Low-carbon scheduling method and system for multi-park integrated energy system

By generating adversarial networks and improving the Shapley value algorithm, high-precision output forecast sequences are dynamically generated and combined with a carbon trading cost optimization model to solve the problems of cumulative forecast deviations and unfair benefit distribution in multi-park integrated energy systems, thereby achieving the minimization of system total costs and dynamic control of carbon emissions.

CN120725334APending Publication Date: 2025-09-30SHANDONG UNIV
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202510805861.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-17
Publication Date
2025-09-30

AI Technical Summary

Technical Problem

The existing scheduling method of multi-park integrated energy system relies on fixed probability distribution to generate renewable energy output scenarios, which leads to the accumulation of prediction deviations. The traditional Shapley value allocation does not consider the impact of prediction errors on benefit fairness, and the multi-objective separation optimization leads to an imbalance between the total system cost and low-carbon goals.

Method used

A high-precision output forecast sequence is dynamically generated by generative adversarial networks, a comprehensive optimization model integrating carbon trading costs and user satisfaction is adopted, and an improved Shapley value algorithm combined with a prediction deviation penalty coefficient is used to improve the fairness of revenue distribution.

Benefits of technology

It achieves high-precision generation of output sequences under complex meteorological conditions, solves the problem of cumulative prediction deviations, realizes global minimization of total system costs and dynamic control of carbon emissions, and improves the precise allocation of energy sharing benefits.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120725334A_ABST
    Figure CN120725334A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of multi-park integrated energy system scheduling, in particular to a low-carbon scheduling method and system for a multi-park integrated energy system, and the method comprises the steps: obtaining renewable energy historical output data and historical meteorological parameters of each park, and generating a renewable energy generation power prediction sequence through a pre-trained generative adversarial network; constructing a comprehensive energy optimization model, inputting a renewable energy power generation power prediction sequence, and outputting an equipment scheduling instruction and an energy interaction strategy of each park; executing the equipment scheduling instruction, monitoring actual renewable energy power generation power data, and calculating a renewable energy power generation power prediction deviation; and calculating the energy interaction cost and the energy sharing income of each park, and calculating the actual marginal contribution value of each park by adopting an improved Shapley value algorithm based on the renewable energy power generation power prediction deviation, the energy interaction cost and the energy sharing income. According to the invention, prediction precision and scheduling reliability can be improved, cost and carbon emission are optimized, and earnings are distributed fairly.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of multi-park integrated energy system scheduling, and specifically to a low-carbon scheduling method and system for a multi-park integrated energy system. Background Art

[0002] Multi-park Integrated Energy Systems (PIESs) are collaborative systems comprised of multiple park-level integrated energy systems (PIESs). Due to their energy coupling and synergy advantages, they are a key technical direction for improving clean energy utilization and reducing carbon emissions. Renewable energy (RE) is clean and flexible, but also intermittent. Inter-park energy complementarity and sharing mechanisms within PIESs, such as the Emissions Trading System (ETS), can mitigate the impact of renewable energy volatility on the power grid to a certain extent. However, balancing low-carbon scheduling with economic benefits remains a technical challenge.

[0003] Existing technologies achieve multi-park integrated energy system scheduling through a combination of a randomized optimization scenario generation method and a traditional Shapley value-based revenue allocation method. The randomized optimization scenario generation method constructs a historical probability distribution model of renewable energy output using a Markov chain, generates a typical output sequence using Monte Carlo simulation, and employs scenario reduction techniques to screen representative scenarios. This method, combined with mixed-integer linear programming, solves the equipment scheduling solution, mitigating the impact of renewable energy fluctuations on scheduling decisions through probabilistic modeling. The traditional Shapley value-based revenue allocation method traverses all possible park alliance combinations, calculates the difference in total alliance costs before and after a single park joins as its marginal contribution, and uses a permutation-combination weighted average formula to determine the allocation weight for each park. This approach can, to a certain extent, ensure fairness in the distribution of benefits and incentive compatibility in multi-agent cooperative games.

