Optimized scheduling method of wind-solar-pumped storage complementary power generation system

Abstract

The invention relates to an optimal scheduling method of a wind-light-pumped storage complementary power generation system. The method comprises the following steps: acquiring wind-light data of the pumped storage complementary power generation system; based on the wind and light data, generating a wind and light scene by using a WGAN-GP network model, and performing scene reduction on the wind and light scene by using a K-means + + method to obtain a typical wind and light scene and a probability thereof; establishing a two-stage distribution robust optimization model considering the wind and light uncertainty, and adopting Camp; the two-stage distribution robust optimization model is solved through a CG algorithm, an optimal scheduling scheme is obtained, and in the solving process, the typical wind and light scene and the probability of the typical wind and light scene serve as the initial scene and the initial probability of the two-stage distribution robust optimization model. Compared with the prior art, the method has the advantages that the wind and light output fluctuation can be stabilized, and the operation safety, economy and stability of the system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system optimization and dispatching, and in particular to an optimization and dispatching method for a wind-solar-pumped-storage complementary power generation system. Background Art

[0002] Affected by factors such as wind speed, light, and climate, wind and solar power output has strong randomness and volatility. Large-scale wind and solar access to the power grid will pose a huge challenge to the safety and stability of the power system operation. Relying on the complementary characteristics between wind, solar, and water, it can only smooth out the fluctuations in wind and solar power output, and cannot completely solve the problem of wind and solar power consumption. The energy storage system can not only alleviate the power fluctuations caused by the intermittent and random nature of wind and solar power output, but also play an important role in promoting wind and solar power consumption and achieving emission reduction and decarbonization. Compared with other energy storage methods, pumped storage has the advantages of long life, large capacity, and zero emissions. Traditional pumped storage power stations generate electricity and store energy during peak and trough loads of the power grid, and there will be certain power losses in the process. The wind-solar-pumped storage multi-energy complementary power generation system built on the basis of pumped storage power stations is an effective means to promote wind and solar power consumption and smooth load fluctuations. The literature "Simulation Research on Line Arrester Based on ATP-EMTP" applies the double-layer optimization theory to the capacity configuration problem of pumped storage power stations. The upper target is to minimize the investment cost of the pumped storage power station, and the lower target is to maximize the net profit of the system. The paper "Optimizing cleaner productions of sustainable energies: Aco-design framework for complementary operations of offshore wind and pumped hydro-storages" established a model of wind power and pumped storage complementary power generation system. The experimental results show that pumped storage power stations can compensate for the volatility of wind power output and significantly improve the total power generation income of the complementary system. The paper "Optimal Dispatch of Composite Energy Storage for Large-Scale Photovoltaic Power Generation System Based on Carbon Trading" uses pumped storage to adjust the fluctuation of photovoltaic output, achieving large-scale photovoltaic grid-connected consumption, so that the power generation system can operate in a low-carbon and economical manner. The above-mentioned paper does not take into account the fluctuation of wind power and photovoltaic output at the same time for the power system containing pumped storage power stations, and the calculation results may tend to be optimistic.

[0003] The randomness and volatility of wind and solar power output cannot be ignored in optimizing scheduling. At present, the research methods for the problem of wind and solar power output uncertainty mainly focus on stochastic optimization and robust optimization. The paper "Robust Optimal Scheduling Method for Power System Considering Hydropower, Photovoltaic and Storage Complementarity and DC Transmission" proposed an optimal scheduling model for power system considering hydropower, photovoltaic and storage complementarity and DC transmission, on which the uncertainty of photovoltaic output was considered, but the uncertainty of wind power output was not considered; the paper "Comprehensive Stochastic Optimization Model for Planning and Operation of Thermal Power Unit Flexibility Transformation" established a stochastic optimization model considering wind power ramping based on the opportunity constraint theory. The paper "Coordination of interdependent natural gas and electricity infrastructures for firming the variability of wind energy in stochastic day-ahead scheduling" uses scenario analysis to deal with the uncertainty of wind power output in the system. Stochastic optimization needs to obtain the probability distribution of uncertain quantities from historical data and select representative scenarios for decision-making. However, the more accurate the parameter description, the more information is required, the larger the amount of calculation, and the more accurate the probability distribution is, which may lead to low model accuracy. The papers "Robust two-stage regional-district scheduling of multi-carrier energy systems with a large penetration of wind power" and "Robust optimization for energy transactions in multi-microgrids under uncertainty" describe the uncertainty of wind and solar power output with box-type uncertainty sets and construct a two-stage robust optimization scheduling model. Robust optimization only optimizes the model under the worst scenario, making the scheduling scheme too conservative and less economical. Summary of the invention

