Effective reactive power reserve planning configuration method for high-proportion new energy power grid

Abstract

The invention discloses an effective reactive power reserve planning configuration method for a high-proportion new energy power grid. The method comprises the following steps: defining effective reactive power reserve of each electrical node of the high-proportion new energy power grid; the method comprises the following steps: constructing an uncertainty full scene set based on historical data of a new energy station, and introducing a random forest to combine with local meteorological data to enrich the scene set; obtaining a typical scene and probability distribution thereof through improved k-means clustering; establishing an effective reactive power reserve planning configuration model considering the uncertainty of the power system and power constraint conditions thereof, and converting nonlinear constraint into linear constraint through a strong second-order cone relaxation method; and solving the effective reactive power reserve planning configuration model through a Gurobi solver under a power constraint condition based on the complete version full scene set after the random forest is enriched, so as to obtain the optimal configuration of the effective reactive power reserve planning. According to the method, the random forest and the distributed robustness are combined to obtain the optimal configuration of the effective reactive power reserve planning of the high-proportion new energy power grid.

Description

Technical Field

[0001] The present invention belongs to the field of power system technology, and specifically relates to the technical field of research on reactive reserve planning and configuration of new-era power systems under large-scale new energy injection in electrical engineering, especially the technical field of planning and configuration optimization considering the dual uncertainty of new energy and load in new power systems and the effectiveness of reserves, and in particular to a method for planning and configuring effective reactive reserves in high-proportion new energy power grids. Background Art

[0002] In recent years, electricity load has grown rapidly. To meet the energy needs of high-quality development of productivity, renewable energy will play an increasingly important role in the energy supply system in the future. As renewable energy becomes increasingly important in the energy supply system, the planning of effective reactive power reserves for power grids with a high proportion of renewable energy faces severe challenges. However, the existing reactive power reserves in traditional power grids are insufficient to cope with the high volatility of new power systems with large-scale integration of renewable energy. Traditional theories and methods are difficult to adapt to the strong volatility and complexity of the new era power grid to ensure the stable operation of the power system after the large-scale injection of renewable energy. To ensure that the system has sufficient voltage stability after the injection of a large amount of renewable energy, the system should have sufficient effective reactive power reserves to support the stable output of large-scale renewable energy. Therefore, solving the planning and configuration problem of the maximum effective reactive power reserve of the renewable energy system is of great significance to the development of clean energy. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies and propose a method for planning and configuring effective reactive power reserves in a high-proportion renewable energy power grid. This invention combines random forests with distributed robustness to obtain the optimal configuration for planning effective reactive power reserves in a high-proportion renewable energy power grid.

[0004] The object of the present invention is achieved through the following technical solution: a method for planning and configuring effective reactive power reserves in a high-proportion new energy power grid, comprising the following steps:

[0005] (1) Define the effective reactive power reserve of each electrical node in a high-proportion renewable energy grid;

[0006] (2) Construct an uncertainty full scenario set based on historical data of new energy stations, and introduce random forest to enrich the scenario set in combination with local meteorological data to obtain a complete version of the full scenario set;

[0007] (3) Based on the complete set of scenes, typical scenes and their probability distribution are obtained by improving k-means clustering;

[0008] (4) Establish an effective reactive power reserve planning and configuration model considering power system uncertainty;

[0009] (5) Establish the power constraints of the effective reactive power reserve planning configuration model and transform the nonlinear constraints in the power constraints into linear constraints through the strong second-order cone relaxation method;

[0010] (6) Based on the complete scenario set enriched by random forest, the effective reactive power reserve planning configuration model is solved by Gurobi solver under the linearized power constraint condition to obtain the optimal configuration of effective reactive power reserve planning.

[0011] Furthermore, the step (1) specifically includes:

[0012] Select the maximum reactive power that can maintain the voltage of the electrical node within the safe range Minimum reactive power Maximum forward reactive output for all possible operating scenarios Negative maximum reactive output The smaller value in the middle is the effective reactive power reserve, that is, the maximum effective reactive power reserve in the positive direction is The maximum negative effective reactive power reserve is The effective reactive power reserve is directional, and the positive effective reactive power reserve and the negative effective reactive power reserve represent that the reactive power reserve devices of the electrical nodes are in the form of a parallel capacitor group and a parallel reactor group, respectively.

[0013] Furthermore, the step (2) specifically includes:

[0014] First, a full uncertainty scenario set is constructed based on the historical data of the new energy station. The full uncertainty scenario set includes the full uncertainty scenario set corresponding to the power source and the full uncertainty scenario set corresponding to the load in the new energy station. The full uncertainty scenario set corresponding to the power source in the new energy station is expressed as:

[0015]

[0016] In the formula, S represents the uncertainty scenario set corresponding to the power source in the new energy station, of which there are L scenarios in total, s l Indicates the lth scenario corresponding to the power supply, scenario s l There are M new energy stations in the system, i is the electrical node number, m is the new energy station number, P l m Represents scene s l The active power output of the new energy station m, Q l m Represents scene s l The reactive power output of the new energy station m in the new energy station is expressed as follows:

[0017]

[0018] Where Z represents the uncertainty scenario set corresponding to the load in the new energy station, of which there are L scenarios in total, z l Indicates the lth scenario corresponding to the load, scenario z l There are D new energy stations in the system, i is the number of the electrical node, d is the number of the new energy station, P l d Represents scene z l Active power output of new energy station d, Q l d Represents scene z l The reactive power output of the new energy station d;

[0019] Then, the original data full scene set X is obtained based on the uncertainty full scene set corresponding to the power source and the uncertainty full scene set corresponding to the load in the new energy station. Bootstrap sampling is performed on the original data full scene set X T times, and N samples are extracted each time to generate T subsets.

