Comprehensive energy system operation method based on spectrum normalization generative adversarial network

Abstract

The invention discloses a comprehensive energy system operation method based on a spectrum normalization generative adversarial network, and the method comprises the steps: obtaining a typical scene set of uncertainty variables, fusing the scene information of the uncertainty variables, formulating a robust uncertainty set through employing 1-norm constraint and box-type constraint, and carrying out the calculation of the robust uncertainty set. Constructing a three-stage robust optimization scheduling model taking the lowest operation cost as a target; and converting the three-stage robust optimization scheduling model into a single-layer mixed integer model, and performing solution by using a solver to obtain a scheduling scheme. According to the scheduling scheme, the influence of multiple uncertainties such as the superior power grid electricity price and the source load output on the economical efficiency and the stability of the integrated energy system can be reduced, and a decision maker can balance the economic efficiency and the conservative property of the IES by selecting proper uncertain parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of hybrid energy systems, and in particular to a method for operating an integrated energy system based on a spectral normalization generative adversarial network. Background Art

[0002] With the gradual depletion of fossil energy, efficient energy utilization has become the focus of research by experts and scholars in the energy field. The integrated energy system (IES) couples four types of energy, namely cold, heat, electricity and gas, to provide users with multiple load requirements such as cold, heat and electricity, which can achieve complementary utilization of energy, reduce system operating costs, improve energy utilization efficiency and reduce pollution emissions. For an integrated energy system that includes renewable energy, the existence of multiple uncertainties such as the randomness and intermittency of power generation and the volatility of various loads makes it impossible to accurately predict the power generation and load, and thus it is impossible to formulate an accurate scheduling plan for the system, which has a great impact on the stability and economy of the operation of the integrated energy system.

[0003] A large number of literatures have considered the uncertainty of source and load in the study of optimal scheduling of integrated energy systems and have achieved rich research results. For example, Zhang Jinhui, Wang Xu, Jiang Chuanwen, et al., Optimal Scheduling Method for Multi-grid Coupling Integrated Energy System Considering Traffic Flow Uncertainty [J]. Power System Technology, 2019, 43(9): 3081-3089, etc., considered the energy supply and demand imbalance caused by the spatiotemporal uncertainty of new energy vehicle operation when constructing the optimal scheduling model of the integrated energy system.

[0004] The above literature provides many valuable research methods for the uncertainty problems in the operation of integrated energy systems and has achieved rich research results. However, in the existing research results, generally only the uncertainty of a single source load in the system is considered or only one method is used to deal with multiple uncertainties, while the characteristics of the uncertainties such as renewable energy generation and load forecasting errors in the system are ignored, which is somewhat one-sided. In fact, when there is randomness in renewable energy generation, its probability distribution function is determined, so it is suitable to be processed by random methods; when there are errors in load forecasting, its probability distribution and membership relationship are difficult to determine, so it is suitable to be processed by interval methods or robust optimization methods. Compared with interval methods, robust optimization methods generally focus on extreme cases, require constraints to be met in the worst case, and the optimization results obtained are often too conservative. Summary of the invention

[0005] The purpose of the present invention is to provide a method for operating an integrated energy system based on a spectral normalization generative adversarial network to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: The integrated energy system operation method based on spectral normalization generative adversarial network includes: Obtain a typical scenario set of uncertain variables, integrate the scenario information of uncertain variables, use 1-norm constraints and box constraints to formulate a robust uncertainty set, and build a three-stage robust optimization scheduling model with the goal of minimizing operating costs; The three-stage robust optimization scheduling model is transformed into a single-layer mixed integer model and solved using a solver to obtain a scheduling solution.

[0007] Furthermore, a typical scenario set of uncertain variables is obtained, the scenario information of uncertain variables is integrated, and the 1-norm constraint and box constraint are used to formulate a robust uncertainty set. Specifically, the spectral norm is introduced to normalize the weight matrix of the discriminator in WGAN-GP to constrain the discriminant performance, and the global 1-Lipschitz constraint is realized by calculating the maximum singular value of the weight matrix of each layer of the discriminator.

[0008] Furthermore, the objective function of WGAN-GP is as follows: (1); Assume the input renewable energy data set X , the discriminator includes the discriminator , , , , , and ; , for any input X have: (2); The set of weights and biases of each layer of the label discriminator : (3); Then we get: (4); according to have to: (5); Weight Perform normalization; (6); For any X : (7); Use inequality (7) to calculate the maximum singular value of the weight matrix of each layer; in, for The spectral norm of is the unit distance, yes No. The weight of layer discrimination, yes No. The deviation of layer discrimination, is the total number of discriminator layers, yes The weight of the discriminant output, yes No. The dimension of the layer discriminant output, yes No. The weight of layer discrimination, yes The dimension of the first layer discriminant output, yes The weight of the first layer of discrimination, is the spectral norm of A, is the spectral norm of B, is the spectral norm of AB, The Wasserstein distance between the real scene and the generated scene To satisfy 1-Lipschitz continuity, For real scene The distribution law of is random noise The distribution probability of is the probability of generating a scenario, is the output of the discriminator, is the expected value of the input noise.

[0009] Furthermore, a three-stage robust optimization scheduling model with the goal of minimizing operating costs is constructed, which specifically includes a first-stage scheduling model, a second-stage scheduling model, and a third-stage scheduling model.

[0010] Furthermore, in the first stage scheduling model, the objective function is to minimize the 0-1 variable switching cost. , including the power exchange status between the integrated energy system and the upper power grid, the charging and discharging status of the energy storage equipment, and the objective function The expression is: (8); The constraints are: (9); (10); Where TE is the scheduling period, is the cost coefficient for 0-1 variable switching, is the depreciation cost coefficient of charging and discharging energy storage equipment, , Respectively in t Purchase and sales 0-1 variables at the moment, , Respectively in t The charging and discharging status of the energy storage device at each moment is a 0-1 variable.

[0011] Furthermore, in the second-stage dispatch model, the objective function is to minimize the power exchange cost between the integrated energy system and the upper power grid. , the objective function The expression is: (11); The constraints are: (12); (13); In the formula, , Respectively in t The purchase and sale prices of electricity exchanged between the integrated energy system and the upper power grid at all times, , Respectively in t The power purchased and sold by the integrated energy system and the upper power grid at all times. , They are respectively the maximum values ​​of the purchased power and sold power exchanged between the integrated energy system and the upper-level power grid.

[0012] Furthermore, in the third stage dispatch model, the objective function is to minimize the operating cost of the integrated energy system. , including gas purchase cost , Energy storage equipment operating costs Carbon trading costs , , and The expressions are: (14); (15); (16); The constraints include CHP-CCS-P2G system coupling model constraints, carbon trading constraints, energy storage equipment constraints, gas boiler model constraints, and integrated energy system power balance constraints; In the formula, The price of gas purchased from the upper grid for the integrated energy system. , , Respectively in t Natural gas consumed by cogeneration units, gas boilers and power hydrogen production units at all times.

[0013] Furthermore, the constraint expressions of the CHP-CCS-P2G system coupling model are as follows: (17); (18); (19); (20); (twenty one); (twenty two); (twenty three); (twenty four); (25); (26); (27); (28); In the formula, is the calorific value of natural gas, , are the power generation and heat generation efficiencies of the cogeneration unit, , Respectively in t The electrical power and thermal power output of the cogeneration unit at any given moment, , are the maximum values ​​of the electrical power and thermal power output of the cogeneration unit, is the carbon emission conversion coefficient of the cogeneration unit, For t The carbon emission value of the cogeneration unit at each moment, For t The value of CO2 captured by the carbon capture unit at the moment; The power consumption factor for capturing CO2 in the carbon capture unit, For t The electrical power consumed by the carbon capture unit to capture CO2 at any given moment, The conversion factor between the amount of CO2 required to generate natural gas for the power-to-hydrogen unit and the power consumption, For tThe value of CO2 consumed by the power hydrogen production unit at the moment, For t The electric power consumed by the hydrogen production unit at the moment, The efficiency of the power-to-hydrogen unit to convert natural gas, For t Net output of the CHP-CCS-P2G system operating in coordination at all times.

