Multi-energy micro-grid distribution robust low-carbon economic dispatching method based on deep learning, electronic equipment and medium

Abstract

The invention belongs to the technical field of multi-energy micro-grid system optimization scheduling, and particularly relates to a multi-energy micro-grid distribution robust low-carbon economic scheduling method based on deep learning, electronic equipment and a medium. According to the method, a mathematical model of a multi-energy micro-grid system is established according to coupling characteristics of various energy sources among power systems. In order to improve the economical efficiency and the low-carbon property of the system, a load demand response mechanism and a carbon transaction mechanism are adopted, and an electric heating load demand response model and a reward and punishment type stepped carbon transaction model are constructed. In order to solve the wind and light uncertainty of the integrated energy system and improve the robustness of the system, a scene set of uncertain variables is generated by using a conditional generative adversarial network in deep learning, and the generated scenes are clustered by using a K-means clustering method to obtain typical scenes. In order to obtain more real probability distribution, a fluctuation range of a typical scene is constrained by using a comprehensive norm, and a probability distribution fuzzy set of uncertain variables is obtained. And based on the constructed fuzzy set, the demand response model and the reward and punishment type stepped carbon transaction model, a two-stage distribution robust low-carbon economic optimization model of the multi-energy microgrid is established, in the first stage, an energy storage equipment start-stop plan of the system is determined, and in the second stage, an initial plan is adjusted and supplemented after uncertainties are revealed. And finally, carrying out iterative solution on the established model by utilizing a column and constraint generation method to obtain an optimal scheduling scheme, thereby ensuring the low-carbon property, the economical efficiency and the robustness of the system.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of multi-energy micro-grid system optimization scheduling, and particularly relates to a multi-energy micro-grid distribution robust low-carbon economic scheduling method based on deep learning, an electronic device and a medium. BACKGROUND

[0002] Under the dual pressures of increasingly serious energy shortage and increasingly prominent environmental protection issues, renewable energy (RG) such as wind energy and solar energy has gradually become a key force for global energy transformation due to its clean and pollution-free advantages, and is favored by all circles. Multi-energy micro-grid (MEMG) as a new technology system in the energy field provides a new solution for promoting the efficient use of renewable energy and plays an indispensable role in energy transformation. However, the randomness and volatility of renewable energy, like a double-edged sword, bring both cleanliness and hope to the energy structure, but also bring great challenges to the stable operation and economic performance of the micro-grid. The use of traditional energy will also cause a large amount of carbon emissions, further exacerbating environmental problems. Therefore, ensuring that MEMG can cope with the uncertainty of renewable energy while achieving economic operation and reducing carbon emissions is the key to its sustainable development.

[0003] At present, stochastic programming and robust optimization are two commonly used methods for dealing with the uncertainty of power systems. Stochastic programming makes optimization decisions based on the probability distribution of uncertain variables, but accurate probability distribution is often difficult to obtain in practical applications. Robust optimization ensures the optimization effect of the system in the worst case by constructing an uncertainty set, but this method is often too conservative, leading to optimization results deviating from actual needs and poor economic performance. Distributionally robust optimization (DRO) combines the advantages of stochastic programming and robust optimization, and by constructing a probability distribution fuzzy set, it improves the flexibility and practical application effect of the optimization result while ensuring robustness. In the application of power system optimization scheduling, the DRO method still faces many challenges. Specifically, the fuzzy set construction of DRO is usually based on the distance between the uncertainty variables or the probability distribution, but the current fuzzy set based on moment information is difficult to fully exploit the information contained in the historical data, and the DRO based on distance is usually complex in structure and cumbersome in solving process. Therefore, when constructing the fuzzy set of DRO, a more efficient and reasonable scheme needs to be sought to ensure the comprehensiveness of data information while reducing the complexity of model solving.

[0004] The load demand response mechanism can flexibly adjust users' electricity consumption behavior according to the grid's operating status, achieving peak shaving and valley filling, reducing peak load, and increasing off-peak load, thereby reducing the peak-shaving pressure on microgrids and improving the stability and economy of grid operation. Simultaneously, it can enhance the microgrid's ability to absorb renewable energy, reduce wind and solar curtailment, and promote efficient energy utilization. The reward-and-penalty tiered carbon trading mechanism effectively constrains and incentivizes microgrid carbon emissions by setting carbon emission quotas and tiered carbon trading prices. Microgrids with carbon emissions below their quotas can obtain carbon trading revenue, increasing their economic income; while microgrids exceeding their carbon emission limits need to pay additional carbon trading costs, thus prompting microgrid operators to actively take emission reduction measures to lower carbon emissions. This mechanism helps promote the development of microgrids towards a low-carbon and clean direction, achieving energy conservation and emission reduction goals, while also providing new optimization means for the economic operation of microgrids, enabling them to minimize costs and maximize benefits while meeting environmental constraints. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention proposes a deep learning-based multi-energy microgrid distributed bloater low-carbon economic dispatch method, electronic equipment, and medium, comprising the following steps:

[0006] Step 1: Establish a simplified mathematical model of the main equipment in the multi-energy microgrid (MEMG) and consider the operating constraints of the equipment;

[0007] Step 2: Construct an electric heating load demand response model to adjust user electricity consumption behavior to smooth peak and valley loads, reduce the peak-shaving pressure on microgrids, and improve the stability and economy of grid operation; establish a reward and punishment tiered carbon trading model to constrain microgrid carbon emissions, promote its low-carbon and clean development, and achieve energy conservation and emission reduction.

[0008] Step 3: Analyze the uncertainty of the scenery in MEMG, use conditional generative adversarial networks in deep learning to obtain the scenarios with uncertain variables, and use K-means clustering to cluster the generated scenarios to obtain typical scenarios with uncertain scenery output; in order to obtain a more realistic probability distribution, use a comprehensive norm composed of 1-norm and ∞-norm to constrain the probability distribution of typical scenarios, thereby obtaining a fuzzy set of uncertain probability distribution.

[0009] Step 4: Based on the constructed fuzzy set, load demand response model and reward-penalty tiered carbon trading model, establish a two-stage sub-Bruker optimization scheduling model for MEMG. The first stage determines the start-up and shutdown plan of the system's energy storage equipment, and the second stage aims to find the unit output plan under the worst probability distribution scenario.

[0010] Step 5: Use the column and constraint generation algorithm to iteratively solve the established model to obtain the optimal scheduling scheme, thereby ensuring the system's low carbon footprint, economy, and robustness.

[0011] Furthermore, the MEMG coupling equipment includes wind and solar generator sets, electric boilers, combined heat and power (CHP) equipment, batteries, and thermal storage tanks. The demand side includes electrical and thermal loads, with key equipment as follows:

[0012] A combined heat and power (CHP) unit is a highly efficient energy utilization device that simultaneously converts the chemical energy of fuel into electrical and thermal energy. However, CHP units must consider constraints on electrothermal output power, electrothermal coupling, and ramp-up constraints.

[0013] Energy storage devices include batteries and thermal storage tanks. To absorb excess wind and solar power generation and reduce curtailment rates, energy storage devices are used to store surplus electricity. When energy output is insufficient, the storage devices release energy to meet load demands. However, energy storage devices must consider capacity constraints and power constraints.

[0014] Furthermore, by adopting an electricity and heat load demand response mechanism, users' electricity consumption behavior can be flexibly adjusted to achieve peak shaving and valley filling, reducing peak load and increasing off-peak load, thereby reducing the peak-shaving pressure on the microgrid and effectively improving the stability and economy of grid operation. Introducing a reward-and-penalty tiered carbon trading mechanism, by dividing carbon emissions into different ranges and setting reasonable tiered pricing strategies, can better reduce the system's carbon emissions. This mechanism helps promote the development of microgrids towards a low-carbon and clean direction, achieving energy conservation and emission reduction goals.

[0015] Furthermore, the MEMG source side considers the uncertainty of wind and solar power output. Based on historical data of uncertain variables on the source side, clustering is performed after generating scenarios of uncertain variables using a conditional generative adversarial network. The specific process is as follows:

[0016] Model training: The Conditional Generative Adversarial Network (CGN) consists of two structurally independent deep learning networks: a generator and a discriminator. The generator generates samples, while the discriminator determines whether the generated samples are genuine or fake. The authenticity of the generator's samples is improved through a minimax game between the generator and discriminator. During training, the generator's input is noise, and its output is fake samples. The discriminator's input is historical data with uncertain variables or fake samples from the generator, and its output is the probability that a sample is genuine. Training is complete when the discriminator cannot distinguish between genuine and fake samples. Scene generation: Fixed conditional values ​​and high-dimensional noise following a normal distribution are input into the trained CGN, and the generator outputs a set of scenes.

[0017] Scene clustering utilizes the K-means algorithm to cluster a large number of scenes generated by a conditional generative adversarial network (GAN), selecting scenes with prominent features and high probabilities from the original scenes. The K-means algorithm determines the number of cluster centers N. s It is pre-defined that during the clustering process, N original scenes will be divided into N... s Clusters, where the cluster center of each cluster is the typical scenario s, and the corresponding probability distribution is P. s .