[0020] Then, T decision trees are constructed based on the T subsets to form a random forest. When constructing the t-th tree, the optimal split point is recursively selected, and the splitting criterion is to minimize the Gini impurity.

[0021] Secondly, the classification task is completed based on random forest combined with local meteorological data, which is specifically achieved through soft voting to obtain a set of predicted future scenarios;

[0022] Finally, the original data full-scene set corresponding to the historical data and the predicted future scene set are combined to form a complete full-scene set containing L+T scenes.

[0023] Furthermore, the step (3) specifically includes:

[0024] First, the k-means clustering method is improved by introducing the sum of squares of the range, where the sum of squares of the range is expressed as:

[0025]

[0026] Where, NSSE r Represents the improved r-th cluster C r The sum of squares of the range, c j Represents cluster C r The jth data in Represents cluster C r The average value of all data in, J represents the cluster C r The total number of data in c rmax Represents cluster C r The maximum value in c rmin Represents cluster C r The minimum value in ;

[0027] Then, the improved k-means clustering method is used to perform k-means clustering on the L+T scene sets in the complete version of the full scene set, obtaining R clusters. The elbow method and silhouette coefficient method in the k-means clustering method are combined to determine the number of typical scene clusters as A. From the R clusters, the first A clusters with the smallest sum of squared extreme differences are selected as typical scene clusters to obtain typical scenes.

[0028] Finally, the corresponding typical scene probability distribution is obtained by calculating the proportion of the number of scenes in each typical scene cluster in the total number of scenes.

[0029] Furthermore, the objective function of the effective reactive power reserve planning configuration model is:

[0030]

[0031] f1=C fix ·y i +C fc ·y i ·x i

[0032]

[0033] Where f1 represents the planning configuration cost function; f2 represents the distributed robustness cost; i is the electrical node number, I is the total number of electrical nodes; C fix represents the fixed cost of installing reactive power compensation equipment at node i, C fc Represents the unit capacity cost coefficient; y is a decision variable of 0 or 1, indicating whether to install reactive power compensation equipment; y i is a 0 or 1 decision variable for node i, indicating whether node i should install reactive power compensation equipment; x is a continuous variable indicating the maximum effective reactive power reserve capacity of the planned configuration; x i is a continuous variable, representing the maximum effective reactive reserve capacity of node i; l represents the lth scenario, L is the total number of scenarios; z l represents the reactive reserve unit capacity switching loss coefficient in the lth scenario, U(x) is z l The feasible domain of z il represents the reactive reserve unit capacity switching loss coefficient of node i in the lth scenario; T total Indicates the total number of moments in a single scene, T out,il represents the number of time points when the voltage of node i exceeds the limit in the lth scenario; f3(y,x,z l ) represents the cost function of reactive power reserve equipment operation loss; p l represents the probability of the occurrence of the lth scenario; f4 represents the voltage level deviation penalty function; represents the maximum amplitude of the voltage fluctuation at node i in the lth scenario, and v0 represents the voltage reference value.

[0034] Furthermore, the power constraints of the effective reactive power reserve planning configuration model specifically include:

[0035] ① Power flow constraint, its expression is:

[0036]

[0037] Where, P Gli represents the active power output of power node i in the lth scenario, Q Gli represents the reactive power output of power node i in the lth scenario, P Lli represents the active power of load node i in the lth scenario, Q Lli represents the reactive power of load node i in the lth scenario, V li represents the voltage amplitude of node i in the lth scenario, V lj represents the voltage amplitude of node j in the lth scenario, δ li represents the phase angle of node i in the lth scenario, δ lj represents the phase angle of node j in the lth scenario, G ij represents the real part of the admittance of the branch between node i and node j, B ij represents the imaginary part of the admittance of the branch between node i and node j, S B is the set of all nodes;

[0038] ② Branch power constraint, its expression is:

[0039] Q c li≤y i z il i∈S B

[0040] Q c li≥-y i z il i∈S B

[0041] Where Q c li represents the reactive power compensation capacity of node i in the lth scenario; y i is a 0 or 1 decision variable of node i, indicating whether node i is equipped with reactive power compensation equipment; il represents the reactive reserve unit capacity switching loss coefficient of node i in the lth scenario;

[0042] ③Node voltage constraint, its expression is:

[0043] V limin ≤V li≤V limax

[0044] Where V li represents the voltage amplitude of node i in the lth scenario, V limin Indicates the voltage lower limit of node i in the lth scenario, V limax represents the voltage upper limit of node i in the lth scenario;

[0045] ④ Norm constraint, its expression is:

[0046]

[0047] In the formula, ||·|| ∞ represents the infinite norm, p l represents the probability of the lth scenario occurring, The initial probability of the lth scenario occurring, L represents the total number of scenarios, θ ∞ is the allowed change value of probability under the infinite norm constraint, α ∞ is the confidence level.