[0014] Furthermore, after CCS-P2G absorption and conversion, the actual carbon emissions of the CHP-CCS-P2G system and the total carbon emissions of the microgrid It is expressed as the following formula: (29); (30); In the formula, , is the carbon emission coefficient of the gas boiler and the upper power grid, The electric power purchased by the integrated energy system from the superior power grid.

[0015] Furthermore, the carbon trading constraint condition expression is as follows: (31); (32); In the formula, is the electrothermal conversion coefficient, , The carbon emission coefficients of gas-fired units and coal-fired units, respectively, and the carbon quota of the integrated energy system in the carbon trading market , Selling carbon allowances on behalf of the integrated energy system (or free allowances allocated to the integrated energy system).

[0016] Furthermore, the energy storage device constraint expression is as follows: (33); (34); (35); (36); In the formula, , Respectively in t and t- The capacity of the energy storage device at 1 moment, , Respectively in tThe charging and discharging power of energy storage devices at all times. Energy storage devices include ESS and HSD. , are the charging and discharging efficiency of the energy storage device, , are the upper and lower limits of the energy storage device capacity, , are the maximum values ​​of charging and discharging power of the energy storage device respectively.

[0017] Furthermore, the constraint condition expression of the gas boiler model is as follows: (37); (38); In the formula, is the thermal power output of the gas boiler, For gas boiler operating efficiency, It is the upper limit of the thermal power output of the gas boiler.

[0018] Furthermore, the power balance constraint of the integrated energy system includes the electric power balance expression as follows: (39); And the thermal power balance, the expression is as follows: (40); In the formula, For t The output of the wind turbine at any moment, , Respectively in t The charging and discharging power of the energy storage unit at every moment, is the electrical load, , Respectively in t The charging and discharging power of the thermal energy storage unit at all times, For t When heat load.

[0019] Furthermore, a simplification method combining double-layer CCG, Karush-Kuhn-Tucker conditions, and big-M method is adopted to transform the three-stage robust optimization scheduling model into a single-layer mixed integer model, and then the solver is used to solve it to obtain a scheduling scheme, which specifically includes: obtaining a three-stage robust optimization scheduling model based on multiple uncertainty processing of polyhedral uncertainty sets, decomposing and reconstructing the three-stage robust optimization scheduling model to obtain a single-layer mixed integer model, and then solving it through the GUROBI solver to obtain a scheduling scheme.

[0020] Furthermore, the expression of the three-stage robust optimization scheduling model based on the multiple uncertainty processing of the polyhedral uncertainty set is as follows: (41); (42); In the formula, is the uncertainty set of the power interaction price between the integrated energy system and the upper power grid in the second stage, is the uncertainty set of sources and loads in the third stage of integrated energy system scheduling, , are electricity price and source load uncertainty variables, and are the expected values ​​of electricity price and source load uncertainty variables, , are the maximum and minimum values ​​of the uncertainty variable of electricity price, , They are The maximum and minimum values ​​of the uncertain 0-1 variable, , are the maximum and minimum values ​​of the source load uncertainty variables, , They are The maximum and minimum values ​​of the uncertain 0-1 variable, , They are and The uncertainty control parameters, is the electrical load, For t When heat load, , Respectively in t The purchase and sale prices of electricity exchanged between the integrated energy system and the upper power grid at all times, For t The output of the wind turbine at any moment.

[0021] Furthermore, the full matrix formula of the three-stage robust optimization scheduling model is as follows: (43); (44); Where TE is the scheduling period, and the coefficient matrix of the first stage constraint condition is , , , the coefficient matrix of the second stage constraints , , , , the coefficient matrix of the third stage constraints , , , , , , , the sets of decision variables in the first, second and third stages are , , , , Respectively in t Purchase and sales 0-1 variables at the moment, , Respectively in t The charging and discharging status of the energy storage device at the moment 0-1 variable, , Respectively in t The power purchased and sold by the integrated energy system and the upper power grid at all times. , Respectively in t The electrical power and thermal power output of the cogeneration unit at any given moment, For t The electrical power consumed by the carbon capture unit to capture CO2 at any given moment, For t The electric power consumed by the hydrogen production unit at the moment, , Respectively in t The charging and discharging power of the energy storage device at all times, It is the thermal power output of gas boiler.

[0022] Furthermore, the three-stage robust optimization scheduling model is decomposed and reconstructed to obtain a single-layer mixed integer model, which specifically includes: using the inner and outer nested CCG algorithm to decompose and reconstruct the three-stage robust optimization scheduling model into the main problem MP and the sub-problem SP, and the expression is as follows: (45); (46); After decomposition by the CCG algorithm, the multi-layer nested structure of the three-stage robust optimization scheduling model is transformed into a single-layer MILP structure using the KKT condition, thereby obtaining a single-layer mixed integer model. The specific transformation process is as follows: MP is expressed as (47); SP is expressed as (48); The main problem goal of the decomposed sub-problem is to solve the transaction power between the integrated energy system and the upper power grid under the worst transaction electricity price. The sub-problem goal of the sub-problem is to solve the unit dispatch plan under the worst integrated energy system source-load scenario. The expression is as follows: (49); (50); In the formula, is the worst solution to the uncertainty element obtained from the subproblem, and the auxiliary variable , auxiliary variables .

[0023] Furthermore, the GUROBI solver is used to solve the problem, and the scheduling solution includes: 1) Initialize the outer CCG algorithm problem, and initialize the upper and lower bounds of the outer layer as , , initialize the number of iterations , determine the convergence termination condition threshold ; 2) Solve the main problem of the outer CCG algorithm and obtain the optimal solution , update the lower bound of the outer problem , check whether the convergence condition is met ,in , if the convergence condition is met, the solution is terminated; otherwise, the solution continues; 3) Solve the inner CCG algorithm problem and initialize the inner CCG algorithm problem. The upper and lower bounds are initialized as , , initialize the number of iterations , determine the convergence termination condition threshold ; 4) Solve the main problem of the inner CCG algorithm and obtain the optimal solution , update the lower bound of the inner problem ; 5) Solve the inner CCG algorithm sub-problem and obtain the optimal solution , update the upper bound of the inner problem , check whether the convergence condition is met , if the convergence condition is met, terminate the inner problem and jump to 3); otherwise 6) continue to solve; 6) Add the variables and constraints of the second and third stages to the outer CCG algorithm main problem, and initialize the number of iterations Add one and return to 2).

[0024] To achieve the above object, the present invention also provides the following technical solutions: Integrated energy system operation equipment based on spectral normalization generative adversarial network, including: The construction module is used to obtain a typical scenario set of uncertain variables, integrate the scenario information of uncertain variables, use 1-norm constraints and box constraints to formulate a robust uncertainty set, and build a three-stage robust optimization scheduling model with the goal of minimizing operating costs; The scheduling module is used to transform the three-stage robust optimization scheduling model into a single-layer mixed integer model by using a simplification method combining the double-layer CCG algorithm, the Karush-Kuhn-Tucker condition, and the big-M method, and solve it using the solver to obtain a scheduling plan.

[0025] To achieve the above object, the present invention also provides the following technical solutions: A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0026] To achieve the above object, the present invention also provides the following technical solutions: A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any one of the methods described above.

[0027] Compared with the prior art, the present invention has the following beneficial effects: In this paper, a three-stage robust scheduling method is proposed to mitigate the impact of multiple uncertainties of upper grid electricity prices and source (load) on the economic efficiency and safety of IES. In order to manage multiple uncertainties, the spectral normalized generative adversarial network (SNWGAN) and K-means++ clustering are used to generate initial scenarios of uncertainty variables. Based on these initial scenarios, robust polyhedral uncertainty sets are constructed using 1-norm and box constraints, and a three-stage robust optimization scheduling model with the goal of minimizing operating costs is developed. Then, the model is simplified using a combination of double-layer column and constraint generation (CCG), Karush-Kuhn-Tucker (KKT) conditions and Big-M method, and directly solved using the commercial solver GUROBI.