[0018] Furthermore, considering that the scene probability distribution obtained by scene clustering has a certain error, in order to make the probability distribution closer to the true probability distribution, the comprehensive norm is used to evaluate the probability distribution P of typical scenes. s0 The fluctuation range is limited, as follows:

[0019]

[0020] Among them, Ω p Let θ1 and θ be the set of probability distributions for the scene. ∞ This represents the allowable deviation value for the probability distribution.

[0021] Furthermore, a two-stage sub-bluer low-carbon economic optimization model for MEMG is established. The first-stage model formulates an initial decision plan without considering uncertainties, with the objective function being to minimize the start-up and shutdown costs of MEMG energy storage devices. The second-stage model adjusts the decisions of the first stage to cope with actual uncertainties, with the objective function being to minimize energy interaction costs, energy storage system maintenance costs, wind and solar curtailment losses, operating costs of cogeneration equipment and electric boiler equipment, load demand response costs, and carbon trading costs.

[0022]

[0023] Among them, C a This indicates the start-up and shutdown costs of energy storage equipment (batteries, thermal storage tanks); Indicates the start / stop status of the energy storage device during time period t; C EX C ES C loss C OM C DR C c These represent energy interaction costs, energy storage system maintenance costs, wind and solar curtailment losses, operating costs of cogeneration equipment and electric boiler equipment, load demand response costs, and carbon trading costs, respectively.

[0024] Furthermore, a column and constraint generation algorithm is used to divide the problem into two parts: the main problem and the subproblems, which are solved alternately. The content of the column and constraint generation algorithm is as follows:

[0025] 1) Initialization: Set the number of iterations k = 1, the convergence accuracy α, the upper bound UB = +∞, and the lower bound LB = -∞;

[0026] 2) By solving the main problem, the optimal decision variables and optimal solution for the first stage are obtained, and the load balancer (LB) is updated;

[0027] 3) Based on the partial decision values ​​obtained from the main problem, solve the sub-problems to obtain the probability distribution of the worst-case wind and solar power output and the objective function value of the main problem, and update UB;

[0028] 4) Convergence judgment: If UB-LB≤α is satisfied, then stop the iteration and obtain the optimal solution; otherwise, add the worst probability distribution constraint to the first stage main problem and return to step 2) to continue iterating and solving, k=k+1.

[0029] According to another aspect of the present invention, an electronic device is provided, including a processor and a memory, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the method as described above.

[0030] According to another aspect of the present invention, a computer-readable storage medium is provided, which stores a computer program / instructions that, when executed by a processor, implement the method described above.

[0031] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0032] The method of this invention is illustrated in the following figures:

[0033] Figure 1 This is a flowchart of the deep learning-based multi-energy microgrid distributed blue rod low-carbon economic dispatch method, electronic equipment, and media of the present invention.

[0034] Figure 2 This is a diagram of the multi-energy microgrid structure established in this invention;

[0035] Figure 3 The structural diagram of the conditional generation adversarial network for this invention;

[0036] Figure 4 This is a diagram showing the electrical load demand response results of the present invention;

[0037] Figure 5 This is a graph showing the heat load demand response results of the present invention;

[0038] Figure 6 This is a diagram showing the power dispatching results of the present invention;

[0039] Figure 7 This is a diagram showing the thermal energy scheduling results of the present invention;

[0040] Figure 8 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown. Detailed Implementation

[0041] The following is in conjunction with the appendix Figures 1-7 The technical solutions in this invention will be further described in detail below.

[0042] Appendix Figure 1 This is a flowchart of a deep learning-based multi-energy microgrid distributed bloater low-carbon economic dispatch method, electronic equipment, and media. The method includes the following steps:

[0043] Step 1: Establish a simplified mathematical model of the main equipment in the multi-energy microgrid (MEMG) and consider the operating constraints of the equipment;

[0044] Step 2: Construct an electric heating load demand response model to adjust user electricity consumption behavior to smooth peak and valley loads, reduce the peak-shaving pressure on microgrids, and improve the stability and economy of grid operation; establish a reward and punishment tiered carbon trading model to constrain and incentivize microgrid carbon emissions, promote its low-carbon and clean development, and achieve energy conservation and emission reduction.

[0045] Step 3: Analyze the uncertainty of the scenery in MEMG, use conditional generative adversarial networks in deep learning to obtain the scenarios with uncertain variables, and use K-means clustering to cluster the generated scenarios to obtain typical scenarios with uncertain scenery output; in order to obtain a more realistic probability distribution, use a comprehensive norm composed of 1-norm and ∞-norm to constrain the probability distribution of typical scenarios, thereby obtaining a fuzzy set of uncertain probability distribution.