[0004] However, the scheduling methods of multi-park integrated energy systems still have significant shortcomings: random optimization relies on the fixed probability distribution assumption of historical data, and cannot dynamically capture the impact of complex meteorological conditions and equipment status on the fluctuations of renewable energy output, resulting in the accumulation of deviations between the predicted sequence and the actual output; the traditional Shapley value algorithm only distributes benefits based on idealized marginal contributions, and does not consider the fluctuations in energy interaction costs caused by prediction errors, which makes the actual contribution of high-deviation parks overestimated or underestimated; the existing scheduling model separates and optimizes carbon emission costs, user satisfaction and energy interaction costs, lacks a multi-objective coordination mechanism, and it is difficult to achieve the global optimization of the total system cost and low-carbon goals. Summary of the Invention

[0005] In response to the technical problems that the existing scheduling method of multi-park integrated energy system relies on fixed probability distribution to generate renewable energy output scenarios, resulting in accumulated prediction deviations, the traditional Shapley value allocation does not consider the impact of prediction errors on benefit fairness, and the multi-objective separation optimization leads to an imbalance in the trade-off between the total system cost and low-carbon goals, the present application provides a low-carbon scheduling method and system for multi-park integrated energy system, which dynamically generates high-precision output prediction sequences through generative adversarial networks, integrates a comprehensive optimization model of carbon trading costs and user satisfaction to achieve multi-objective collaboration, and adopts an improved Shapley value algorithm combined with a prediction deviation penalty coefficient to improve the fairness of benefit distribution.

[0006] In a first aspect, the present application provides a low-carbon scheduling method for a multi-park integrated energy system, comprising the following steps: S1. Obtain historical renewable energy output data and historical meteorological parameters for each park in a multi-park integrated energy system. Use a pre-trained generative adversarial network to generate a corresponding renewable energy power generation prediction sequence for each park. The generator generates the prediction sequence based on each park's historical power generation data and meteorological parameters, and the discriminator verifies the authenticity of the prediction sequence through adversarial training. S2. Build a comprehensive energy optimization model. Input the forecasted renewable energy generation power series for each park into the model. The optimization goal is to minimize the total system cost while satisfying the constraints. Output the equipment scheduling instructions and energy interaction strategy for each park. The total system cost includes energy interaction costs and carbon trading costs that take into account user satisfaction. S3. Execute the equipment dispatch instructions for each park and monitor the actual renewable energy power generation data to obtain the actual renewable energy power generation sequence for each park. Calculate the predicted renewable energy power generation deviation for each park based on the predicted renewable energy power generation sequence and the actual renewable energy power generation sequence:

[0007] in, is the forecast deviation of renewable energy power generation in the nth park; is the actual power generation sequence of renewable energy in the nth park is the actual power generation sequence of renewable energy in the nth park, ; S4. Calculate the energy interaction costs and energy sharing benefits of each park based on the energy interaction strategy. Use the improved Shapley value algorithm to calculate the actual marginal contribution of each park based on the forecast deviation of renewable energy power generation, energy interaction costs, and energy sharing benefits. The formula is:

[0008]

[0009] in, is the actual marginal contribution value of the nth park; is the confidence error adjustment coefficient; The highest value among all the forecast deviations of renewable energy power generation in the park; is the original marginal contribution value of the nth park without considering the error adjustment, calculated using the traditional Shapley value algorithm; S5. Allocate the benefits of the multi-park integrated energy system based on the actual marginal contribution of each park.

[0010] It should be further explained that in step S1, the loss function of the generator is:

[0011] in, For noisy data Normal distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the generated data, ranging from 0 to 1; The loss function of the discriminator is:

[0012] in, It is real data obey the distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the real data, ranging from 0 to 1; The training process of the generative adversarial network is: The generator and the discriminator perform adversarial training by alternately optimizing the objective function to approach the Nash equilibrium and obtain a pre-trained generative adversarial network. The objective function is:

[0013] in, is the adversarial loss of the generative adversarial network.

[0014] It should be further explained that the generator uses the LSTM network structure to generate the prediction sequence, and its output function for:

[0015] in, is the nonlinear mapping of the LSTM unit; is the weight matrix of the generator; is the bias term; is a noise vector that obeys the normal distribution; The historical output and historical meteorological parameters; The discriminator uses convolutional neural network to extract data features, and its feature extraction function for:

[0016] in, is the feature map of the convolutional neural network; is the convolution kernel weight; is the bias term; is generated data or real data series.

[0017] It should be further explained that in step S2, the objective function of the comprehensive energy optimization model is:

[0018] in, is the total cost of individual operation of the nth park; T is the total number of time periods; is the carbon trading cost of the park in the tth period considering user satisfaction:

[0019] is the carbon trading cost of the park in period t, The cost of adapting the peak load benefits to users in this park; The cost of electricity purchased from the distribution station in the park during the tth period; The electricity sales revenue of the distribution station in the park in the tth period; is the gas purchase cost of the gas distribution station in the park during the tth period; is the gas sales revenue of the gas distribution station in the park during the tth period; is the PT operating cost of the park in the tth period; is the operating cost of the P2G equipment in the park during the tth period.