[0028] The main conclusions are as follows: 1) Scenario generation and reduction: SNWGAN and K-means++ clustering algorithms are used to generate and reduce initial scenarios of uncertain variables. This method effectively uses historical datasets to generate scenarios that are similar but not identical to historical data.

[0029] 2) Enhanced robustness: Compared with traditional deterministic optimization models and other robust optimization (RO) models, the proposed three-stage RO model enhances the system's ability to mitigate uncertainty and is more in line with actual scheduling scenarios. In addition, the adopted algorithm can solve the optimal model solution in a shorter time (22.81 seconds).

[0030] 3) Sensitivity analysis: Through the sensitivity analysis of uncertainty parameters, it is found that as the uncertainty parameters increase, the total optimization cost of the system increases and the robustness of the scheduling plan increases. Decision makers can balance the economic efficiency and conservatism of IES by selecting appropriate uncertainty parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The figure is a flow chart of the method of the present invention.

[0032] Figure 2 This is the IES structure diagram.

[0033] Figure 3 It is a diagram of the solution process of the double-layer CCG algorithm of the present invention.

[0034] Figure 4 The Wasserstein distance is used to evaluate the generated scene graph in the present invention.

[0035] Figure 5 The present invention adopts the Fleche perceptual distance to evaluate and generate the scene graph.

[0036] Figure 6 The t-SNE distribution is used to evaluate the generated scene graph in the present invention.

[0037] Figure 7 The Davidson-Dinburg index coefficient of the present invention is used to calculate the similarity graph between each cluster.

[0038] Figure 8 FIG. 4 is a diagram showing the power output optimization results of the IES of the present invention.

[0039] Fig. 9 This is a diagram showing the heat output optimization results of the IES of the present invention.

[0040] Fig.10 This is a diagram of electricity purchase prices during the worst scenario analysis of the present invention.

[0041] Fig.11 This is the electricity price diagram for the worst scenario analysis of the present invention.

[0042] Fig.12 This is the worst output and electricity price distribution diagram for the worst scenario of source load in the first stage of the present invention.

[0043] Fig.13This is the worst output and electricity price distribution diagram for the worst scenario of source load in the second stage of the present invention.

[0044] Fig.14 This is the worst output and electricity price distribution diagram for the worst scenario of source load in the third stage of the present invention.

[0045] Fig.15 This is a diagram showing the influence of the uncertain parameters of the present invention on the total cost of optimizing the energy base.

[0046] Fig.16 This is a diagram of the equipment model of the method of the present invention.

[0047] Fig.17 It is a diagram of the internal structure of the computer device of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] See also Figures 1 to 17 , the present invention provides a technical solution: In order to reduce the impact of multiple uncertainties such as the electricity price of the upper power grid and the output of sources and loads on the economy and stability of the integrated energy system, a scenario generation method based on spectral normalization generative adversarial network is proposed. First, a typical scenario set of uncertain variables is obtained; second, the scenario information of uncertain variables is integrated, and the 1-norm constraint and box constraint are used to formulate a robust uncertainty set, and a three-stage robust optimization scheduling model with the lowest operating cost as the goal is constructed. Then, a simplification method combining double-layer CCG, Karush-Kuhn-Tucker conditions, and big M method is used to transform the three-stage robust model into a single-layer mixed integer model, and the commercial solver GUROBI is used to solve it. Finally, the effectiveness of the proposed model and method is verified through example simulation. CCG: column and constraint algorithm, Karush-Kuhn-Tucker: Karush Kuhn Tucker.

[0050] 1. Typical scene generation based on SNWGAN and K-means++ clustering 1.1 SNWGAN WGAN-GP is one of the most important derivative models of GAN. It uses the Gradient Penalty (GP) gradient penalty term to make the Wasserstein distance satisfy the Lipschitz continuity condition. Lipschitz continuity ensures the existence of the optimal discriminant function and the unique Nash equilibrium, which is crucial for the stable training of the model.

[0051] WGAN-GP consists of two neural networks: a generator and G ), and the other is the discriminator (discriminator, D ); G Using random noise as input, it is used to mine the intrinsic characteristics of real sample data, thereby generating a large number of new data samples similar to the real data; D Taking real data and generated data as input, the goal is to distinguish real data from generated data as much as possible. The objective function of WGAN-GP is as follows.

[0052] (1); However, the GP term introduced by WGAN-GP can only make the Wasserstein distance approximately satisfy the Lipschitz continuity, which makes it difficult to achieve global constraints. This paper introduces the spectral norm to normalize the weight matrix in the discriminant network to further constrain its discriminant performance.

[0053] Assume the input is X , the discriminator representation , for any input X Both: (2); Overall label discriminator weights and biases: (3); Then we can get: (4); according to We can get: (5); Normalize the weights: (6); In summary, for any X : (7); Using the above inequality, the global 1-Lipschitz constraint can be implemented by simply calculating the maximum singular value of the weight matrix of each layer.

[0054] for The spectral norm of is the weight of each layer of the discriminator, is the bias of each layer of the discriminator, is the lth layer of the discriminator; Total number of discriminator layers; is the discriminator output degree matrix, The Wasserstein distance between the real scene and the generated scene To satisfy 1-Lipschitz continuity, For real scene The distribution probability is random noise The distribution probability of is the distribution probability of generating the scene.

[0055] 2. Integrated energy system structure and three-stage RO scheduling model 2.1 Integrated energy system structure The integrated energy system constructed in this paper mainly includes combined heating and power (CHP) unit, power to gas (P2G) unit, carbon capture and storage (CCS) unit, wind turbine (WT), gas boiler (GB), electric energy storage (ESS), thermal energy storage (HSD), Superior Grid (SG), Natural Gas Network (NSN) and electric load (EL), thermal load (HL), as shown in the following figure. Figure 2 shown.

[0056] 2.2 Three-stage RO scheduling model for integrated energy system 2.2.1 First-stage scheduling model 1) Objective function In the first stage, we take the minimization of the 0-1 variable state switching cost as the objective function, including the power exchange state between IES and the upper grid, and the charging and discharging state of the energy storage device. The objective function expression is: (8); 2) Constraints (9); (10); 2.2.2 Second-stage scheduling model 1) Objective function In the second stage, we take the minimization of the power exchange cost between IES and the upper power grid as the objective function, and the objective function expression is: (11); , They are t The purchase and sale prices of electricity when IES exchanges power with the upper-level power grid at each moment; , They are t At each moment, IES exchanges purchased power and sold power with the upper-level power grid.

[0057] 2) Constraints (12); (13); 2.2.3 The third stage scheduling model 1) Objective function The third stage takes minimizing IES operating costs as the objective function, including gas purchase costs , Energy storage operating costs Carbon trading costs ( The specific calculation process is detailed in (b) of 2) constraints.

[0058] (14); (15); (16); In the formula, The gas price that IES pays for purchasing gas from the parent grid; is the depreciation cost coefficient of energy storage charging and discharging; 2) Constraints (a) Constraints of the CHP-CCS-P2G system coupling model (17); (18); (19); (20); (twenty one); (twenty two); (twenty three); (twenty four); (25); (26); (27); (28); In the formula, Natural gas consumed by CHP units; is the calorific value of natural gas, , They are the power generation and heat generation efficiency of the CHP unit respectively; , are the output electric power and thermal power of the CHP unit respectively; , are the maximum values ​​of the electrical power and thermal power output of the CHP unit, respectively; is the carbon emission conversion coefficient of the CHP unit; is the carbon emission value of the CHP unit; The value of CO2 captured for CCS; The electricity consumption factor for capturing CO2 for the CCS system; The electrical power consumed to capture CO2 for CCS; The conversion factor between the amount of CO2 required to generate natural gas for P2G and the electricity consumption; The value of CO2 consumed by P2G; The electrical power consumed by P2G; The efficiency of P2G conversion to natural gas; It is the net output of the CHP-CCS-P2G system operating in coordination.