[0046] Step 4: Based on the constructed fuzzy set, load demand response model and reward-penalty tiered carbon trading model, establish a two-stage sub-Bruker optimization scheduling model for MEMG. The first stage determines the system's equipment start-up and shutdown plan, and the second stage aims to find the unit output plan under the worst probability distribution scenario.

[0047] Step 5: Use the column and constraint generation algorithm to iteratively solve the established model to obtain the optimal scheduling scheme, thereby ensuring the system's low carbon footprint, economy, and robustness.

[0048] As described above, this invention introduces a load demand response mechanism and a reward-and-penalty tiered carbon trading mechanism to improve the system's economic efficiency and low-carbon performance. Secondly, a distributed robust low-carbon economic optimization model is used to describe the wind and solar uncertainties of multi-energy microgrids, and a column and constraint generation method is employed to iteratively solve the model, thereby improving the economic efficiency and robustness of the integrated energy system while ensuring its low-carbon performance.

[0049] Appendix Figure 2 The simplified mathematical model and related constraints of the main equipment in the multi-energy microgrid system structure are as follows:

[0050] 1) Combined heat and power equipment

[0051] To improve energy efficiency, this invention employs a combined heat and power (CHP) system that mixes and burns hydrogen and natural gas to simultaneously produce electricity and heat. The constraints considered are as follows:

[0052]

[0053] Where P CHP (t) represents the output power of CHP during time period t, derived from the output electrical power P. CHP,e (t) and output thermal power P CHP,h (t) composition; P CHP,g (t) represents the natural gas consumption power of CHP; η CHP Energy conversion efficiency; c m Thermoelectric ratio; P CHP,min and P CHP,max These represent the upper and lower limits of CHP output power, respectively; ΔP CHP,min and ΔP CHP,max These represent the upper and lower limits of the CHP slope constraint, respectively.

[0054] 2) Electric boiler

[0055] An electric boiler is a heating device that converts electrical energy into heat energy. Its operation must meet the following constraints:

[0056]

[0057] Among them, P EB,e (t), P EB,t (t), P EB,t Represent the power consumption, heat output, and rated heat output of the electric boiler during time period t, respectively; η EB This indicates the power consumption efficiency of the electric boiler.

[0058] 3) Energy storage system

[0059] Energy storage devices, including batteries and thermal storage tanks, must meet the following constraints during operation:

[0060]

[0061] Here, ES stands for battery or thermal storage tank. These represent the charging / discharging power and the maximum charging / discharging power limit of the energy storage device during time period t, respectively. and These are integer variables of 0-1, representing the charging and discharging states of the energy storage device, respectively; C(t), C... ES,min C ES,max These represent the capacity of the energy storage device at time t, the lower limit of the energy storage device's capacity, and the upper limit of the energy storage device's capacity, respectively. and These represent the charging and discharging efficiency of the energy storage device.

[0062] 5) Power balance constraints

[0063] In a power grid, it is necessary to ensure a balance of supplied power, which can be expressed as follows:

[0064]

[0065] in, P PV (t), P WT (t), P L (t), These represent the electrical power purchased from the grid, the electrical power sold to the grid, the photovoltaic power generation, the wind turbine power generation, the battery discharge power, the battery charging power, the user demand load, the transferable load, and the load that can be reduced during time period t, respectively.

[0066] Because the thermal system is connected to the electrical system, the thermal system also needs to maintain power balance at all times. The power balance expression for thermal energy is as follows:

[0067]

[0068] in, H L (t), These represent the heat release power of the thermal storage tank, the heat charging power of the thermal storage tank, the user's demand for heat load, the transferable heat load, and the reduceable heat load, respectively, during time period t.

[0069] Step 2: In order to effectively improve the system's economy and low carbon emissions, the load demand response mechanism and the reward-and-penalty tiered carbon trading mechanism are used together to achieve the system's economical and low carbon operation.

[0070] Load demand response mechanisms can flexibly adjust users' electricity consumption behavior based on the grid's operating status, achieving peak shaving and valley filling, reducing peak loads, and increasing off-peak loads, thereby reducing the peak-shaving pressure on microgrids and improving the stability and economy of grid operation. Electric and heat loads within multi-energy microgrids are not only interconnected through conversion equipment but also participate in demand response at their respective time dimensions. Based on flexible load control methods, they can be divided into base loads and flexible loads. Flexible loads include loads that can be reduced and loads that can be transferred.