[0020] It should be further explained that the carbon trading costs of each park are calculated based on the tiered carbon reward and penalty mechanism, including: Calculate the carbon emission rights that each park needs to trade in the carbon trading market :

[0021] in, is the actual carbon emissions, for carbon emission quotas; >0 means the park needs to purchase carbon emission rights in the carbon trading market. <0The park needs to sell carbon emission rights in the carbon trading market; Calculate the carbon trading cost of the tth park ;

[0022] in, is the function for calculating carbon trading costs; A preset threshold for carbon emission rights; is the benchmark carbon price, determined by the carbon trading market; and is the preset penalty coefficient.

[0023] It should be further explained that carbon emission quotas The calculation formula is:

[0024] in, Additional carbon emission allowances for P2G; Initial carbon emission quota without P2G:

[0025] Initial carbon emission allowances for electricity purchased from distribution stations; Initial carbon emission allowances for gas turbines; It is the initial carbon emission quota of the cogeneration unit.

[0026] It should be further explained that The calculation formula is:

[0027] Where, It is the preset user benefit distribution ratio coefficient; It is the additional carbon emission quota brought about by users adjusting the indoor temperature.

[0028] It should be further explained that step S2 also includes calculating user satisfaction , the steps include: Calculate user satisfaction with thermal energy usage :

[0029] Calculate user satisfaction with heating costs :

[0030] Calculating user satisfaction :

[0031] in, The indoor temperature target value set by the user in time period t; The minimum indoor temperature value acceptable to the user during time period t; is the time series of indoor temperature actually detected during period t; The expected value of heating energy cost set by the user in period t; The upper limit of the maximum heat energy cost acceptable to the user in period t; The minimum heat energy cost acceptable to the user in period t; 、 is the preset weight coefficient, satisfying .

[0032] It should be further explained that The calculation method is:

[0033] Where J is the preset number of historical periods, indicating the number of temperature lags considered in modeling; represents the indoor temperature during time period tj; represents the water supply temperature during time period tj; represents the outdoor temperature during time period tj; 、 、 is the weight coefficient corresponding to the historical period, which is determined by fitting historical data.

[0034] It should be further explained that the constraints of step S2 include: Grid node balance constraints:

[0035] Gas network node balance constraints:

[0036] Heat network node balance constraints:

[0037] in, is the total electrical load during time period t; The exchange power of the distribution station in time period t; is the wind turbine power consumption in period t; is the photovoltaic power consumption in period t; is the power consumption of the P2G device in time period t; is the power generated by the gas turbine during period t; is the electric power of the cogeneration unit in period t; Exchange power for the storage device during time period t; is the line active power loss flowing from node i in period t; is the line active power loss injected into node j during period t; is the total gas load during period t; is the gas exchange volume of the gas distribution station in time period t; is the P2G gas production in period t; is the gas consumption of the gas turbine in period t; is the gas exchange volume of the gas storage equipment in time period t; is the gas loss from node i during period t; is the gas loss injected into node j during period t; is the total heat load during period t; is the heating power of the waste heat boiler in time period t; is the thermal power of the cogeneration unit in time period t.

[0038] It should be further explained that the constraints also include satisfying the return water temperature The dynamic equation of:

[0039] in, 、 、 is the weight coefficient of the return water temperature model.

[0040] It should be further explained that in step S4, the following method is used to calculate : Calculate the objective function:

[0041] Where N is the total number of parks in the multi-park integrated energy system, The operating cost of the nth park is:

[0042] The cost generated by the interaction between the nth park and the mth park; The cost generated by the interaction between the nth park and the mth park; The energy sharing income of the nth park is:

[0043] S is any sub-alliance that does not include park n; is the number of members of sub-alliance S; is the original cooperation value of sub-alliance S; is the probability weight of park n when joining sub-alliance S.

[0044] It should be further explained that The generation formula is:

[0045] in, Represents a Gaussian distribution function with a mean of 0 and a variance of 1; is the confidence error of the generator; is the confidence error of the discriminator.