[0048] Furthermore, the nonlinear constraints in the power constraints are converted into linear constraints by using the strong second-order cone relaxation method, specifically including:

[0049] The nonlinear power flow constraints and node voltage constraints are replaced by the strong second-order cone relaxation method to transform the mixed integer nonlinear optimization problem into a linear optimization problem, which can be expressed as:

[0050]

[0051] Q Cli ≤|x i y i |i∈S B

[0052] c ij =c ji ,s ij =-s ji i,j∈S B

[0053]

[0054] Where c ij represents the real voltage product between node i and node j, c ji represents the real voltage product between node j and node i, s ij represents the imaginary voltage product between nodes i and j, s ji represents the imaginary voltage product between node j and node i, c ii represents the square of the voltage modulus at node i, c jjrepresents the square of the voltage modulus at node j, V imin Represents the lower voltage limit of node i, V imax represents the voltage upper limit of node i, and ||·||2 represents the L2 norm.

[0055] Furthermore, the step (6) specifically includes:

[0056] The objective function of the effective reactive reserve planning and configuration model is decomposed to obtain the main problem corresponding to the planning and configuration cost function and the subproblems corresponding to the distributed robustness cost. The subproblems are decomposed according to different scenarios and then solved independently. When solving the subproblems, based on the complete set of scenarios enriched by random forests, the Gurobi solver is used to solve the subproblems under linearized power constraints to obtain the probability distribution with the highest robustness cost. This probability distribution is then substituted into the main problem for iterative calculation to achieve the solution of the two-stage distributed robust optimization model:

[0057]

[0058] Where x * represents the first-stage variable determined after solving the main problem, p l represents the probability of the occurrence of the lth scenario, Ω represents the uncertain set of scenario probabilities, ξ l represents the lth scene, w l represents auxiliary variables, Indicates the auxiliary variable introduced at the nth iteration, U(x,ξ l ) is w l The feasible domain of represents the second stage constraint; ultimately the optimal configuration of effective reactive power reserve planning is obtained.

[0059] The beneficial effects of the present invention are as follows: the present invention fully considers the uncertainty of renewable energy and the effectiveness of reactive power reserves, combines the strong second-order cone relaxation method with the column and constraint algorithm to solve the problem, and achieves the economically optimal planning configuration of the maximum effective reactive power reserve planning configuration. The present invention verifies the effectiveness of the proposed model through real examples, effectively solving the planning and configuration optimization problem of the maximum effective reactive power reserve in power systems with large-scale renewable energy injection, and provides a new idea and method for reactive power reserve planning in new power systems. In the scenario of large-scale renewable energy injection into the power grid, the present invention combines the random forest algorithm with distributed robust optimization, fully considering the dual uncertainty of load and renewable energy and the robustness of effective reactive power reserve configuration. Through real examples, it is verified that the optimal configuration of the maximum effective reactive power reserve planning can be achieved, while balancing economic efficiency and reactive power reserve effectiveness. The present invention can reduce the invalid reserve by 76.2% and the voltage deviation rate by 67.3%, and the configuration cost is reduced by 12.8% compared with the traditional robust model. It not only ensures the stable operation of the power system under large-scale renewable energy injection, but also achieves the economic optimization of the maximum effective reactive power reserve planning configuration. It provides theoretical support and practical paradigm for the stable and economic operation of high-proportion renewable energy power grids, and is of great significance to the development of clean energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The definition and comparison diagram of reactive power reserve;

[0061] Figure 2 It is the relationship curve between the sum of squares of the range and the number of clusters;

[0062] Figure 3 It is the relationship curve between silhouette coefficient and number of clusters;

[0063] Figure 4 Configuring scheme diagram for effective reactive power reserve planning;

[0064] Figure 5 A comparison chart of reactive power reserves with and without considering fluctuations in renewable energy sources;

[0065] Figure 6 This is a curve diagram showing the impact of the proportion of new energy on reactive power reserves;

[0066] Figure 7 Flowchart of the method for planning and configuring effective reactive power reserves for a high proportion of renewable energy power grid. DETAILED DESCRIPTION

[0067] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. When the following description refers to the drawings, identical numbers in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims. It should be understood that the foregoing general description and the detailed description that follows are exemplary and illustrative only and do not limit the present application.

[0068] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. As used in this application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0069] It should be understood that although the terms first, second, third, etc. may be used in this application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of..." or "when..." or "in response to determination." Moreover, the term "comprises," "comprising," or any other variant thereof is intended to cover non-exclusive inclusion, so that the process or method comprising a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process or method. In the absence of further restrictions, the elements defined by the statement "comprising a..." do not exclude the presence of other identical elements in the process, method, article, or device comprising the elements.

[0070] The present invention will be described in detail below with reference to the accompanying drawings. Unless there is any conflict, the features of the following embodiments and implementations may be combined with each other.

[0071] See also Figure 7 The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid of the present invention specifically comprises the following steps:

[0072] (1) The effective reactive power reserve of the high-proportion new energy grid of the present invention is different from the concept of the effective reactive power reserve of the traditional synchronous machine. Therefore, it is necessary to first define the effective reactive power reserve of each electrical node (substation) of the high-proportion new energy grid.

[0073] Specifically, the maximum reactive power that can maintain the voltage of the electrical node within the safe range in the actual project is selected. Minimum reactive power Maximum forward reactive output for all possible operating scenarios Negative maximum reactive output The smaller value in the middle is the effective reactive power reserve, and its definition is shown in the figure below. Figure 1 As shown, the maximum effective reactive power reserve in the forward direction is The maximum negative effective reactive power reserve is Among them, the effective reactive power reserve is directional. The positive effective reactive power reserve and the negative effective reactive power reserve represent the reactive power reserve devices of the electrical nodes in the form of parallel capacitor groups and parallel reactor groups, respectively.