[0059] After CCS-P2G absorption and conversion, the actual carbon emissions of CHP-CCS-P2G and the total carbon emissions of the microgrid It can be expressed as: (29); (30); In the formula, , is the carbon emission coefficient of GB and the upper-level power grid; It is the electric power purchased by the integrated energy system from the superior power grid (assuming that all the electric energy from the power grid comes from coal-fired units).

[0060] (b) Carbon trading constraints At present, the allocation mechanism of domestic carbon trading is mostly free allocation. The IES initial carbon emission quota model is as follows: (31); In the formula, for the free quotas allocated to the integrated energy system; , are the carbon emission coefficients of gas-fired units and coal-fired units respectively.

[0061] Therefore, the integrated energy system actually participates in the carbon quota in the carbon trading market. , a positive value represents that the integrated energy system purchases carbon quotas, and a negative value represents that the integrated energy system sells carbon quotas.

[0062] (32); (c) Energy storage equipment constraints (33); (34); (35); (36); (d) Gas boiler model and its constraints (37); (38); (e) IES power balance constraints IES power balance includes electrical power balance and thermal power balance, and the constraints are formula (39) and formula (40) respectively.

[0063] (39); (40); 3. Multiple uncertainty RO modeling and model solving 3.1 Multiple uncertainty processing based on polyhedral uncertainty sets Considering the uncertainty of the power price of the upper grid and the IES source load, a polyhedron uncertainty set is used to describe their fluctuation range. is the uncertainty set of the power interaction price between IES and the upper grid in the second stage, and is modeled as follows: (41); is the uncertainty set of source and load in the third stage IES scheduling, and the model is as follows: (42); 3.2 Full matrix formula of the three-stage RO model To facilitate the understanding of the following text, the full matrix formula of the above model is as follows: (43); (44); In the formula, , , is the coefficient matrix of the first stage constraints, , , , is the coefficient matrix of the second stage constraints, , , , , , , is the coefficient matrix of the third stage constraints, , , are the sets of decision variables for the three stages respectively.

[0064] 3.3 RO model decomposition and reconstruction The three-stage RO model constructed in this paper is a non-deterministic polynomial (NP)-hard problem and cannot be solved directly. In order to solve the computational challenge, we use the inner and outer nested CCG algorithm to decompose and reconstruct the original problem into the main problem (MP) and subproblems (SP), as shown below: (45); (46); In the formula, , is an auxiliary variable.

[0065] After decomposition by the CCG algorithm, the original problem is transformed into a nested convex optimization problem. The KKT condition can be used to transform the original multi-layer nested structure into a single-layer MILP problem, which can be directly solved by the GUROBI solver. The specific transformation process is shown in Equations (47)-(52).

[0066] The main problem is shown in formula (47): (47); * is the worst solution to the uncertainty element obtained from the subproblem.

[0067] The expression of the sub-problem is shown in formula (48).

[0068] (48); The subproblems after CCG transformation are still nested and cannot be solved directly, so the CCG algorithm is used to decompose the inner subproblems again: The main problem objective of the decomposed sub-problem is to solve the transaction power between IES and the upper power grid under the worst transaction electricity price, and the sub-problem objective of the sub-problem is to solve the unit dispatch plan under the worst IES source-load scenario, as shown in (49) and (50).

[0069] (49); (50); 3.4 RO model solution process The reconstructed and decomposed RO model can be solved by the GUROBI solver. The specific solution steps are as follows: 1) Initialize the outer CCG algorithm questions, and initialize the upper and lower bounds respectively , , initialize the number of iterations , determine the convergence termination condition threshold ; 2) Solve the main problem of the outer CCG and obtain the optimal solution , update the lower bound of the outer problem , check whether the convergence condition is met ,in , if the convergence condition is met, the solution is terminated; otherwise, the solution continues; 3) Solve the inner CCG problem and initialize the inner CCG algorithm problem. The upper and lower bounds are initialized as , , initialize the number of iterations , determine the convergence termination condition threshold ; 4) Solve the main problem of the inner CCG and obtain the optimal solution , update the lower bound of the inner problem ; 5) Solve the inner CCG subproblem and obtain the optimal solution , update the upper bound of the inner problem , check whether the convergence condition is met , if the convergence condition is met, terminate the inner problem and jump to 3); otherwise 6) continue to solve; 6) Add the variables and corresponding constraints of the second and third stages to the outer CCG main problem, and let , return to 2).

[0070] 4. Example Analysis 4.1 Example Setup This example uses a comprehensive energy system demonstration project in North China as the simulation object. The system unit parameters are detailed in Table 1, the electricity price forecast values ​​are shown in Table 2, and the unit operation coefficients in this article are from relevant literature. The SNWGAN model data comes from the two-year (January 1, 2022 to December 31, 2023) source-load measured historical data of the park. The data time interval is 1 hour, of which 80% of the output data is used as a training set and 20% of the data is used as a test set. The SNWGAN model in the example is built using the deep learning framework Pytorch, and the number of iterations is set to 20,000. The computer is configured with an Intel Core i7 processor with a main frequency of 1.8GHz and 16GB of memory.

[0071] Table 1 System parameters and coefficients

[0072] Table 2 Forecasted electricity prices

[0073] 4.2 SNWGAN Performance Verification To verify the quality of the scenes generated by SNWGAN, this section uses Wasserstein distance, Fréchet perceptual distance, and t-SNE to evaluate the generated scenes.

[0074] (1) Wasserstein distance: The goal of model training is to make the loss as small as possible, that is, to minimize the Wasserstein distance between the true distribution and the generated distribution.

[0075] (2) Fréchet perceptual distance: Calculates the distance between the true sample distribution and the generated sample distribution to measure the quality of the generated samples. The lower the score, the closer the two distributions are.

[0076] (3) t-SNE distribution: can be used to visualize high-dimensional data. It converts the similarity between data points into joint probabilities and attempts to minimize the KL divergence between the joint probabilities of low-dimensional embeddings and high-dimensional data.

[0077] The results of these three evaluation indicators are as follows Figures 4 to 6 As shown in Figure 2, it can be seen that the distribution of generated scenes is very close to the distribution of real scenes, which verifies the accuracy of the scenes generated by SNWGAN.

[0078] 4.3 Clustering Results The Davies-Bouldin Index (DBI) coefficient calculates the similarity between each cluster, and the average of all similarities is used to measure the clustering effect. The farther the distance between clusters, the smaller the DBI value, and the better the clustering result. Figure 7It can be seen that for WT output, the optimal number of clusters is 3; for EL, the optimal number of clusters is 2; for HL, the optimal number of clusters is 2. Typical scenarios of WT, EL, and HL using probability weighted calculation are shown in Figure 7 The initial expected value.

[0079] 4.4IES Optimal Scheduling Solution The three-stage robust optimization iteration process is shown in Table 3. According to Table 3, under the hardware conditions of this paper, the model reaches convergence after 2 outer iterations, the algorithm running time is 22.81s, and the solution speed meets the day-ahead scheduling requirements.

[0080] Table 3 Inner and outer nested CCG iterative solution process

[0081] After the iteration converges, the electrical and thermal distribution diagrams are as follows Figures 8 to 9 As shown. Figure 8 It can be found that the wind power output in IES plays a big role, indicating that the wind power consumption level is very high during the dispatching period. During the night [3:00-6:00] period, the load demand is low, the wind power output is high, the CHP unit output is significantly reduced, and the energy storage is charged to store excess electricity. During the peak load period [16:00:21:00], due to the "heat-to-electricity" operation mode, the power output of the CHP unit is limited, and the IES energy supply cannot meet the load demand. The shortfall is supplemented by the energy storage discharge.

[0082] like Fig. 9 As shown in Figure 1, the heat load is mainly provided by CHP, thermal energy storage and GB. When the heat energy exceeds the heat load demand, the excess heat energy can be stored and released when the heat energy is insufficient.