[0046] In a second aspect, the present application provides a low-carbon scheduling system for a multi-park integrated energy system, which is used to implement the low-carbon scheduling method for the multi-park integrated energy system, including: The data acquisition module is used to obtain the historical renewable energy output data and historical meteorological parameters of each park in the multi-park integrated energy system; Renewable energy generation power prediction module, which is used to generate a corresponding renewable energy generation power prediction sequence for each park through a pre-trained generative adversarial network; The integrated energy optimization module is used to build an integrated energy optimization model, input the renewable energy power generation prediction sequence of each park into the integrated energy optimization model, and output the equipment scheduling instructions and energy interaction strategies of each park; The instruction execution and deviation calculation module is used to execute the equipment scheduling instructions of each park and monitor the actual power generation data of renewable energy, obtain the actual power generation sequence of renewable energy in each park, and calculate the predicted deviation of renewable energy power generation in each park based on the predicted power generation sequence of renewable energy and the actual power generation sequence of renewable energy; The marginal contribution evaluation module is used to calculate the energy interaction cost and energy sharing benefits of each park based on the energy interaction strategy. Based on the forecast deviation of renewable energy power generation, energy interaction cost and energy sharing benefits, the improved Shapley value algorithm is used to calculate the actual marginal contribution value of each park; The profit distribution module is used to distribute the profits of the multi-park integrated energy system according to the actual marginal contribution value of each park.

[0047] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein the processor is used to implement the steps of the above-mentioned low-carbon scheduling method for a multi-park integrated energy system when executing the computer program.

[0048] In a fourth aspect, the present application provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the low-carbon scheduling method for the multi-park integrated energy system.

[0049] It can be seen from the above technical solutions that this application has the following advantages: 1. This application dynamically generates renewable energy power generation prediction sequences through a pre-trained generative adversarial network, and uses the adversarial training mechanism of the generator and discriminator to capture the time dependence and fluctuation pattern of the output data. It solves the problem of prediction deviation accumulation caused by random optimization relying on a fixed probability distribution, achieves high-precision generation of output sequences under complex meteorological conditions, and provides dynamic and reliable data input for multi-park scheduling.

[0050] 2. This application constructs a comprehensive energy optimization model that includes energy interaction costs and carbon trading costs that take into account user satisfaction. Through a layered carbon reward and punishment mechanism and a heat network demand response satisfaction quantification function, it solves the synergy imbalance problem caused by multi-objective separation optimization in existing technologies, and achieves global minimization of total system costs and dynamic control of carbon emissions under the constraints of electricity-gas-heat coupling.

[0051] 3. This application adopts an improved Shapley value algorithm, embeds the renewable energy forecast deviation as a weight factor into the marginal contribution calculation, and dynamically corrects the actual contribution value of the park through the confidence error adjustment coefficient. It solves the problem of benefit distribution distortion caused by traditional methods that ignore forecast errors, and realizes the precise allocation of energy sharing benefits and forecast risk responsibilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0053] Figure 1 This is a flow chart of a low-carbon scheduling method for a multi-park integrated energy system in one embodiment of the present application.

[0054] Figure 2 This is an architectural diagram of a multi-park integrated energy system in one embodiment of the present application.

[0055] Figure 3 This is a diagram of a dynamic reward and punishment mechanism based on GAN error classification in one embodiment of the present application; Figure 3 In the figure, (a) is the wind turbine prediction and error graph of GAN; (b) is the wind turbine reward and penalty coefficient diagram under different errors; (c) is the photovoltaic prediction and error map of GAN; (d) is a diagram of photovoltaic reward and penalty coefficients under different errors.

[0056] Figure 4 This is a diagram showing the results of generating a multi-park renewable energy output scenario in one embodiment of the present application; Figure 4 In the middle, (a) is the WT1 image generated by GAN in PIES A; (b) is the WT2 image generated by GAN in PIES A; (c) is the PV1 image generated by GAN in PIES A; (d) is the PV1 image generated by GAN in PIES B; (e) is the WT1 image generated by GAN in PIES C; (f) is the PV1 map generated by GAN in PIES C.

[0057] Figure 5 This is a diagram of power interaction and output power of each park in one embodiment of the present application; Figure 5 In, (a) is the power interaction diagram between the parks; (b) is the output power diagram of each park.

[0058] Figure 6 is a diagram of natural gas and heat output for each park in one embodiment of the present application; Figure 6 In the figure, (a) is the natural gas output diagram of each park; (b) is the heat output diagram of each park.

[0059] Figure 7 This is a comparison chart of system operating indicators in different scenarios in an embodiment of the present application; Figure 7 In the figure, (a) is a comparison chart of electricity purchase costs; (b) is a comparison chart of GT carbon emissions; (c) is a comparison chart of PT carbon emissions; (d) is a comparison chart of indoor temperatures.