[0071] 1) Transferable load

[0072] Transferable loads must ensure that the total transfer amount is zero, as shown in the following model:

[0073]

[0074] in, These represent the minimum value of the transferable electrical load, the maximum value of the transferable electrical load, and the transferable electrical load, respectively. Minimum transferable heat load, maximum transferable heat load, transferable heat load.

[0075] 2) Load can be reduced

[0076]

[0077] in, These represent the electrical load that can be reduced and the maximum electrical load that can be reduced, respectively. These represent the heat load that can be reduced and the maximum heat load that can be reduced, respectively.

[0078]

[0079] Among them, C DR Indicates the load demand response cost; ω tr and ω re These represent the transferable load compensation factor and the load reduction compensation factor, respectively.

[0080] Compared to traditional carbon trading mechanisms, reward-based tiered carbon trading mechanisms, by dividing carbon emissions into different tiers and setting reasonable tiered pricing strategies, can better reduce the system's carbon emissions. Compared to traditional tiered carbon trading mechanisms, reward-based tiered carbon trading mechanisms, by introducing reward coefficients, can further enhance the system's incentive to participate in carbon trading, thereby improving the system's low-carbon profile. The reward-based tiered carbon trading model is shown below:

[0081]

[0082] M CET =M buy +M CHP (12)

[0083]

[0084] Among them, M CHP Indicates the initial carbon allowance obtained by CHP; M buy This represents the initial carbon allowance obtained by a multi-energy microgrid from purchasing electricity from the external grid; P CHP,e (t) represents the amount of electricity generated by CHP during time period t; P CHP,h (t) represents the heat generated by CHP during time period t; λ represents the thermoelectric conversion factor. This represents the electricity purchased by the multi-energy microgrid during time period t; This represents the decentralization quota per unit of thermal energy consumption; Indicates the power purchase quota; M CET Indicates the initial carbon allowance for a multi-energy microgrid; D CET E represents the actual carbon emissions of a multi-energy microgrid; CET This represents the tradable carbon emissions of a multi-energy microgrid; The carbon transfer coefficient of CHP is represented by ∈; the carbon transfer coefficient of electricity purchased from external grids is represented by d; the carbon emission range is represented by p; the carbon base price is represented by u; the penalty coefficient is represented by ω; and C represents the reward coefficient. C This indicates the cost of carbon trading.

[0085] Step 3 involves generating scenarios with uncertain variables based on conditional generative adversarial networks (CGAN), and then using the K-means algorithm to reduce the generated scenarios. The specific implementation process is as follows.

[0086] 1) Scene generation

[0087] CGAN structure as Figure 3 As shown, it consists of two structurally independent deep learning networks: a generator and a discriminator. On the input side of the CGAN generator, random noise Z is combined with conditional Y and used as the input to the generator G, which outputs a sample x' = G(Z|Y). The CGAN discriminator D not only needs to determine the similarity between the generated sample distribution p(x') and the true sample distribution p(x), but also needs to determine whether the generated sample x' satisfies condition c. The loss functions of the generator and discriminator in CGAN are shown below.

[0088] L G =-E x' [D(G(Z|Y))] (15)

[0089] L D =-E x [D(X|Y)+E x'[D(G(Z|Y))] (16)

[0090] Where E represents the expected value of the corresponding distribution; D(~) represents the discriminator function. The generator aims to increase the discriminator output value for generated samples x', while the discriminator aims to decrease the discriminator output value for generated samples x' and increase the discriminator output value for real samples x. Therefore, the training objective of CGAN is a conditional minimax game.

[0091]

[0092] The distance between the real sample distribution and the generated sample distribution is measured using the Wasserstein distance metric, which effectively addresses issues such as training difficulties and mode collapse. The Wasserstein distance in CGAN can be represented as:

[0093]

[0094] Equation (14) represents the optimization of the discriminator D's parameter θ (D) This enables the discriminator to maximize the expected value E of the true sample. x [D(X)] and the expected value E of the generated sample x' The difference between [D(G(Z))].

[0095] 2) Scene clustering

[0096] The trained CGAN is used to generate N scenic power generation scenes. As N increases, the complexity of the model solution increases continuously. Therefore, it is necessary to reduce the number of generated scenes to decrease the computational load. The K-means method is chosen to cluster N scenes. s A typical scenario, the main steps of which are as follows:

[0097] 1) Let the number of clusters be N s Randomly select N s One scene is used as the initial cluster center;

[0098] 2) Calculate the distance between the remaining scenes and the center point, assign each scene to the nearest cluster center, and then divide the original scenes into m clusters;

[0099] 3) For each cluster, a new cluster center is selected, and the sum of the distances from all points to the cluster center is calculated. If the sum of the distances decreases after the replacement, the new cluster center is selected.