[0083] 4.5 Uncertainty analysis 4.5.1 Worst Scenario Analysis Figure 10 to Figure 11 is the worst scenario for the electricity price in the second stage. It can be seen that there is a certain error between the actual electricity purchase price obtained in the second stage robust optimization and the electricity purchase price predicted a day ago. From the worst electricity purchase price, it can be seen that the electricity price in the period of 20:00-03:00 is at the upper limit of the robust range. The electricity purchase price in these periods is higher than the generated electricity purchase price scenario. The actual electricity purchase cost of IES is higher than the predicted electricity purchase cost, which reflects that the IES operator has increased the operating cost in order to resist the impact of uncertainty. Similarly, there is some error between the actual electricity sales price optimized in the second stage and the electricity sales price predicted a day ago. The electricity sales prices of 12:00-13:00 and 19:00-21:00 are at the lower limit of the robust range, indicating that the actual electricity sales price is lower than the electricity sales price predicted at that time, because IES reduces the electricity sales revenue at that time to prevent uncertainty risks.

[0084] Figure 12 to Figure 14 This is the worst scenario for source load in the third stage. It can be seen that the worst output of wind power in the third stage is similar to the distribution of electricity sales prices. In most periods, it is located at the lower extreme of the fluctuation range. The actual output is lower than the expected value. The system needs to purchase additional electricity from the outside or increase the unit output to fill the difference between the actual output and the predicted output. Therefore, this scenario has the strongest robustness. The actual power of multi-energy load is higher than the expected value, which is exactly the opposite of the distribution results of wind power and electricity sales prices, but the impact is the same, that is, in order to resist the impact of robustness during fluctuation periods, the system needs to purchase additional electricity from the outside or increase the unit output.

[0085] 4.5.2 Comparison of Uncertain Optimization Methods In order to verify the advantages of the three-stage adjustable robust optimization method in this paper in dealing with uncertain problems, this section discusses four different uncertainty modeling methods. The control variable method is used, and the uncertainty variables are regarded as experimental variables. The other conditions are consistent with the initial settings of the example, as follows: Scenario 1: The three-stage robust optimization proposed in this paper considering the uncertainty of the upstream power grid electricity price and source load.

[0086] Scenario 2: Three-stage robust optimization of the robust model considering the uncertainty of upstream grid electricity prices.

[0087] Scenario 3: Three-stage robust optimization of the robust model considering source-load uncertainty.

[0088] Scenario 4: Using a deterministic optimization method, the predicted values ​​of the power price of the upper grid and the source load power are used as the optimization configuration data.

[0089] The IES operating cost in each scenario is shown in Table 4.

[0090] Table 4 IES operating cost results under each scenario

[0091] From Table 4, we can see that: Compared with Scenario 1, the total cost of Scenario 2 is reduced by 15.47%, among which the gas purchase cost and carbon trading cost are significantly reduced, by 13.5% and 32.71% respectively. This is because the uncertainty of source and load in IES is ignored, which greatly reduces the output of CHP units and reduces the gas purchase and carbon trading costs. Although Scenario 2 improves the economic efficiency of IES operation, the scheduling plan formulated is less compatible with the actual source and load output, exaggerating the system's ability to absorb new energy.

[0092] Compared with Scenario 1, the total cost of Scenario 3 has little change, among which the power exchange cost is reduced by 120.9584 yuan, the gas purchase cost is reduced by 19 yuan, and the carbon trading cost is reduced by 0.2318 yuan. It can be seen that considering the uncertainty of the upper-level power grid electricity price will increase the power interaction cost between IES and the upper-level power grid to offset the uncertainty of the electricity price, while having little impact on other costs in the system.

[0093] Compared with Scenario 1, Scenario 4 does not consider both uncertainties, which reduces the total operating cost of the system by 18.1%, including a 33.74% reduction in carbon trading costs, a 19.89% reduction in power exchange costs, and a 15.15% reduction in gas purchase costs. The system economy is significantly improved. However, this does not mean that deterministic optimization is better than uncertain optimization. This is because the uncertainty risk and re-dispatch cost caused by the forecast error of the upper-level power grid electricity price and source-load output are not considered in deterministic optimization, but in fact the uncertainty risk is underestimated, which will lead to higher actual dispatch costs.

[0094] 4.6 Sensitivity analysis of uncertain parameters In the robust optimization model, the magnitude of the uncertainty parameter represents the total duration of uncertainty occurring during the entire scheduling period. Therefore, it reflects the degree of interference of uncertainty on the robust optimization model. Fig.15 The influence of the uncertainty parameter size on the total optimization cost of IES is explained.

[0095] Fig.15 It is shown that as the uncertainty of electricity prices and source loads increases, the system operating costs also rise, indicating an increase in robustness. This positive correlation is due to the fact that a larger uncertainty parameter means that more uncertainty scenarios of transaction prices and wind / solar output are considered during the optimization process. Therefore, during the investment phase, the system continuously increases the installed capacity of storage, and during the dispatch process, the control cost increases to mitigate multiple uncertainties. This leads to an increase in the investment cost of the device and an increase in operating costs, resulting in an overall increase in total costs. The analysis shows that decision makers can effectively balance the economic efficiency and conservatism of IES by selecting appropriate uncertainty parameters.

[0096] 4 Conclusion In order to mitigate the impact of multiple uncertainties in upper grid electricity prices and source loads on the economic efficiency and security of IES, a three-stage robust scheduling method is proposed. In order to manage multiple uncertainties, spectral normalized generative adversarial networks (SNWGAN) and K-means++ clustering are used to generate initial scenarios of uncertainty variables. Based on these initial scenarios, robust polyhedral uncertainty sets are constructed using 1-norm and box constraints, and a three-stage robust optimization scheduling model with the objective of minimizing operating costs is developed. Then, the model is simplified using a combination of double-layer column and constraint generation (CCG), Karush-Kuhn-Tucker (KKT) conditions and Big-M method, and directly solved using the commercial solver GUROBI. The main conclusions are as follows: 1) Scenario generation and reduction: SNWGAN and K-means++ clustering algorithms are used to generate and reduce initial scenarios of uncertain variables. This method effectively uses historical datasets to generate scenarios that are similar but not identical to historical data.

[0097] 2) Enhanced robustness: Compared with traditional deterministic optimization models and other robust optimization (RO) models, the proposed three-stage RO model enhances the system's ability to mitigate uncertainty and is more in line with actual scheduling scenarios. In addition, the adopted algorithm solves the optimal model solution in 22.81 seconds.

[0098] 3) Sensitivity analysis: Through the sensitivity analysis of uncertainty parameters, it is found that as the uncertainty parameters increase, the total optimization cost of the system increases and the robustness of the scheduling plan increases. Decision makers can balance the economic efficiency and conservatism of IES by selecting appropriate uncertainty parameters.

[0099] like Fig.16 As shown, the integrated energy system operation device based on spectral normalization generative adversarial network includes a building module and a scheduling module; The construction module is used to obtain a typical scenario set of uncertain variables, integrate the scenario information of uncertain variables, use 1-norm constraints and box constraints to formulate a robust uncertainty set, and build a three-stage robust optimization scheduling model with the goal of minimizing operating costs; The scheduling module is used to transform the three-stage robust optimization scheduling model into a single-layer mixed integer model by using a simplification method that combines the double-layer CCG algorithm, the Karush-Kuhn-Tucker condition, and the big-M method, and use the solver to solve it to obtain a scheduling plan.

[0100] like Fig.17As shown, the proposed integrated energy system operation device based on spectral normalization generative adversarial network can adopt various types of computing devices such as desktop computers, notebooks, PDAs and cloud servers. This device is composed of multiple components such as processors and memories, but is not limited thereto and may include other components or different combinations. The device examples provided here are for illustrative purposes only and are not restrictive.

[0101] The processor can be various types of devices, such as CPU, DSP, ASIC, FPGA, etc., and can also include programmable logic devices, discrete gates, transistor logic devices and discrete hardware components, etc. The memory is used to store computer programs and modules, and the processor runs or executes the computer program modules stored in the memory and calls the data stored in the memory, thereby realizing various functions of the integrated energy system based on spectral normalization generative adversarial network.

[0102] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0103] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the method for operating an integrated energy system based on a spectral normalization generative adversarial network.