[0074] (2) Based on the historical data of new energy stations, an uncertainty full scenario set is constructed, and random forest is introduced to enrich the scenario set in combination with local meteorological data to obtain a complete version of the full scenario set.

[0075] Specifically, new energy mainly exists in the form of wind farms and photovoltaic power plants. The output fluctuations of both have daily periodicity, so one day is selected as a scenario. First, based on the historical data of new energy stations, an uncertainty full scenario set is constructed. The uncertainty full scenario set includes the uncertainty full scenario set corresponding to the power source in the new energy station and the uncertainty full scenario set corresponding to the load. The uncertainty full scenario set corresponding to the power source in the new energy station is expressed as:

[0076] S={s1,s2,...,s l ,...,s L}

[0077] s l =((i,P l 1 ,Q l 1 ),(i,P l 2 ,Q l 2 ),...,(i,P l m ,Q l m ),...,(i,P l M ,Q l M ))

[0078]

[0079] In the formula, S represents the uncertainty scenario set corresponding to the power source in the new energy station, of which there are L scenarios in total, s lIndicates the lth scenario corresponding to the power supply, scenario s l There are M new energy stations in the system, i is the electrical node number, m is the new energy station number, P l m Represents scene s l The active power output of the new energy station m, Q l m Represents scene s l The reactive power output of the new energy station m in the new energy station is expressed as follows:

[0080]

[0081] Where Z represents the uncertainty scenario set corresponding to the load in the new energy station, of which there are L scenarios in total, z l Indicates the lth scenario corresponding to the load, scenario z l There are D new energy stations in the system, i is the number of the electrical node, d is the number of the new energy station, P l d Represents scene z l Active power output of new energy station d, Q l d Represents scene z l Reactive power output of new energy station d.

[0082] Then, the original data full scene set X={S,Z}={(s1,z1),(s2,z2),...,(s l ,z l ),...,(s L ,z L )}, perform T bootstrap sampling on the original data full scene set X, extract N samples each time, and generate T subsets:

[0083]

[0084] Where D t represents the t-th subset; represents the nth sample in the tth subset, and also represents the nth sample extracted from the original data full scene set X. Each sample in the original data full scene set X It is drawn with probability 1 / N.

[0085] It should be understood that bootstrap sampling is a powerful statistical resampling technique. It is a nonparametric method for estimating the sampling distribution of a statistic, particularly useful when the theoretical distribution is difficult to derive or when the sample size is small. The core idea of ​​bootstrap sampling is to simulate the sampling distribution by repeatedly sampling with replacement from the original data.

[0086] Then, T decision trees are constructed based on the T subsets to form a random forest. When constructing the t-th tree, the optimal split point is recursively selected, and the splitting criterion is to minimize the Gini impurity.

[0087] Secondly, the classification task is completed based on random forest combined with local meteorological data, which is specifically achieved through soft voting to obtain a set of predicted future scenarios.

[0088] It should be noted that classification tasks can be divided into hard voting and soft voting. Among them, hard voting can be understood as: counting the number of predictions for each category in T trees; the final prediction result is the category with the most votes. Soft voting can be understood as: counting the probability of each category predicted in T trees (usually determined by the proportion of samples in each category in the leaf node); averaging the category probabilities predicted by T trees; and selecting the category with the largest average probability as the final prediction result of the classification task. Soft voting contains information about the confidence of the model and is usually more effective. Therefore, soft voting is used to implement classification tasks, which can be expressed by the following formula:

[0089]

[0090] Where h t (x) represents the predicted output of the t-th tree, x represents the local weather data as the input of the tree, It represents the average probability that the predicted output of T trees is category k, where K represents the total number of categories. Represents the final predicted category of the classification task. Taking the local meteorological data as the input of the tree, we get the prediction output of T trees, that is, the predicted future scene set.

[0091] Finally, the original data full-scene set corresponding to the historical data and the predicted future scene set are combined to form a complete full-scene set containing L+T scenes.

[0092] (3) Based on the complete version of the full scene set, typical scenes and the probability distribution of each typical scene are obtained through improved k-means clustering.

[0093] Specifically, the traditional k-means clustering method uses the sum of squared errors within the cluster to measure the current cluster C r The specific formula for the degree of polymerization is as follows:

[0094]

[0095] Where, SSE r Represents the rth cluster C r The degree of polymerization, c j Represents cluster C r The jth data in Represents cluster C r The average value of all data in, J represents the cluster C r Considering the volatility of new energy, the maximum reactive reserve capacity configuration is planned for the source end of the new power grid. The sum of squares of all data in the cluster SSE r The ability to characterize the boundary conditions of new energy fluctuations is limited. Here, the k-means clustering method is improved by introducing the sum of squared ranges, where the sum of squared ranges is expressed as:

[0096]

[0097] Where, NSSE r Represents the improved r-th cluster C r The sum of squares of the range, c rmax Represents cluster C r The maximum value in c rmin Represents cluster C r Then, the improved k-means clustering method is used to perform k-means clustering on the L+T scene sets in the complete version of the full scene set to obtain R clusters; the elbow method and silhouette coefficient method in the k-means clustering method are combined to determine the number of typical scene clusters as A, for example, Figure 2 and Figure 3 The following are experimental results obtained using the elbow and silhouette coefficient methods in the k-means clustering method. The optimal number of clusters is 6, indicating the number of typical scene clusters. The first A clusters with the smallest sum of squared range differences (SSS) from the R clusters are selected as typical scene clusters to obtain typical scenes. Finally, by calculating the proportion of scenes in each typical scene cluster to the total number of scenes (L+T), the corresponding typical scene probability distribution is obtained as the initial probability distribution for that scene.