[0004] The purpose of the present invention is to provide an optimization scheduling method for a wind-solar-pumped storage complementary power generation system to improve the operational safety and stability of the system.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] An optimization scheduling method for a wind-solar-pumped-storage complementary power generation system comprises the following steps:

[0007] Obtain wind and solar data including pumped storage complementary power generation system;

[0008] Based on the wind and solar data, the WGAN-GP network model is used to generate wind and solar scenes, and the K-means++ method is used to reduce the wind and solar scenes to obtain typical wind and solar scenes and their probabilities;

[0009] A two-stage distributed blue-robust optimization model considering the uncertainty of wind and solar power is established, and the C&CG algorithm is used to solve the two-stage distributed blue-robust optimization model to obtain the optimal scheduling scheme. In the solution process, the typical wind and solar power scenes and their probabilities are used as the initial scenes and initial probabilities of the two-stage distributed blue-robust optimization model.

[0010] Furthermore, the objective function used in the WGAN-GP network model training process is:

[0011]

[0012] In the formula, G and D are the generator and discriminator respectively, E(·) is the expected function, G(·) is the generating function, and D(·) is the discriminator function. is the penalty term, ||·||2 represents the 2-norm, ρ is the penalty coefficient, ε is a random number that satisfies a uniform distribution on the interval [0,1], P d is the probability distribution of the real historical data x of Fengguang, P z is the probability distribution of random noise z.

[0013] Furthermore, in the process of reducing the scene by using the K-means++ method, the Davis-Bourdin index and the silhouette coefficient are introduced for quantitative calculation to determine the optimal number of clusters.

[0014] Furthermore, the calculation expression of the Davis-Bourdin index is:

[0015]

[0016] Where DBI(i) is the Davis-Bourdin index of the ith cluster, N is the total number of clusters, and C i , C j Represent the centers of the i-th and j-th clusters respectively; S i , S j Respectively represent all samples in cluster i, j to cluster center C i , C j The average distance between clusters is also called the intra-cluster distance; d(C i ,C j ) represents the cluster center C i and C j The distance between.

[0017] Furthermore, the calculation expression of the silhouette coefficient is:

[0018]

[0019] Where s(u) represents the silhouette coefficient of the u-th sample, b(u) represents the average distance between the u-th sample and other samples outside the cluster, and a(u) is the average distance between the u-th sample and all samples in the cluster.

[0020] Furthermore, the two-stage distributed robust optimization model considering the uncertainty of wind and solar power includes two stages. The first stage considers the cost of purchasing electricity from the external power grid and the operating cost of pumped storage, and optimizes the scheduling strategy according to the predicted wind and solar power output in order to achieve the best cost-effectiveness. The second stage considers the uncertainty of wind and solar power. On the basis of the first stage, the unit output is adjusted according to the actual wind and solar power output to solve the optimal scheduling plan under the worst scenario.

[0021] Furthermore, the expression of the two-stage distributed robust optimization model is:

[0022]

[0023] I k ≤Qε

[0024] Fx+Gy≤l

[0025] in:

[0026]

[0027] In the formula, x is the decision variable of the first stage, including the cost of purchasing electricity from the external power grid and the operating cost of pumped storage, y k is the flexible adjustment variable in the second stage under scenario k, c represents the corresponding cost coefficient, ε represents the prediction vector of wind and solar, P k is the occurrence probability of scene k, N is the number of discrete scenes, D, H, I, Q, F, G, a, b, e, r, l are the correlation coefficient matrix and constant column vector, is the initial probability value of scene k, θ1 and θ ∞ are the probability allowable deviation limits under the corresponding constraints of 1-norm and ∞-norm, M is the number of generated samples, α1, α ∞ is the uncertainty confidence.

[0028] Furthermore, by introducing 0-1 auxiliary variables to perform equivalent transformation, the 1-norm and ∞-norm corresponding constraints are linearized respectively, wherein the 1-norm corresponding constraint is transformed as follows:

[0029]

[0030] The corresponding constraint transformation of the ∞-norm is as follows:

[0031]

[0032] In the formula, and They are the probability distribution p k Positive and negative shifts relative to the original probability distribution, and 0-1 flags for positive and negative offsets respectively.