[0104] If the integrated energy system operation method based on spectral normalization generative adversarial network is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device.

[0105] Based on such understanding, the present invention implements all or part of the processes in the above-mentioned integrated energy system operation method based on spectral normalization generative adversarial network, and can also be completed by instructing related hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of the above-mentioned integrated energy system operation method based on spectral normalization generative adversarial network can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or preset intermediate form, etc.

[0106] Computer-readable storage media may refer to any entity or device that can carry computer program code, including recording media, USB flash drives, mobile hard drives, disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), software distribution media, etc.

[0107] It should be noted that the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable storage media do not include electrical carrier signals and telecommunication signals.

[0108] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for operating an integrated energy system based on a spectral normalized generative adversarial network, characterized in that: include: Obtain a typical scenario set of uncertain variables, integrate the scenario information of uncertain variables, use 1-norm constraints and box constraints to formulate a robust uncertainty set, and build a three-stage robust optimization scheduling model for operating costs; The three-stage robust optimization scheduling model is transformed into a single-layer mixed integer model and solved using a solver to obtain a scheduling solution.

2. The method according to claim 1, characterized in that A typical scene set of uncertain variables is obtained, the scene information of uncertain variables is integrated, and the 1-norm constraint and box constraint are used to formulate a robust uncertainty set. Specifically, the spectral norm is introduced to normalize the weight matrix of the discriminator in WGAN-GP to constrain the discriminant performance, and the global 1-Lipschitz constraint is realized by calculating the singular values ​​of the weight matrix of each layer of the discriminator.

3. The method according to claim 2, characterized in that The objective function of WGAN-GP is as follows: （1）； Assume the input renewable energy data set X , the discriminator includes the discriminator , , , , , and ; , for the input X have: （2）； The set of weights and biases of each layer of the label discriminator : （3）； get: （4）； according to have to: （5）； Weight Perform normalization; （6）； For any X : (7); Use inequality (7) to calculate the singular values ​​of the weight matrix of each layer; in, for The spectral norm of is the unit distance, yes No. The weight of layer discrimination, yes No. The deviation of layer discrimination, is the total number of discriminator layers, yes The weight of the discriminant output, yes No. The dimension of the layer discriminant output, yes No. The weight of layer discrimination, yes The dimension of the first layer discriminant output, yes The weight of the first layer of discrimination, is the spectral norm of A, is the spectral norm of B, is the spectral norm of AB, The Wasserstein distance between the real scene and the generated scene To satisfy 1-Lipschitz continuity, For real scene The distribution law of is random noise The distribution probability of is the probability of generating a scenario, is the output of the discriminator, is the expected value of the input noise.

4. The method according to claim 1, characterized in that The three-stage robust optimization scheduling model for constructing operating costs specifically includes the first-stage scheduling model, the second-stage scheduling model and the third-stage scheduling model.

5. The method according to claim 4, characterized in that In the first stage scheduling model, the 0-1 variable switching cost is used as the objective function , including the power exchange status between the integrated energy system and the upper power grid, the charging and discharging status of the energy storage equipment, and the objective function The expression is: （8）； The constraints are: （9）； （10）； Where TE is the scheduling period, is the cost coefficient for 0-1 variable switching, is the depreciation cost coefficient of charging and discharging energy storage equipment, , Respectively in t Purchase and sales 0-1 variables at the moment, , Respectively in t The charging and discharging status of the energy storage device at each moment is a 0-1 variable.

6. The method according to claim 5, characterized in that In the second stage dispatch model, the power exchange cost between the integrated energy system and the upper power grid is taken as the objective function. , the objective function The expression is: （11）； The constraints are: （12）； （13）； In the formula, , Respectively in t The purchase and sale prices of electricity exchanged between the integrated energy system and the upper power grid at all times, , Respectively in t The power purchased and sold by the integrated energy system and the upper power grid at all times. , They are respectively the maximum values ​​of the purchased power and sold power exchanged between the integrated energy system and the upper-level power grid.

7. The method according to claim 6, characterized in that In the third stage dispatch model, the operating cost of the integrated energy system is used as the objective function , including gas purchase cost , Energy storage equipment operating costs Carbon trading costs , , and The expressions are: （14）； （15）； （16）； The constraints include CHP-CCS-P2G system coupling model constraints, carbon trading constraints, energy storage equipment constraints, gas boiler model constraints, and integrated energy system power balance constraints; In the formula, The price of gas purchased from the upper grid for the integrated energy system. , , Respectively in t The natural gas consumed by the cogeneration unit, gas boiler and power hydrogen production unit at all times, , Respectively in t The charging and discharging power of the energy storage device at all times.

8. The method according to claim 7, characterized in that The constraint expressions of the CHP-CCS-P2G system coupling model are as follows: （17）； （18）； （19）； （20）； （21）； （22）； （23）； （24）； （25）； （26）； （27）； （28）； In the formula, is the calorific value of natural gas, , are the power generation and heat generation efficiencies of the cogeneration unit, , Respectively in t The electrical power and thermal power output of the cogeneration unit at any given moment, , are the maximum values ​​of the electrical power and thermal power output of the cogeneration unit, is the carbon emission conversion coefficient of the cogeneration unit, For t The carbon emission value of the cogeneration unit at each moment, For t The value of CO2 captured by the carbon capture unit at the moment; The power consumption factor for capturing CO2 in the carbon capture unit, For t The electrical power consumed by the carbon capture unit to capture CO2 at any given moment, The conversion factor between the amount of CO2 required to generate natural gas for the power-to-hydrogen unit and the power consumption, For t The value of CO2 consumed by the power hydrogen production unit at the moment, For t The electric power consumed by the hydrogen production unit at the moment, The efficiency of the power-to-hydrogen unit to convert natural gas, For t Net output of the CHP-CCS-P2G system operating in coordination at all times.

9. The method according to claim 8, characterized in that After CCS-P2G absorption and conversion, the actual carbon emissions of the CHP-CCS-P2G system and the total carbon emissions of the microgrid It is expressed as the following formula: （29）； （30）； In the formula, , is the carbon emission coefficient of the gas boiler and the upper power grid, The electric power purchased from the upper power grid for the integrated energy system. is the carbon emission value of the cogeneration unit, Natural gas consumed by gas boilers and power hydrogen production units, It is the purchased power exchanged between the integrated energy system and the upper-level power grid.

10. The method according to claim 9, characterized in that The carbon trading constraint condition expression is as follows: （31）； （32）； In the formula, is the electrothermal conversion coefficient, , The carbon emission coefficients of gas-fired units and coal-fired units, respectively, and the carbon quota of the integrated energy system in the carbon trading market , Selling carbon allowances on behalf of integrated energy systems, For t The thermal power output of the gas boiler at all times.

11. The method according to claim 10, characterized in that The energy storage device constraint expression is as follows: （33）； （34）； （35）； （36）； In the formula, , Respectively in t and t- The capacity of the energy storage device at 1 moment, , are the charging and discharging efficiency of the energy storage device, , are the upper and lower limits of the energy storage device capacity, , are the maximum values ​​of charging and discharging power of the energy storage device respectively.

12. The method according to claim 11, characterized in that The constraint expressions of the gas boiler model are as follows: （37）； （38）； In the formula, is the thermal power output of the gas boiler, For gas boiler operating efficiency, It is the upper limit of the thermal power output of the gas boiler.

13. The method according to claim 12, characterized in that The power balance constraint of the integrated energy system includes the electric power balance expression as follows: （39）； And the thermal power balance, the expression is as follows: （40）； In the formula, For t The output of the wind turbine at any moment, , Respectively in t The charging and discharging power of the energy storage unit at every moment, is the electrical load, , Respectively in t The charging and discharging power of the thermal energy storage unit at all times, For t When heat load.

14. The method according to claim 1, wherein: The three-stage robust optimization scheduling model is converted into a single-layer mixed integer model and solved by a solver to obtain a scheduling plan, which specifically includes: obtaining a three-stage robust optimization scheduling model based on multiple uncertainty processing of polyhedral uncertainty sets, decomposing and reconstructing the three-stage robust optimization scheduling model to obtain a single-layer mixed integer model, and then solving it through the GUROBI solver to obtain a scheduling plan.