[0100] 4) Repeat steps 2 and 3 until the cluster centers no longer change;

[0101] 5) The algorithm converges, and the final cluster centers and clusters are output.

[0102] The comprehensive norm mentioned in step 3 consists of the 1-norm and the ∞-norm, and is used to construct the sub-Bruker optimized fuzzy set. Its expression is as follows:

[0103]

[0104] Where Pr{} represents the probability scale; N and N s θ1 and θ2 represent the number of scenes generated by CGAN and the number of scenes after clustering, respectively. ∞ Let the permissible deviations of the probability distributions be represented as follows:

[0105]

[0106] Where α1 and α ∞ This represents the confidence level that the probability distribution needs to satisfy.

[0107] From formulas (19) and (20), the confidence set of the probability distribution can be derived as follows:

[0108]

[0109] In the specific implementation of step 4, a two-stage sub-Bruker low-carbon economic optimization model is established based on the mathematical model of the main equipment, the load demand response mechanism, the reward-and-penalty tiered carbon trading model, and the probability distribution fuzzy set. The objective of the first stage is to formulate an initial decision plan without considering uncertainties. The objective function of the first stage is expressed as follows:

[0110]

[0111] Where x represents the decision variable for the first stage, C a This indicates the start-up and shutdown costs of energy storage devices; This indicates the start-up and shutdown status of the energy storage device during time period t.

[0112] The goal of the second stage is to find the probability distribution of the worst-case discrete source-load scenario and adjust the decisions made in the first stage. The objective function of the second stage is expressed as follows:

[0113]

[0114] Where y represents the decision variable for the second stage; C EX C ES C loss C OM C DR C c These represent energy interaction costs, energy storage system maintenance costs, wind and solar curtailment losses, operating costs of cogeneration equipment and electric boiler equipment, load demand response costs, and carbon trading costs, respectively.

[0115] In the specific implementation of step 5, the established sub-Bruker optimization problem can be summarized into the following general formula:

[0116]

[0117] Where a is the cost coefficient corresponding to the first stage; b is the cost coefficient in the objective function of the second stage; p s Let Ax ≤ e represent the probability distribution of the actual scenario; let Ax ≤ e represent the constraints related to the decision variable x in the first stage; let By ≤ f represent the constraints related to the decision variable y in the second stage; and let Ex + Fy ≤ g represent the constraints related to both the decision variable in the first stage and the decision variable in the second stage.

[0118] Factor (24) represents a multi-level optimization problem that cannot be solved directly by commercial solvers. Therefore, the model is divided into a main problem and sub-problems, and the column and constraint generation method is used to solve iteratively.

[0119] By solving the main problem, we can obtain the optimal solution of the decision variables in the first stage and the upper bound of the model (24). The main problem can be expressed as:

[0120]

[0121] Where η is the introduced auxiliary variable; k is the number of iterations for model solving; and K is the maximum number of iterations.

[0122] When the solution to the main problem is known, the subproblem is solved by finding the probability distribution of the worst-case scenario within the confidence interval, thus obtaining the lower bound of model equation (24). The subproblem can be expressed as:

[0123]

[0124] The specific solution process for the constraint generation method is as follows:

[0125] 1) Initialization: Set the number of iterations k = 1, the convergence accuracy α, the upper bound UB = +∞, and the lower bound LB = -∞;

[0126] 2) By solving the main problem, the optimal decision variables and optimal solution for the first stage are obtained, and the load balancer (LB) is updated;

[0127] 3) Based on the partial decision values ​​obtained from the main problem, solve the sub-problems to obtain the probability distribution of the worst-case wind and solar power output and the objective function value of the main problem, and update UB;

[0128] 4) Convergence judgment: If UB-LB≤α is satisfied, then stop the iteration and obtain the optimal solution; otherwise, add the worst probability distribution constraint to the first stage main problem and return to step 2) to continue iterating and solving, k=k+1.

[0129] By solving the established two-stage bibliometric low-carbon economic optimization model, the optimal scheduling plan for each MEMG unit can be obtained. Figure 4 and Figure 5 The final demand value after the demand response of the electric heating load can be obtained respectively. (See attached diagram) Figure 6 The optimal scheduling result of electricity is determined by... Figure 6 It is evident that renewable energy provides the majority of electricity. Combined heat and power (CHP) equipment can provide flexible power supply to the system. Batteries store energy during periods of low demand and release it during peak demand periods, improving system flexibility. Electric boilers can play a role in transferring and storing electrical power, absorbing some of the output from renewable energy sources. (Appendix) Figure 7 To optimize the scheduling of thermal energy, combined heat and power (CHP) equipment and electric boilers work together to meet the heat load, while thermal storage tanks, as a buffer device, also play an important role in balancing heat supply and demand.