[0060] Figure 8 It is a schematic block diagram of a low-carbon scheduling system for a multi-park integrated energy system in one embodiment of the present application.

[0061] Figure 9 It is a schematic diagram of the hardware structure of an electronic device in one embodiment of the present application. DETAILED DESCRIPTION

[0062] In order to make the application objectives, features, and advantages of this application more obvious and easy to understand, the technical solutions protected by this application will be clearly and completely described below using specific embodiments and drawings. Obviously, the embodiments described below are only part of the embodiments of this application, not all of them. Based on the embodiments in this patent, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this patent.

[0063] The following describes in detail the low-carbon scheduling method involved in this application. Specific details such as specific system structures and technologies are provided for illustrative purposes rather than for limitation, so as to provide a thorough understanding of the embodiments of this application. However, it should be clear to those skilled in the art that this application can also be implemented in other embodiments without these specific details.

[0064] In the low-carbon scheduling method involved in this application, the term "comprising" is used to indicate the presence of the described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0065] To facilitate the clear description of the technical solutions of this application, the words "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or order of execution, and the words "first" and "second" do not necessarily mean different.

[0066] The phrases "one embodiment" or "some embodiments" described in this application mean that the specific features, structures, or characteristics described in the embodiment are included in one or more embodiments of the application. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in other embodiments," etc. that appear in different places in this application do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized.

[0067] The following are some explanations of terms in this plan to facilitate a better understanding of this plan: Emissions Trading (ETS): ETS refers to the economic activity of buying and selling carbon emission allowances through a trading market, based on initial government-allocated carbon emission quotas allocated by a multi-park integrated energy system. When actual system emissions fall below their quotas, the system can sell the surplus for a profit; when they exceed the quota, the system must purchase allowances. Trading utilizes a tiered reward and penalty mechanism, with larger allowance gaps resulting in higher unit price penalties and reward coefficients. Allowance calculations include the carbon absorption contribution of P2G equipment, and trading revenue is shared with users participating in demand response.

[0068] Generative Adversarial Networks (GANs) are deep learning models consisting of a generator (G) and a discriminator (D). These two are trained through an adversarial game, aiming to enable the generator to learn the distribution of real data and generate realistic samples. The process is as follows: First, the generator receives random noise (such as a normally distributed random vector) and generates fake samples through neural network mapping. Next, the discriminator receives both the fake samples and real data samples output by the generator and determines the authenticity of the samples (outputting a probability value between 0 and 1, with 1 indicating complete authenticity). The generator then optimizes its parameters through backpropagation, aiming to "deceive" the discriminator, so that the generated fake samples more closely resemble the distribution of real data. The discriminator simultaneously optimizes its parameters with the goal of correctly distinguishing real from fake samples. This cycle repeats until the discriminator can no longer accurately distinguish real from fake samples. At this point, the samples generated by the generator approximate the distribution of real data, reaching a relatively balanced "Nash equilibrium."

[0069] Traditional Shapley Value Algorithm: The traditional Shapley Value Algorithm (Shapley Value Algorithm) originates from cooperative game theory and is used to solve the problem of fair distribution of benefits or costs in multi-person cooperative games. Its core idea is to determine each participant's distribution value based on their marginal contribution to the cooperative alliance, ensuring that the distribution result satisfies axioms such as efficiency, symmetry, dummy variables, and additivity.

[0070] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0071] The low-carbon scheduling method provided in the embodiment of the present application is executed by a computer device, and accordingly, the low-carbon scheduling system of the multi-park integrated energy system runs in the computer device.

[0072] Figure 1 This is a flow chart of a low-carbon scheduling method according to an embodiment of the present application. Figure 1 The execution entity can be a low-carbon scheduling system for a multi-park integrated energy system. According to different needs, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0073] like Figure 1 As shown in Figure 2, the low-carbon dispatch method for the multi-park integrated energy system includes: Step S1: Obtain the historical renewable energy output data and historical meteorological parameters of each park in the multi-park integrated energy system, and generate a corresponding renewable energy power generation prediction sequence for each park through a pre-trained generative adversarial network. The generator generates the prediction sequence based on the historical power generation data and meteorological parameters of each park, and the discriminator verifies the authenticity of the prediction sequence through adversarial training.