[0033] Furthermore, the step of solving the two-stage distributed robust optimization model using the C&CG algorithm includes:

[0034] (1) Decomposing the two-stage robust optimization model into a main problem and sub-problems;

[0035] (2) Initialize uncertain variables;

[0036] (3) Solve the main problem to obtain the optimal solution of the current iteration and update the new lower bound;

[0037] (4) Solve the subproblem to obtain the worst scenario probability distribution of the current iteration and update the new upper bound;

[0038] (5) Determine whether the convergence condition is met. If so, output the optimal solution and the iteration ends. If not, add auxiliary variables and related constraints and return to step (3) for iterative calculation until the convergence condition is met and the optimal solution is output.

[0039] Furthermore, the main problem is described as:

[0040]

[0041] The sub-problem is described as:

[0042]

[0043] In the formula, c represents the corresponding cost coefficient, x is the decision variable of the first stage, including the cost of purchasing electricity from the external power grid and the operating cost of pumped storage, “*” represents the optimal solution, a and b are, respectively, P k j is the probability value of the jth iteration of scenario k, J is the number of iterations, is the second-stage flexible adjustment variable at the jth iteration under scenario k, c represents the corresponding cost coefficient, represents the predicted vector of the scenery at the jth iteration under scene k, P k is the occurrence probability of scene k, N is the number of discrete scenes, y k is the flexible adjustment variable in the second stage under scenario k, ε k The predicted vector representing the scenery of scene k.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) The present invention combines the WGAN-GP network model and the K-means++ method to generate clusters for the uncertainty scenarios of wind and solar power as the initial solution, and establishes a two-stage distributed robust optimization model by considering the uncertainty of wind and solar power. The C&CG algorithm is used to iteratively solve and approximate the optimal solution, thereby obtaining the optimal scheduling strategy and improving the operation safety and stability of the system.

[0046] (2) The present invention combines new energy with pumped-storage power stations, utilizing the good energy storage and regulation function of pumped-storage power stations to better track load changes, compensate for fluctuations in wind and solar power output, and achieve peak shaving and valley filling.

[0047] (3) The first stage of the two-stage robust optimization model of the present invention takes into account the system cost and predicted output, and can optimize the system scheduling decision. The second stage considers the uncertainty of wind and solar power on the basis of the first stage and adjusts the unit output to solve the optimal scheduling plan under the worst scenario while being able to achieve good economy.

[0048] (4) The present invention adopts the C&CG algorithm to solve the two-stage distributed robust optimization model. During the solution process, auxiliary variables and constraints related to the sub-problems are continuously introduced to improve the solution efficiency and better cope with the dynamics of the system in actual situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0050] Figure 2 This is a schematic diagram of the GAN scene generation method of the present invention;

[0051] Figure 3 It is the C&CG algorithm flow chart of the present invention;

[0052] Figure 4 It is a structural block diagram of the wind-solar-pumped-storage complementary power generation system of the present invention;

[0053] Figure 5 It is the renewable energy consumption diagram of the present invention;

[0054] Figure 6 This is a load comparison diagram before and after considering pumped storage in the present invention. DETAILED DESCRIPTION

[0055] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0056] This embodiment provides an optimization scheduling method for a wind-solar-pumped-storage complementary power generation system. Figure 1 As shown, the method comprises the following steps:

[0057] S1. Scenery scene generation based on WGAN-GP.

[0058] Generative Adversarial Network (GAN) is an unsupervised machine learning method proposed by Goodfellow et al. GAN consists of a generator G and a discriminator D. The GAN model structure is shown in the figure below. Figure 2 As shown in Figure 2, the core idea of ​​GAN is to train the dynamic game process between the generator and the discriminator. This paper assumes that the real historical data of Fengguang is x, and its probability distribution is P d There exists a set of random noise vectors n whose probability distribution satisfies P n . GAN can build P d and P n The mapping relationship between them is obtained by training the generator and the discriminator to obtain the generated samples that meet the distribution relationship of the real data, and the probability distribution P is obtained. g The training process is completed by two deep neural networks, namely the generator G(n; θ G ) and the discriminator D(x;θ D ), where θ G and θ D are the weights of the corresponding neural networks.

