distribution network.

9. A computer device, characterized in that, include: A memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the active distribution network distributed robust optimization scheduling method considering grid-load-storage coordination as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the steps of the active distribution network distributed robust optimization scheduling method that takes into account grid-load-storage coordination as described in any one of claims 1 to 7.

Images