[0098] (4) Establish an effective reactive power reserve planning and configuration model considering the uncertainty of the new power system under the massive injection of new energy, so as to find the probability distribution of the planning configuration scheme with the highest robustness cost and give the optimal planning configuration scheme of effective reactive power reserve based on this distribution.

[0099] Furthermore, the purpose of establishing an effective reactive power reserve planning and configuration model is to configure effective reactive power reserves for the new power system with large-scale new energy injection. Specifically, the objective function of the effective reactive power reserve planning and configuration model is:

[0100]

[0101] f1=C fix ·y i +C fc ·y i ·x i

[0102]

[0103] Where f1 represents the planning configuration cost function; f2 represents the distributed robustness cost; i is the electrical node number, I is the total number of electrical nodes; C fix represents the fixed cost of installing reactive power compensation equipment at node i, C fc represents the unit capacity cost coefficient, and its unit is RMB 400,000 / Mvar; y is a decision variable of 0 or 1, indicating whether reactive power compensation equipment is installed. y = 0 indicates that reactive power compensation equipment is not installed, and y = 1 indicates that reactive power compensation equipment is installed. i is a 0 or 1 decision variable for node i, indicating whether node i should install reactive power compensation equipment; x is a continuous variable indicating the maximum effective reactive power reserve capacity of the planned configuration; x i is a continuous variable, representing the maximum effective reactive reserve capacity of node i; l represents the lth scenario, L is the total number of scenarios; z l represents the reactive reserve unit capacity switching loss coefficient in the lth scenario, U(x) is z l The feasible domain of z il represents the reactive reserve unit capacity switching loss coefficient of node i in the lth scenario; T total Indicates the total number of moments in a single scene, T out,il represents the number of time points when the voltage of node i exceeds the limit in the lth scenario; f3(y,x,z l ) represents the cost function of reactive power reserve equipment operation loss; p l represents the probability of the occurrence of the lth scenario; f4 represents the voltage level deviation penalty function; represents the maximum amplitude of the voltage fluctuation at node i in the lth scenario, v0 represents the voltage reference value, and the per-unit value is 1.

[0104] (5) Establish the power constraints of the effective reactive power reserve planning configuration model, and transform the nonlinear constraints in the power constraints into linear constraints through the strong second-order cone relaxation method.

[0105] Furthermore, the power constraints of the effective reactive power reserve planning configuration model include:

[0106] ① Power flow constraint, its expression is:

[0107]

[0108] Where, P Gli represents the active power output of power node i in the lth scenario, Q Gli represents the reactive power output of power node i in the lth scenario, P Lli represents the active power of load node i in the lth scenario, Q Lli represents the reactive power of load node i in the lth scenario, V li represents the voltage amplitude of node i in the lth scenario, V lj represents the voltage amplitude of node j in the lth scenario, δ li represents the phase angle of node i in the lth scenario, δ lj represents the phase angle of node j in the lth scenario, G ij represents the real part of the admittance of the branch between node i and node j, B ij represents the imaginary part of the admittance of the branch between node i and node j, S B is the set of all nodes.

[0109] ② Branch power constraint, its expression is:

[0110] Q c li≤y i z il i∈S B

[0111] Q c li≥-y i z il i∈S B

[0112] Where Q cli represents the reactive power compensation capacity of node i in the lth scenario, which does not exceed the upper limit and lower limit of reactive reserve capacity configured at node i; y i is a 0 or 1 decision variable of node i, indicating whether node i is equipped with reactive power compensation equipment; il represents the reactive reserve unit capacity switching loss coefficient of node i in the lth scenario.

[0113] ③Node voltage constraint, its expression is:

[0114] V limin ≤V li ≤V limax

[0115] Where V li represents the voltage amplitude of node i in the lth scenario, V limin Indicates the voltage lower limit of node i in the lth scenario, V limaxrepresents the voltage upper limit of node i in the lth scenario.

[0116] ④ Norm constraint, its expression is:

[0117]

[0118] In the formula, ||·|| ∞ represents the infinite norm, p l represents the probability of the lth scenario occurring, The initial probability of the lth scenario occurring, L represents the total number of scenarios, θ ∞ is the allowed change value of probability under the infinite norm constraint, α ∞ is the confidence level.

[0119] Furthermore, the nonlinear constraints in the power constraints are converted into linear constraints through the strong second-order cone relaxation method. Specifically, the strong second-order cone relaxation method is used to replace the nonlinear power flow constraints and node voltage constraints, and the mixed integer nonlinear optimization problem is converted into a linear optimization problem to avoid local optimal solutions. It can be expressed as:

[0120]

[0121] Q Cli ≤|x i y i |i∈S B

[0122] c ij =c ji ,s ij =-s ji i,j∈S B

[0123]

[0124] Where c ij represents the real voltage product between node i and node j, c ji represents the real voltage product between node j and node i, s ij represents the imaginary voltage product between nodes i and j, s ji represents the imaginary voltage product between node j and node i, c ii represents the square of the voltage modulus at node i, c jj represents the square of the voltage modulus at node j, V imin Represents the lower voltage limit of node i, V imax represents the voltage upper limit of node i, and ||·||2 represents the L2 norm.