[0059] The loss functions of the generator G and the discriminator D can be defined as follows:

[0060]

[0061] Where L G and L D Denote the loss functions of the generator and discriminator respectively; E(·) is the expected function; G(·) is the generating function; D(·) is the discriminator function. L G The smaller the value of , the closer the samples generated by the generator are to the actual data, and the stronger the ability of the discriminator to distinguish the authenticity of the data. The objective function of the game process is expressed as:

[0062]

[0063] The objective function of WGAN-GP overall training is expressed as:

[0064]

[0065] where ||·||2 represents the 2-norm, ρ is the penalty coefficient, ε is a random number that satisfies a uniform distribution on the interval [0,1].

[0066] S2. Scenery scene reduction based on K-means++.

[0067] This step is to construct typical wind and solar uncertainty scenarios. The wind and solar scene generation method based on WGAN-GP and the K-means++ method are used for scene clustering to obtain typical wind and solar scenes and their probabilities, and the obtained typical wind and solar scenes are used as the initial scenarios for distributed robust optimization.

[0068] In this embodiment, the K-means++ algorithm is used to cluster a large number of scenery scenes generated by WGAN-GP. Since the number of clusters has a great influence on the clustering effect, scheduling model construction and computing efficiency, it is necessary to determine the optimal number of clusters.

[0069] To this end, typical internal validity indicators such as Davis-Bourdin Index (DBI) and Silhouette Coefficient (SC) are introduced for quantitative analysis, both of which are widely used in multi-wind farm correlation and load curve classification.

[0070] Assuming that the k samples in the data set are clustered into three categories, the DBI is expressed as follows:

[0071]

[0072] Where DBI(i) is the Davis-Bourdin index of the ith cluster, N is the total number of clusters, and C i , C j Represent the centers of the i-th and j-th clusters respectively; S i , S j Respectively represent all samples in cluster i, j to cluster center C i , C j The average distance between clusters is also called the intra-cluster distance; d(C i ,C j ) represents the cluster center C i and C j The distance between.

[0073] SC stands for:

[0074]

[0075] Among them, s(u) represents the silhouette coefficient of the u-th sample, b(u) represents the average distance between the u-th sample and other samples outside the cluster, and a(u) is the average distance between the u-th sample and all samples in the cluster.

[0076] S3. Establish a two-stage distributed optimization model considering the uncertainty of wind and solar power.

[0077] The first stage of this two-stage distributed robust optimization model is the pre-scheduling stage, that is, the uncertainty of wind and solar power is not considered, and the second stage solves the optimal scheduling plan under the worst scenario.

[0078] First, the first-stage model is established. The first stage is the pre-dispatch stage, which takes into account the cost of purchasing electricity from the external power grid and the operating cost of pumped storage, and optimizes the dispatch decision of the complementary power generation system according to the predicted wind and solar output, in order to achieve the best cost-effectiveness.

[0079] Secondly, the second-stage model is established. The second stage considers the impact of wind and solar uncertainty on system operation. Under the premise of the first stage, the unit output is adjusted according to the actual wind and solar output to solve the optimal scheduling plan under the worst scenario.

[0080] The two-stage distributed robust optimization model is expressed in matrix as:

[0081]

[0082] I k ≤Qε

[0083] Fx+Gy≤l

[0084] In the formula, x is the decision variable of the first stage, including the cost of purchasing electricity from the external power grid and the operating cost of pumped storage; c represents the corresponding cost coefficient; y k is the flexible adjustment variable in the second stage under scene k; ε represents the prediction vector of scenery; P k is the occurrence probability of scene k; N is the number of discrete scenes; D, H, I, Q, F, G, a, b, e, r, l are the correlation coefficient matrix and constant column vector.

[0085] Construct a set of confidence levels for a probability distribution:

[0086]

[0087] In the formula, is the initial probability value of scene k; θ1 and θ ∞ are the probability allowable deviation limits under the corresponding constraints of 1-norm and ∞-norm respectively.

[0088] Among them, the probability allowed deviation values ​​θ1 and θ ∞ It is expressed as:

[0089]

[0090] The comprehensive norm confidence interval contains absolute value constraints, which need to be linearized and 0-1 auxiliary variables introduced for equivalent transformation. The 1-norm transformation is as follows:

[0091]

[0092] In the formula, and They are the probability distribution p k Positive and negative shifts relative to the original probability distribution; and 0-1 flags for positive and negative offsets respectively.