15. The method according to claim 14, characterized in that The expression of the three-stage robust optimization scheduling model obtained by multiple uncertainty processing based on polyhedral uncertainty sets is as follows: （41）； （42）； In the formula, is the uncertainty set of the power interaction price between the integrated energy system and the upper power grid in the second stage, is the uncertainty set of sources and loads in the third stage of integrated energy system scheduling, , are electricity price and source load uncertainty variables, and are the expected values ​​of electricity price and source load uncertainty variables, , are the maximum and minimum values ​​of the uncertainty variable of electricity price, , They are The maximum and minimum values ​​of the uncertain 0-1 variable, , are the maximum and minimum values ​​of the source load uncertainty variables, , They are The maximum and minimum values ​​of the uncertain 0-1 variable, , They are and The uncertainty control parameter, is the electrical load, For t When heat load, , Respectively in t The purchase and sale prices of electricity exchanged between the integrated energy system and the upper power grid at all times, For t The output of the wind turbine at any moment.

16. The method according to claim 15, characterized in that The full matrix formula of the three-stage robust optimization scheduling model is as follows: （43）； （44）； Where TE is the scheduling period, and the coefficient matrix of the first stage constraint condition is , , , the coefficient matrix of the second stage constraints , , , , the coefficient matrix of the third stage constraints , , , , , , , the sets of decision variables in the first, second and third stages are , , , , Respectively in t Purchase and sales 0-1 variables at the moment, , Respectively in t The charging and discharging status of the energy storage device at the moment 0-1 variable, , Respectively in t The power purchased and sold by the integrated energy system and the upper power grid at all times. , Respectively in t The electrical power and thermal power output of the cogeneration unit at any given moment, For t The electrical power consumed by the carbon capture unit to capture CO2 at any given moment, For t The electric power consumed by the hydrogen production unit at the moment, , Respectively in t The charging and discharging power of the energy storage device at all times, It is the thermal power output of gas boiler.

17. The method according to claim 16, characterized in that The three-stage robust optimization scheduling model is decomposed and reconstructed to obtain a single-layer mixed integer model, which specifically includes: using the inner and outer nested CCG algorithm to decompose and reconstruct the three-stage robust optimization scheduling model into the main problem MP and the sub-problem SP, and the expression is as follows: （45）； （46）； After decomposition by the CCG algorithm, the multi-layer nested structure of the three-stage robust optimization scheduling model is transformed into a single-layer MILP structure using the KKT condition, thereby obtaining a single-layer mixed integer model. The specific transformation process is as follows: MP is expressed as (47); SP is expressed as (48); The main problem goal of the decomposed sub-problem is to solve the transaction power between the integrated energy system and the upper power grid under the worst transaction electricity price. The sub-problem goal of the sub-problem is to solve the unit dispatch plan under the worst integrated energy system source-load scenario. The expression is as follows: (49)； (50)； In the formula, is the worst solution to the uncertainty element obtained from the subproblem, and the auxiliary variable , auxiliary variables .

18. The method according to claim 17, characterized in that Then the GUROBI solver is used to solve the problem, and the scheduling solution includes: 1) Initialize the outer CCG algorithm problem, and initialize the upper and lower bounds of the outer layer as , , initialize the number of iterations , determine the convergence termination condition threshold ; 2) Solve the main problem of the outer CCG algorithm and obtain the optimal solution , update the lower bound of the outer problem , check whether the convergence condition is met ,in , if the convergence condition is met, the solution is terminated; otherwise, the solution continues; 3) Solve the inner CCG algorithm problem and initialize the inner CCG algorithm problem. The upper and lower bounds are initialized as , , initialize the number of iterations , determine the convergence termination condition threshold ; 4) Solve the main problem of the inner CCG algorithm and obtain the optimal solution , update the lower bound of the inner problem ; 5) Solve the inner CCG algorithm sub-problem and obtain the optimal solution , update the upper bound of the inner problem , check whether the convergence condition is met , if the convergence condition is met, terminate the inner problem and jump to 3); otherwise 6) continue to solve; 6) Add the variables and constraints of the second and third stages to the outer CCG algorithm main problem, and initialize the number of iterations Add one and return to 2).

19. Integrated energy system operation equipment based on spectral normalization generative adversarial network, characterized in that: include: The construction module is used to obtain a typical scenario set of uncertain variables, integrate the scenario information of uncertain variables, use 1-norm constraints and box constraints to formulate a robust uncertainty set, and build a three-stage robust optimization scheduling model for operating costs; The scheduling module transforms the three-stage robust optimization scheduling model into a single-layer mixed integer model and uses the solver to solve it to obtain the scheduling plan.

20. The device according to claim 19, characterized in that A typical scene set of uncertain variables is obtained, the scene information of uncertain variables is integrated, and the 1-norm constraint and box constraint are used to formulate a robust uncertainty set. Specifically, the spectral norm is introduced to normalize the weight matrix of the discriminator in WGAN-GP to constrain the discriminant performance, and the global 1-Lipschitz constraint is realized by calculating the singular values ​​of the weight matrix of each layer of the discriminator.

21. The device according to claim 20, characterized in that The objective function of WGAN-GP is as follows: （1）； Assume the input renewable energy data set X , the discriminator includes the discriminator , , , , , and ; , for the input X have: （2）； The set of weights and biases of each layer of the label discriminator : （3）； get: （4）； according to have to: （5）； Weight Normalize: （6）； For any X : （7）； Use inequality (7) to calculate the singular values ​​of the weight matrix of each layer; Among them, of Spectral norm, is the unit distance, yes No. The weight of layer discrimination, yes No. The deviation of layer discrimination, is the total number of discriminator layers, yes The weight of the discriminant output, yes No. The dimension of the layer discriminant output, yes No. The weight of layer discrimination, yes The dimension of the first layer discriminant output, yes The weight of the first layer of discrimination, yes The spectral norm of yes The spectral norm of yes The spectral norm of The Wasserstein distance between the real scene and the generated scene To satisfy 1-Lipschitz continuity, For real scene The distribution law of is random noise The distribution probability of is the probability of generating a scenario, is the output of the discriminator, is the expected value of the input noise.

22. The device according to claim 19, characterized in that The three-stage robust optimization scheduling model for constructing operating costs specifically includes the first-stage scheduling model, the second-stage scheduling model and the third-stage scheduling model.

23. The device according to claim 22, characterized in that In the first stage scheduling model, the 0-1 variable switching cost is used as the objective function , including the power exchange status between the integrated energy system and the upper power grid, the charging and discharging status of the energy storage equipment, and the objective function The expression is: （8）； The constraints are: （9）； （10）； Where TE is the scheduling period, is the cost coefficient for 0-1 variable switching, is the depreciation cost coefficient of charging and discharging energy storage equipment, , Respectively in t Purchase and sales 0-1 variables at the moment, , Respectively in t The charging and discharging status of the energy storage device at each moment is a 0-1 variable.

24. The device according to claim 23, characterized in that In the second stage dispatch model, the power exchange cost between the integrated energy system and the upper power grid is taken as the objective function. , the objective function The expression is: （11）； The constraints are: （12）； （13）； In the formula, , Respectively in t The purchase and sale prices of electricity exchanged between the integrated energy system and the upper power grid at all times, , Respectively in t The power purchased and sold by the integrated energy system and the upper power grid at all times. , They are respectively the maximum values ​​of the purchased power and sold power exchanged between the integrated energy system and the upper-level power grid.

25. The device according to claim 24, characterized in that In the third stage dispatch model, the operating cost of the integrated energy system is used as the objective function , including gas purchase cost , Energy storage equipment operating costs Carbon trading costs , , and The expressions are: （14）； （15）； （16）； The constraints include CHP-CCS-P2G system coupling model constraints, carbon trading constraints, energy storage equipment constraints, gas boiler model constraints, and integrated energy system power balance constraints; In the formula, The price of gas purchased from the upper grid for the integrated energy system. , , Respectively in t The natural gas consumed by the cogeneration unit, gas boiler and power hydrogen production unit at all times, , Respectively in t The charging and discharging power of the energy storage device at all times.