[0125] (6) Based on the complete scenario set enriched by random forest, the effective reactive power reserve planning configuration model is solved by Gurobi solver under the linearized power constraint condition to obtain the optimal configuration of effective reactive power reserve planning.

[0126] Specifically, the objective function of the effective reactive reserve planning and configuration model is decomposed to obtain a main problem corresponding to the planning and configuration cost function and a subproblem corresponding to the distributed robustness cost. Since the minimization problems solved under different scenarios in the subproblems are independent of each other, the subproblems can be decomposed according to different scenarios and then solved independently in sequence. Based on the complete set of scenarios enriched by random forests, the subproblems are solved using the Gurobi solver under linearized power constraints to obtain the probability distribution with the highest robustness cost. This probability distribution is then substituted into the main problem for iterative calculation to achieve the solution of the two-stage distributed robust optimization (DRO) model:

[0127]

[0128] Where x * represents the first-stage variable determined after solving the main problem, p l represents the probability of the occurrence of the lth scenario, Ω represents the uncertain set of scenario probabilities, ξ l represents the lth scene, w l represents auxiliary variables, Indicates the auxiliary variable introduced at the nth iteration, U(x,ξ l ) is w l The feasible domain of represents the second stage constraint; ultimately the optimal configuration of effective reactive power reserve planning is obtained.

[0129] Furthermore, after obtaining the optimal configuration of the effective reactive power reserve plan, its effectiveness can be evaluated and then compared with the configuration costs of other models to verify the effectiveness of the method described in the present invention. The evaluation index is shown in the following formula:

[0130]

[0131]

[0132] Where Q l,IRPR is the total reactive power reserve of the lth scenario, Q li,IRPR is the invalid reactive power reserve of the lth scene node i, N ineff is the total number of nodes with invalid reactive power reserve; R ineff is the ratio of invalid reactive power reserve nodes, N plan The total number of nodes optimized for reactive power reserve planning and configuration, VDR is the voltage deviation rate, V itis the actual voltage of node i at the tth moment in the scenario, V B,i is the reference voltage of node i, N is the total number of nodes, T is the total number of time; VFR is the voltage fluctuation rate, V i t It is the maximum value of the actual voltage amplitude of node i in a certain time domain.

[0133] The following describes in detail the method for planning and configuring effective reactive power reserves in a high-proportion new energy power grid according to embodiments of the present invention, and the purpose and effect of the present invention will become more apparent.

[0134] This example illustrates how the method described herein can optimize the effective reactive power reserve planning and configuration of a renewable energy grid, particularly in the context of large-scale renewable energy injection. This example was performed on a personal computer equipped with an Intel Core i5-12500H CPU @ 2.50GHz, 32.0GB of RAM, and an NVIDIA GTX 3060 GPU. The program was written in Python 3.10.9, and the optimization framework used was Gurobi 10.0.3.

[0135] In this example, a real grid in a province in eastern China was used as a construction example. This grid contains 6,188 nodes, 8,687 branches, and 73 large conventional thermal power plants. The total number of renewable energy station nodes in this example is 267, most of which are located offshore in the northeastern waters and onshore in the northern part of the province. The provincial grid has 863 525kV interconnection nodes, with key interconnection nodes in the northern and eastern regions being significantly affected by renewable energy fluctuations. Due to confidentiality requirements, the network topology cannot be presented here.

[0136] First, according to the content described in step (1), define the effective reactive power reserve of each electrical node in the high-proportion new energy grid, such as Figure 1 As shown. The maximum effective reactive power reserve in the forward direction is The maximum negative effective reactive power reserve is

[0137] Then, according to the contents described in step (2) and step (3), based on the established new power system uncertainty full scenario set model, the improved k-means combined with the elbow method and the silhouette coefficient method is used to determine the number of typical scenarios. The relationship curves of the sum of squares of the range difference and the number of clusters and the relationship curves of the silhouette coefficient and the number of clusters are shown as follows: Figure 2 and Figure 3 As shown. Figure 1 、 Figure 2 The experimental results of the graph feedback show that the number of typical scene clusters is 6. By calculating the proportion of the number of scenes in each typical scene cluster to the total number of scenes, the corresponding typical scene probability distribution is obtained, and the results are shown in Table 1.

[0138] Table 1: Probability distribution of typical scenarios

[0139]

[0140]

[0141] The probability distribution of each typical scenario and the fluctuation conditions of each scenario are then obtained from the clustering results, and then the switching loss coefficient and the maximum voltage fluctuation amplitude are obtained.

[0142] Secondly, the probability distribution, switching loss coefficient, and maximum voltage fluctuation amplitude are substituted into the two-stage DRO model for optimization to find the planning configuration scheme with the highest robustness cost near the initial fuzzy probability distribution. The planning configuration scheme with the highest robustness cost is obtained as follows: Figure 4 The nodes selected for effective reactive power reserve planning and configuration optimization are 174 key communication nodes in the northern region of the province. According to the optimization results, 42 of these nodes do not require effective reactive power reserve configuration, 54 nodes require only positive effective reactive power reserve configuration, 17 nodes require only negative effective reactive power reserve configuration, and only 61 busbar nodes require both positive and negative effective reactive power reserve configuration.