[0093] Similarly, the ∞-norm is linearized to:

[0094]

[0095] Finally, p k is the probability distribution of the actual scenario.

[0096] S4, C&CG algorithm solution.

[0097] The C&CG algorithm (Cutting-and-Conquering Algorithm) is used to solve the two-stage distributed robust optimization model. Auxiliary variables and constraints related to sub-problems are continuously introduced during the solution process to improve the solution efficiency.

[0098] For the proposed two-stage distributed robust optimization model, the C&CG algorithm is used to decompose it into a main problem and sub-problems for iterative solution. The C&CG algorithm flow chart is shown in Figure 3 The main problem can be described as:

[0099]

[0100] The subproblem is to solve the main problem given x * On this basis, we find the worst scenario probability distribution within the confidence interval, provide iterative operations for the main problem, and obtain the lower bound of the model. The sub-problem can be expressed as:

[0101]

[0102] In the formula, “*” indicates the optimal solution; J indicates the number of iterations;

[0103] Since the constraints of the inner and outer min-max optimization problems are unrelated, they can be solved in the order of inner first and outer later, as shown in the following formula:

[0104]

[0105] This example uses the historical data of a pumped storage power station. Combined with the regional wind and solar power generation data, a wind, solar and pumped storage combined system operation simulation model is established. 24 hours is selected as a cycle, and according to the load demand and weather data, an 850kW wind turbine and a 0.2kW / piece photovoltaic cell are selected.

[0106] In order to verify the effectiveness of the proposed model, this paper sets up the following two scenarios for comparative analysis.

[0107] Case 1: There is no pumped storage power station, and wind power and photovoltaic power generation operate independently.

[0108] Case 2: There is a pumped storage power station, which uses a wind-solar-pumped storage combined system for economic operation.

[0109] The embodiment of the present invention designs an optimization scheduling method for a pumped storage complementary power generation system taking into account the uncertainty of wind and solar power. Under the background of a high proportion of renewable energy access to the power system, a distributed robust optimization model of a complementary power generation system consisting of wind power, photovoltaic power, and pumped storage power stations is established. The system structure diagram is shown in FIG. Figure 4 As shown in the figure. In the first stage, the cost of purchasing electricity from the external power grid and the operating cost of pumped storage are considered, and the system dispatch decision is optimized according to the predicted wind and solar output. In the second stage, based on the premise of the first stage, the unit output is adjusted according to the actual wind and solar output to solve the optimal dispatch plan under the worst scenario. The simulation results are shown in the figure. Figure 5 , Figure 6 As shown in the figure, the simulation results show that this design makes the originally fluctuating wind power and photovoltaic power generation become a stable and dispatchable multi-source energy storage system, smoothing load fluctuations and achieving peak load shifting. In addition, the system's wind and solar power abandonment problems are alleviated, and the ability to absorb renewable energy is effectively improved.

[0110] In summary, the complementary power generation system optimization scheduling method designed in the embodiment of the present invention combines pumped storage power stations with renewable energy power generation, which can smooth out the output fluctuations of wind and solar power. At the same time, considering the uncertainty of wind and solar power output, the complementary power generation system including pumped storage is optimized and scheduled, thereby improving the operation safety, economy and stability of the system. The effectiveness of the present invention is verified through example simulation.

[0111] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0112] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An optimization scheduling method for a wind-solar-pumped storage complementary power generation system, characterized in that: The following steps are involved: Obtain wind and solar data including pumped storage complementary power generation system; Based on the wind and solar data, the WGAN-GP network model is used to generate wind and solar scenes, and the K-means++ method is used to reduce the wind and solar scenes to obtain typical wind and solar scenes and their probabilities; A two-stage distributed blue-robust optimization model considering the uncertainty of wind and solar power is established, and the C&CG algorithm is used to solve the two-stage distributed blue-robust optimization model to obtain the optimal scheduling scheme. In the solution process, the typical wind and solar power scenes and their probabilities are used as the initial scenes and initial probabilities of the two-stage distributed blue-robust optimization model.

2. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 1 is characterized in that: The objective function used in the WGAN-GP network model training process is: In the formula, G and D are the generator and discriminator respectively, E(·) is the expected function, G(·) is the generating function, and D(·) is the discriminator function. is the penalty term, ||·||2 represents the 2-norm, ρ is the penalty coefficient, ε is a random number that satisfies a uniform distribution on the interval [0,1], P d is the probability distribution of the real historical data x of Fengguang, P z is the probability distribution of random noise z.

3. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 1 is characterized in that: In the process of scene reduction of scenery scenes using the K-means++ method, the Davis-Bourdin index and the silhouette coefficient are introduced for quantitative calculation to determine the optimal number of clusters.

4. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 3 is characterized in that: The calculation expression of the Davis-Bourdin index is: Where DBI(i) is the Davis-Bourdin index of the ith cluster, N is the total number of clusters, and C i , C j Represent the centers of the i-th and j-th clusters respectively; S i , S j Respectively represent all samples in cluster i, j to cluster center C i , C j The average distance between clusters is also called the intra-cluster distance; d(C i ,C j ) represents the cluster center C i and C j The distance between.

5. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 3 is characterized in that: The calculation expression of the silhouette coefficient is: Where s(u) represents the silhouette coefficient of the u-th sample, b(u) represents the average distance between the u-th sample and other samples outside the cluster, and a(u) is the average distance between the u-th sample and all samples in the cluster.

6. The optimization scheduling method of a wind-solar-pumped-storage complementary power generation system according to claim 1 is characterized in that: The two-stage distributed robust optimization model considering the uncertainty of wind and solar power includes two stages. The first stage considers the cost of purchasing electricity from the external power grid and the operating cost of pumped storage, and optimizes the scheduling strategy according to the predicted wind and solar power output in order to achieve the best cost-effectiveness. The second stage considers the uncertainty of wind and solar power. On the basis of the first stage, the unit output is adjusted according to the actual wind and solar power output to solve the optimal scheduling plan under the worst scenario.

7. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 6 is characterized in that: The expression of the two-stage distributed robust optimization model is: Ouch k ≤Qε Fx+Gy≤l in: In the formula, x is the decision variable of the first stage, including the cost of purchasing electricity from the external power grid and the operating cost of pumped storage, y k is the flexible adjustment variable in the second stage under scenario k, c represents the corresponding cost coefficient, ε represents the prediction vector of wind and solar, P k is the occurrence probability of scene k, N is the number of discrete scenes, D, H, I, Q, F, G, a, b, e, r, l are the correlation coefficient matrix and constant column vector, is the initial probability value of scene k, θ1 and θ ∞ are the probability allowable deviation limits under the corresponding constraints of 1-norm and ∞-norm, M is the number of generated samples, α1, α ∞ is the uncertainty confidence.

8. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 7 is characterized in that: By introducing 0-1 auxiliary variables to perform equivalent transformation, the 1-norm and ∞-norm corresponding constraints are linearized respectively, wherein the 1-norm corresponding constraint is transformed as follows: The corresponding constraint transformation of the ∞-norm is as follows: In the formula, and They are the probability distribution p k Positive and negative shifts relative to the original probability distribution, and 0-1 flags for positive and negative offsets respectively.

9. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 1, characterized in that: The step of using the C&CG algorithm to solve the two-stage distributed robust optimization model includes: (1) Decomposing the two-stage robust optimization model into a main problem and sub-problems; (2) Initialize uncertain variables; (3) Solve the main problem to obtain the optimal solution of the current iteration and update the new lower bound; (4) Solve the subproblem to obtain the worst scenario probability distribution of the current iteration and update the new upper bound; (5) Determine whether the convergence condition is met. If so, output the optimal solution and the iteration ends. If not, add auxiliary variables and related constraints and return to step (3) for iterative calculation until the convergence condition is met and the optimal solution is output.

10. The optimization scheduling method of a wind-solar-pumped storage complementary power generation system according to claim 9, characterized in that: The main problem is described as: The sub-problem is described as: In the formula, c represents the corresponding cost coefficient, x is the decision variable of the first stage, including the cost of purchasing electricity from the external power grid and the operating cost of pumped storage, "*" represents the optimal solution, and a and b are respectively, is the probability value of the jth iteration of scenario k, J is the number of iterations, is the second-stage flexible adjustment variable at the jth iteration under scenario k, c represents the corresponding cost coefficient, represents the predicted vector of the scenery at the jth iteration under scene k, P k is the occurrence probability of scene k, N is the number of discrete scenes, y k is the flexible adjustment variable in the second stage under scenario k, ε k The predicted vector representing the scenery of scene k.

Images