26. The device according to claim 25, characterized in that The constraint expressions of the CHP-CCS-P2G system coupling model are as follows: （17）； （18）； （19）； （20）； （21）； （22）； （23）； （24）； （25）； （26）； （27）； （28）； In the formula, is the calorific value of natural gas, , are the power generation and heat generation efficiencies of the cogeneration unit, , Respectively in t The electrical power and thermal power output of the cogeneration unit at any given moment, , are the maximum values ​​of the electrical power and thermal power output of the cogeneration unit, is the carbon emission conversion coefficient of the cogeneration unit, For t The carbon emission value of the cogeneration unit at each moment, For t The value of CO2 captured by the carbon capture unit at the moment; The power consumption factor for capturing CO2 in the carbon capture unit, For t The electrical power consumed by the carbon capture unit to capture CO2 at any given moment, The conversion factor between the amount of CO2 required to generate natural gas for the power-to-hydrogen unit and the power consumption, For t The value of CO2 consumed by the power hydrogen production unit at the moment, For t The electric power consumed by the hydrogen production unit at the moment, The efficiency of the power-to-hydrogen unit to convert natural gas, For t Net output of the CHP-CCS-P2G system operating in coordination at all times.

27. The device according to claim 26, characterized in that After CCS-P2G absorption and conversion, the actual carbon emissions of the CHP-CCS-P2G system and the total carbon emissions of the microgrid It is expressed as the following formula: （29）； （30）； In the formula, , is the carbon emission coefficient of the gas boiler and the upper power grid, The electric power purchased from the upper power grid for the integrated energy system. is the carbon emission value of the cogeneration unit, Natural gas consumed by gas boilers and power hydrogen production units, It is the purchased power exchanged between the integrated energy system and the upper-level power grid.

28. The device according to claim 27, characterized in that The carbon trading constraint condition expression is as follows: （31）； （32）； In the formula, is the electrothermal conversion coefficient, , The carbon emission coefficients of gas-fired units and coal-fired units, respectively, and the carbon quota of the integrated energy system in the carbon trading market , Selling carbon allowances on behalf of integrated energy systems, For t The thermal power output of the gas boiler at all times.

29. The device according to claim 28, characterized in that The energy storage device constraint expression is as follows: （33）； （34）； （35）； （36）； In the formula, , Respectively in t and t- The capacity of the energy storage device at 1 moment, , are the charging and discharging efficiency of the energy storage device, , are the upper and lower limits of the energy storage device capacity, , are the maximum values ​​of charging and discharging power of the energy storage device respectively.

30. The device according to claim 29, characterized in that The constraint expressions of the gas boiler model are as follows: （37）； （38）； In the formula, is the thermal power output of the gas boiler, For gas boiler operating efficiency, It is the upper limit of the thermal power output of the gas boiler.

31. The device according to claim 30, characterized in that The power balance constraint of the integrated energy system includes the electric power balance expression as follows: （39）； And the thermal power balance, the expression is as follows: （40）； In the formula, For t The output of the wind turbine at any moment, , Respectively in t The charging and discharging power of the energy storage unit at every moment, is the electrical load, , Respectively in t The charging and discharging power of the thermal energy storage unit at all times, For t When heat load.

32. The apparatus of claim 19, wherein: The three-stage robust optimization scheduling model is converted into a single-layer mixed integer model and solved by a solver to obtain a scheduling plan, which specifically includes: obtaining a three-stage robust optimization scheduling model based on multiple uncertainty processing of polyhedral uncertainty sets, decomposing and reconstructing the three-stage robust optimization scheduling model to obtain a single-layer mixed integer model, and then solving it through the GUROBI solver to obtain a scheduling plan.

33. The device according to claim 32, characterized in that The expression of the three-stage robust optimization scheduling model obtained by multiple uncertainty processing based on polyhedral uncertainty sets is as follows: （41）； （42）； In the formula, is the uncertainty set of the power interaction price between the integrated energy system and the upper power grid in the second stage, is the uncertainty set of sources and loads in the third stage of integrated energy system scheduling, , are electricity price and source load uncertainty variables, and are the expected values ​​of electricity price and source load uncertainty variables, , are the maximum and minimum values ​​of the uncertainty variable of electricity price, , They are The maximum and minimum values ​​of the uncertain 0-1 variable, , are the maximum and minimum values ​​of the source load uncertainty variables, , They are The maximum and minimum values ​​of the uncertain 0-1 variable, , They are and The uncertainty control parameter, is the electrical load, For t When heat load, , Respectively in t The purchase and sale prices of electricity exchanged between the integrated energy system and the upper power grid at all times, For t The output of the wind turbine at any moment.

34. The device according to claim 33, characterized in that The full matrix formula of the three-stage robust optimization scheduling model is as follows: （43）； （44）； Where TE is the scheduling period, and the coefficient matrix of the first stage constraint condition is , , , the coefficient matrix of the second stage constraints , , , , the coefficient matrix of the third stage constraints , , , , , , , the sets of decision variables in the first, second and third stages are , , , , Respectively in t Purchase and sales 0-1 variables at the moment, , Respectively in t The charging and discharging status of the energy storage device at the moment 0-1 variable, , Respectively in t The power purchased and sold by the integrated energy system and the upper power grid at all times. , Respectively in t The electrical power and thermal power output of the cogeneration unit at any given moment, For t The electrical power consumed by the carbon capture unit to capture CO2 at any given moment, For t The electric power consumed by the hydrogen production unit at the moment, , Respectively in t The charging and discharging power of the energy storage device at all times, It is the thermal power output of gas boiler.

35. The device according to claim 34, characterized in that The three-stage robust optimization scheduling model is decomposed and reconstructed to obtain a single-layer mixed integer model, which specifically includes: using the inner and outer nested CCG algorithm to decompose and reconstruct the three-stage robust optimization scheduling model into the main problem MP and the sub-problem SP, and the expression is as follows: （45）； （46）； After decomposition by the CCG algorithm, the multi-layer nested structure of the three-stage robust optimization scheduling model is transformed into a single-layer MILP structure using the KKT condition, thereby obtaining a single-layer mixed integer model. The specific transformation process is as follows: MP is expressed as (47); SP is expressed as (48); The main problem goal of the decomposed sub-problem is to solve the transaction power between the integrated energy system and the upper power grid under the worst transaction electricity price. The sub-problem goal of the sub-problem is to solve the unit dispatch plan under the worst integrated energy system source-load scenario. The expression is as follows: (49)； (50)； In the formula, is the worst solution to the uncertainty element obtained from the subproblem, and the auxiliary variable , auxiliary variables .

36. The device according to claim 35, characterized in that Then the GUROBI solver is used to solve the problem, and the scheduling solution includes: 1) Initialize the outer CCG algorithm problem, and initialize the upper and lower bounds of the outer layer as , , initialize the number of iterations , determine the convergence termination condition threshold ; 2) Solve the main problem of the outer CCG algorithm and obtain the optimal solution , update the lower bound of the outer problem , check whether the convergence condition is met ,in , if the convergence condition is met, the solution is terminated; otherwise, the solution continues; 3) Solve the inner CCG algorithm problem and initialize the inner CCG algorithm problem. The upper and lower bounds are initialized as , , initialize the number of iterations , determine the convergence termination condition threshold ; 4) Solve the main problem of the inner CCG algorithm and obtain the optimal solution , update the lower bound of the inner problem ; 5) Solve the inner CCG algorithm sub-problem and obtain the optimal solution , update the upper bound of the inner problem , check whether the convergence condition is met , if the convergence condition is met, terminate the inner problem and jump to 3); otherwise 6) continue to solve; 6) Add the variables and constraints of the second and third stages to the outer CCG algorithm main problem, and initialize the number of iterations Add one and return to 2).

37. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method according to any one of claims 1 to 18 when executing the computer program.

38. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 18 are implemented.

Images