[0143] Then, the busbar node is selected and the planning configuration scheme after considering the effectiveness of reactive power reserve optimization is compared with the planning configuration scheme without considering the effectiveness. Figure 5 As shown in the figure, the energy penetration rate in the example is gradually increased, and the total amount of reactive power reserve changes with the increase of the proportion of new energy with and without considering the effectiveness. Figure 6 Then, three scenarios were randomly selected from the historical data scenario set to compare the voltage stability before and after configuring the effective reactive power reserve to verify the effectiveness of the reactive power reserve, as shown in Table 2.

[0144] Table 2: Comparison of voltage conditions before and after effective reactive power reserve configuration

[0145]

[0146] Finally, the data-driven robust optimization model was compared with the traditional robust optimization model (RO) and the stochastic optimization model (SO). The results are shown in Table 3. The DRO model combines the advantages of both RO and SO. Compared to the SO model, it considers the allowable deviation of the scenario probability distribution based on the existing scenario probabilities and takes into account adverse scenarios, resulting in a more robust result. Compared with the overly conservative RO model, its optimization results are more economical while retaining robustness.

[0147] Table 3: Comparison of total configuration costs of three optimization models

[0148]

[0149] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for planning and configuring effective reactive power reserves in a high-proportion new energy power grid, characterized in that: The following steps are involved: (1) Define the effective reactive power reserve of each electrical node in a high-proportion renewable energy grid; (2) Construct an uncertainty full scenario set based on historical data of new energy stations, and introduce random forest to enrich the scenario set in combination with local meteorological data to obtain a complete version of the full scenario set; (3) Based on the complete set of scenes, typical scenes and their probability distribution are obtained by improving k-means clustering; (4) Establish an effective reactive power reserve planning and configuration model considering power system uncertainty; (5) Establish the power constraints of the effective reactive power reserve planning configuration model and transform the nonlinear constraints in the power constraints into linear constraints through the strong second-order cone relaxation method; (6) Based on the complete scenario set enriched by random forest, the effective reactive power reserve planning configuration model is solved by Gurobi solver under the linearized power constraint condition to obtain the optimal configuration of effective reactive power reserve planning.

2. The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid according to claim 1 is characterized in that: The step (1) specifically includes: Select the maximum reactive power that can maintain the voltage of the electrical node within the safe range Minimum reactive power Maximum forward reactive output for all possible operating scenarios Negative maximum reactive output The smaller value in the middle is the effective reactive power reserve, that is, the maximum effective reactive power reserve in the positive direction is The maximum negative effective reactive power reserve is The effective reactive power reserve is directional, and the positive effective reactive power reserve and the negative effective reactive power reserve represent that the reactive power reserve devices of the electrical nodes are in the form of a parallel capacitor group and a parallel reactor group, respectively.

3. The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid according to claim 1 is characterized in that: The step (2) specifically includes: First, a full uncertainty scenario set is constructed based on the historical data of the new energy station. The full uncertainty scenario set includes the full uncertainty scenario set corresponding to the power source and the full uncertainty scenario set corresponding to the load in the new energy station. The full uncertainty scenario set corresponding to the power source in the new energy station is expressed as: In the formula, S represents the uncertainty scenario set corresponding to the power source in the new energy station, of which there are L scenarios in total, s l Indicates the lth scenario corresponding to the power supply, scenario s l There are M new energy stations in the system, i is the electrical node number, m is the new energy station number, P l m Represents scene s l The active power output of the new energy station m, Q l m Represents scene s l The reactive power output of the new energy station m in the new energy station is expressed as follows: Where Z represents the uncertainty scenario set corresponding to the load in the new energy station, of which there are L scenarios in total, z l Indicates the lth scenario corresponding to the load, scenario z l There are D new energy stations in the system, i is the number of the electrical node, d is the number of the new energy station, P l d Represents scene z l Active power output of new energy station d, Q l d Represents scene z l The reactive power output of the new energy station d; Then, the original data full scene set X is obtained based on the uncertainty full scene set corresponding to the power source and the uncertainty full scene set corresponding to the load in the new energy station. Bootstrap sampling is performed on the original data full scene set X T times, and N samples are extracted each time to generate T subsets. Then, T decision trees are constructed based on the T subsets to form a random forest. When constructing the t-th tree, the optimal split point is recursively selected, and the splitting criterion is to minimize the Gini impurity. Secondly, the classification task is completed based on random forest combined with local meteorological data, which is specifically achieved through soft voting to obtain a set of predicted future scenarios; Finally, the original data full-scene set corresponding to the historical data and the predicted future scene set are combined to form a complete full-scene set containing L+T scenes.

4. The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid according to claim 1 is characterized in that: The step (3) specifically includes: First, the k-means clustering method is improved by introducing the sum of squares of the range, where the sum of squares of the range is expressed as: Where, NSSE r Represents the improved r-th cluster C r The sum of squares of the range, c j Represents cluster C r The jth data in Represents cluster C r The average value of all data in, J represents the cluster C r The total number of data in c rmax Represents cluster C r The maximum value in c rmin Represents cluster C r The minimum value in ; Then, the improved k-means clustering method is used to perform k-means clustering on the L+T scene sets in the complete version of the full scene set, obtaining R clusters. The elbow method and silhouette coefficient method in the k-means clustering method are combined to determine the number of typical scene clusters as A. From the R clusters, the first A clusters with the smallest sum of squared extreme differences are selected as typical scene clusters to obtain typical scenes. Finally, the corresponding typical scene probability distribution is obtained by calculating the proportion of the number of scenes in each typical scene cluster in the total number of scenes.