[0130] According to another aspect of the present invention, an electronic device is also provided. Figure 8 A schematic block diagram of an electronic device according to an embodiment of the present invention is shown. Figure 8 As shown, the electronic device 700 includes a processor 710 and a memory 720. The memory 720 stores a computer program, which the processor 710 executes to implement the method described above.

[0131] According to another aspect of the present invention, a computer-readable storage medium is also provided. The storage medium stores a computer program / instructions that, when executed by a processor, implement the method described above. The storage medium may, for example, include a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disc read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0132] Those skilled in the art will readily understand the implementation structure, working principle, and beneficial effects of electronic devices and computer-readable storage media by reading the above methods. For the sake of brevity, further details will not be elaborated here.

[0133] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of the invention. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the invention as claimed in the appended claims.

[0134] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0135] In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0136] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0137] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, this approach should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with fewer features than all of those in a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0138] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or apparatus so disclosed can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0139] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0140] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules in the electronic device according to embodiments of the present invention. The present invention can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing some or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0141] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0142] The above description is merely a specific embodiment of the present invention or an explanation of that embodiment. The scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A deep learning-based multi-energy microgrid distributed low-carbon economic dispatch method, electronic equipment, and medium, characterized in that, Includes the following steps: Step 1: Establish a simplified mathematical model of the main equipment in the multi-energy microgrid (MEMG) and consider the operating constraints of the equipment; Step 2: Construct an electric heating load demand response model to adjust user electricity consumption behavior to smooth peak and valley loads, reduce the peak-shaving pressure on microgrids, and improve the stability and economy of grid operation; establish a reward and punishment tiered carbon trading model to constrain and incentivize microgrid carbon emissions, promote its low-carbon and clean development, and achieve energy conservation and emission reduction. Step 3: Analyze the uncertainty of the wind and light in MEMG, use conditional generative adversarial network to obtain the scenarios with uncertain variables, and use K-means clustering to cluster the generated scenarios to obtain typical scenarios with uncertain wind and light output; in order to obtain a more realistic probability distribution, use a comprehensive norm composed of 1-norm and ∞-norm to constrain the probability distribution of typical scenarios, thereby obtaining a fuzzy set of uncertain probability distribution. Step 4: Based on the constructed fuzzy set, load demand response model and reward-penalty tiered carbon trading model, establish a two-stage sub-bluer low-carbon economic optimization scheduling model for MEMG. The first stage determines the system's equipment start-up and shutdown plan, and the second stage aims to find the unit output plan under the worst probability distribution scenario. Step 5: Use the column and constraint generation algorithm to iteratively solve the established model to obtain the optimal scheduling scheme, thereby ensuring the system's low carbon footprint, economy, and robustness.

2. The method according to claim 1, characterized in that, The MEMG coupling equipment includes wind and solar generator sets, electric boilers, combined heat and power units, batteries, and thermal storage tanks. The demand side includes electrical and thermal loads, with key equipment as follows: 1) Combined heat and power equipment To improve energy utilization, this invention employs a combined heat and power (CHP) system that mixes and combusts hydrogen and natural gas to simultaneously produce electricity and heat. Simultaneously, the CHP system needs to consider constraints on electrothermal output power, electrothermal coupling, and ramp-up constraints. 2) Energy storage system Energy storage devices include batteries and thermal storage tanks. To absorb excess wind and solar power generation and reduce curtailment rates, energy storage devices are used to store surplus energy. When energy output is insufficient, the storage devices release energy to meet load demands. Energy storage devices must consider capacity constraints and power constraints, among other factors.

3. The method according to claim 1, characterized in that, To effectively improve the system's economy and low-carbon performance, a load demand response mechanism and a reward-based tiered carbon trading mechanism are used in conjunction to achieve economical and low-carbon operation. Load demand response can flexibly adjust user electricity consumption based on grid operating conditions, achieving peak shaving and valley filling, reducing peak loads and increasing off-peak loads, thereby reducing the peak-shaving pressure on microgrids and improving grid stability and economy. Compared to traditional carbon trading mechanisms, the reward-based tiered carbon trading mechanism, by dividing carbon emissions into different ranges and setting reasonable tiered pricing strategies, can better reduce the system's carbon emissions. Compared to traditional tiered carbon trading mechanisms, the reward-based tiered carbon trading mechanism, by introducing reward coefficients, can further enhance the system's participation in carbon trading, thereby improving the system's low-carbon performance.

4. The method according to claim 1, characterized in that, MEMG considers the uncertainty of wind and solar power output at the source side. Based on historical data of source uncertainties, it generates scenarios of uncertainties using a conditional generative adversarial network and then performs clustering. The specific process is as follows: 1) Model training: Conditional Generative Adversarial Networks (GANs) consist of two structurally independent deep learning networks: a generator and a discriminator. The generator generates samples, while the discriminator determines whether the generated samples are genuine or fake. The authenticity of the generator's samples is improved through a minimax game between the generator and discriminator. During training, the generator's input is noise, and its output is fake samples. The discriminator's input is historical data with uncertain variables or fake samples from the generator, and its output is the probability that a sample is genuine. Training is complete when the discriminator can no longer distinguish between genuine and fake samples. 2) Scene generation: Input fixed conditional values ​​and high-dimensional noise that follows a normal distribution into a trained conditional generative adversarial network, and the generator outputs a set of scenes. 3) Scene clustering: The K-means algorithm is used to cluster a large number of scenes generated by a conditional generative adversarial network, that is, to select scenes with obvious features and high probabilities from the original scenes. The number of cluster centers N in the K-means algorithm is... s It is pre-defined that during the clustering process, N original scenes will be divided into N... s Clusters, where the cluster center of each cluster is the typical scenario s, and the corresponding probability distribution is P. s .

5. The method according to claim 1, characterized in that, Considering that the scene probability distribution obtained by scene clustering has a certain error, in order to make the probability distribution closer to the true probability distribution, the comprehensive norm is used to adjust the scene probability distribution P. s0 The fluctuation range is limited, as follows: Among them, Ω p Let θ1 and θ be the set of probability distributions for the scene. ∞ This represents the allowable deviation value for the probability distribution.

6. The method according to claim 1, characterized in that, A two-stage sub-Blu-ray low-carbon economic optimization model for MEMG is established. The first stage model formulates an initial decision plan without considering uncertainties, and its objective function is to minimize the start-up and shutdown costs of MEMG energy storage equipment. The second stage model adjusts the decision of the first stage to cope with actual uncertainties, and its objective function is to minimize energy interaction costs, energy storage system maintenance costs, wind and solar curtailment losses, operating costs of cogeneration equipment and electric boiler equipment, load demand response costs, and carbon trading costs. Among them, C a This indicates the start-up and shutdown costs of energy storage equipment (batteries, thermal storage tanks); Indicates the start / stop status of the energy storage device during time period t; C EX C ES C loss C OM C DR C c These represent energy interaction costs, energy storage system maintenance costs, wind and solar curtailment losses, operating costs of cogeneration equipment and electric boiler equipment, load demand response costs, and carbon trading costs, respectively. The established two-stage partial Blue bar optimization problem can be summarized into the following general formula: Where a is the cost coefficient corresponding to the first stage; b is the cost coefficient in the objective function of the second stage; p s Let Ax ≤ e represent the probability distribution of the actual scenario; let Ax ≤ e represent the constraints related to the decision variable x in the first stage; let By ≤ f represent the constraints related to the decision variable y in the second stage; and let Ex + Fy ≤ g represent the constraints related to both the decision variable in the first stage and the decision variable in the second stage.

7. The method according to claim 1, characterized in that, The optimization problem (3) is divided into a main problem and subproblems and solved alternately using the column and constraint generation algorithm. The main problem can be expressed as: Where η is the introduced auxiliary variable; k is the number of iterations for model solving; and K is the maximum number of iterations. When the solution to the main problem is known, the subproblem is solved by finding the probability distribution of the worst-case scenario within the confidence interval, thus obtaining the lower bound of model equation (3). The subproblem can be expressed as: The iterative process of the column and constraint generation algorithm is as follows: 1) Initialization: Set the number of iterations k = 1, the convergence accuracy α, the upper bound UB = +∞, and the lower bound LB = -∞; 2) By solving the main problem, the optimal decision variables and optimal solution for the first stage are obtained, and the load balancer (LB) is updated; 3) Based on the partial decision values ​​obtained from the main problem, solve the sub-problems to obtain the probability distribution of the worst-case wind and solar power output and the objective function value of the main problem, and update UB; 4) Convergence judgment: If UB-LB≤α is satisfied, then stop the iteration and obtain the optimal solution; otherwise, add the worst probability distribution constraint to the first stage main problem and return to step 2) to continue iterating and solving, k=k+1.

8. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the method as claimed in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The system stores a computer program / instructions that, when executed by a processor, implement the method as described in any one of claims 1-7.

Images