5. The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid according to claim 1 is characterized in that: The objective function of the effective reactive power reserve planning configuration model is: f1=C fix ·y i +C fc ·y i ·x i Where f1 represents the planning configuration cost function; f2 represents the distributed robustness cost; i is the electrical node number, I is the total number of electrical nodes; C fix represents the fixed cost of installing reactive power compensation equipment at node i, C fc Represents the unit capacity cost coefficient; y is a decision variable of 0 or 1, indicating whether to install reactive power compensation equipment; y i is a 0 or 1 decision variable for node i, indicating whether node i should install reactive power compensation equipment; x is a continuous variable indicating the maximum effective reactive power reserve capacity of the planned configuration; x i is a continuous variable, representing the maximum effective reactive reserve capacity of node i; l represents the lth scenario, L is the total number of scenarios; z l represents the reactive reserve unit capacity switching loss coefficient in the lth scenario, U(x) is z l The feasible domain of z il represents the reactive reserve unit capacity switching loss coefficient of node i in the lth scenario; T total Indicates the total number of moments in a single scene, T out,il represents the number of time points when the voltage of node i exceeds the limit in the lth scenario; f3(y,x,z l ) represents the cost function of reactive power reserve equipment operation loss; p l represents the probability of the occurrence of the lth scenario; f4 represents the voltage level deviation penalty function; represents the maximum amplitude of the voltage fluctuation at node i in the lth scenario, and v0 represents the voltage reference value.

6. The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid according to claim 1 is characterized in that: The power constraints of the effective reactive power reserve planning configuration model specifically include: ① Power flow constraint, its expression is: Where, P Gli represents the active power output of power node i in the lth scenario, Q Gli represents the reactive power output of power node i in the lth scenario, P Lli represents the active power of load node i in the lth scenario, Q Lli represents the reactive power of load node i in the lth scenario, V li represents the voltage amplitude of node i in the lth scenario, V lj represents the voltage amplitude of node j in the lth scenario, δ li represents the phase angle of node i in the lth scenario, δ lj represents the phase angle of node j in the lth scenario, G ij represents the real part of the admittance of the branch between node i and node j, B ij represents the imaginary part of the admittance of the branch between node i and node j, S B is the set of all nodes; ② Branch power constraint, its expression is: Q c li≤y i z il i∈S B Q c li≥-y i z il i∈S B Where Q c li represents the reactive power compensation capacity of node i in the lth scenario; y i is a 0 or 1 decision variable of node i, indicating whether node i is equipped with reactive power compensation equipment; il represents the reactive reserve unit capacity switching loss coefficient of node i in the lth scenario; ③Node voltage constraint, its expression is: In limin ≤V li ≤V limax Where V li represents the voltage amplitude of node i in the lth scenario, V limin Indicates the voltage lower limit of node i in the lth scenario, V limax represents the voltage upper limit of node i in the lth scenario; ④ Norm constraint, its expression is: In the formula, ||·|| ∞ represents the infinite norm, p l represents the probability of the lth scenario occurring, The initial probability of the lth scenario occurring, L represents the total number of scenarios, θ ∞ is the allowed change value of probability under the infinite norm constraint, α ∞ is the confidence level.

7. The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid according to claim 6 is characterized in that: The method of converting the nonlinear constraints in the power constraints into linear constraints by using the strong second-order cone relaxation method specifically includes: The nonlinear power flow constraints and node voltage constraints are replaced by the strong second-order cone relaxation method to transform the mixed integer nonlinear optimization problem into a linear optimization problem, which can be expressed as: Q Cli ≤|x i y i |i∈S B c ij =c ji ,s ij =-s ji i,j∈S B Where c ij represents the real voltage product between node i and node j, c ji represents the real voltage product between node j and node i, s ij represents the imaginary voltage product between nodes i and j, s ji represents the imaginary voltage product between node j and node i, c ii represents the square of the voltage modulus at node i, c jj represents the square of the voltage modulus at node j, V imin Represents the lower voltage limit of node i, V imax represents the voltage upper limit of node i, and ||·||2 represents the L2 norm.

8. The method for planning and configuring effective reactive power reserves of a high-proportion new energy power grid according to claim 1 is characterized in that: The step (6) specifically includes: The objective function of the effective reactive reserve planning and configuration model is decomposed to obtain the main problem corresponding to the planning and configuration cost function and the subproblems corresponding to the distributed robustness cost. The subproblems are decomposed according to different scenarios and then solved independently. When solving the subproblems, based on the complete set of scenarios enriched by random forests, the Gurobi solver is used to solve the subproblems under linearized power constraints to obtain the probability distribution with the highest robustness cost. This probability distribution is then substituted into the main problem for iterative calculation to achieve the solution of the two-stage distributed robust optimization model: Where x * represents the first-stage variable determined after solving the main problem, p l represents the probability of the occurrence of the lth scenario, Ω represents the uncertain set of scenario probabilities, ξ l represents the lth scene, w l represents auxiliary variables, Indicates the auxiliary variable introduced at the nth iteration, U(x,ξ l ) is w l The feasible domain of represents the second stage constraint; ultimately the optimal configuration of effective reactive power reserve planning is obtained.

Images