[0074] By pre-training a generative adversarial network to generate a forecast sequence of renewable energy power generation, a renewable energy uncertainty processing framework based on a generative adversarial network (GAN) was constructed. The generator (G) was used to generate a forecast sequence based on historical power generation data and meteorological parameters, and the discriminator (D) was used to verify the authenticity of the sequence through adversarial training. The adversarial training of the generator and the discriminator approximated the Nash equilibrium, achieving high-precision RE scenario generation, realizing dynamic modeling and high-fidelity generation of output scenarios, and solving the problem of insufficient prediction accuracy caused by traditional random optimization relying on fixed probability distribution, providing reliable input for multi-park scheduling.

[0075] In some specific embodiments, the loss function of the generator is:

[0076] in, For noisy data Normal distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the generated data, ranging from 0 to 1; The loss function of the discriminator is:

[0077] in, It is real data obey the distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the real data, ranging from 0 to 1; The training process of the generative adversarial network is: The generator and the discriminator perform adversarial training by alternately optimizing the objective function to approach the Nash equilibrium and obtain a pre-trained generative adversarial network. The objective function is:

[0078] in, is the adversarial loss of the generative adversarial network.

[0079] By defining the loss functions of the generator and the discriminator and alternately optimizing the adversarial training process, the generator is used to minimize the probability of the generated data being recognized by the discriminator, and the discriminator is used to maximize the ability to distinguish between real and generated data. This achieves the Nash equilibrium approximation of the generative adversarial network and improves the consistency between the predicted sequence and the actual output distribution.

[0080] In some specific embodiments, the generator uses an LSTM network structure to generate a prediction sequence, and its output function for:

[0081] in, is the nonlinear mapping of the LSTM unit; is the weight matrix of the generator; is the bias term; is a noise vector that obeys the normal distribution; The historical output and historical meteorological parameters; The discriminator uses convolutional neural network to extract data features, and its feature extraction function for:

[0082] in, is the feature map of the convolutional neural network; is the convolution kernel weight; is the bias term; is generated data or real data series.

[0083] By using the LSTM network to construct the generator output function and the convolutional neural network to construct the discriminator feature extraction function, LSTM is used to capture the output timing dependency characteristics and CNN is used to extract local data features, which enhances the temporal continuity of the generated sequence and the feature identification ability of the discriminator, solving the problem that traditional linear models cannot handle nonlinear fluctuations.

[0084] Step S2: construct a comprehensive energy optimization model, input the renewable energy power generation power forecast sequence of each park into the comprehensive energy optimization model, and the optimization goal is to minimize the total system cost under the premise of meeting the constraints. Output the equipment scheduling instructions and energy interaction strategy of each park. The total system cost includes energy interaction cost and carbon trading cost considering user satisfaction.

[0085] By constructing a comprehensive energy optimization model that includes energy interaction costs and carbon trading costs that take user satisfaction into consideration, integrating carbon emissions trading (ETS) with user satisfaction, and using a tiered carbon reward and punishment mechanism and heat network demand response (DR) to quantify the benefits of users participating in carbon emission reduction, the prediction sequence is input into the model to solve the optimal scheduling instructions and interaction strategies that meet the electricity-gas-heat network balance constraints, multi-objective collaborative optimization is achieved, minimizing the total system cost while ensuring user comfort, and reducing the difficulty of balancing carbon emissions and economic efficiency.

[0086] During the operation of the multi-park integrated energy system, the ETS mechanism dynamically adjusts carbon trading costs through a hierarchical reward and punishment strategy: when the actual carbon emissions of a park integrated energy system (PIES) are lower than the quota, the park can sell surplus quotas to other parks belonging to the same multi-park integrated energy system to make a profit; if the quota exceeds the standard, it must be purchased at a high price.

[0087] In some specific embodiments, the objective function of the comprehensive energy optimization model is:

[0088] in, is the total cost of individual operation of the nth park; T is the total number of time periods; is the carbon trading cost of the park in the tth period considering user satisfaction:

[0089] is the carbon trading cost of the park in period t, The cost of adapting the peak load benefits to users in this park; The cost of electricity purchased from the distribution station in the park during the tth period; The electricity sales revenue of the distribution station in the park in the tth period; is the gas purchase cost of the gas distribution station in the park during the tth period; is the gas sales revenue of the gas distribution station in the park during the tth period; is the PT operating cost of the park in the tth period; is the operating cost of the P2G equipment in the park during the tth period.

[0090] The objective function comprehensively considers carbon trading costs, energy purchase and sales costs, and user satisfaction, minimizing overall operating costs while satisfying the balance constraints of electricity, gas, and heat network nodes and equipment operating limits. The model enhances system flexibility through the carbon absorption effect of P2G devices and utilizes the DR mechanism to guide users to reduce load during peak hours, forming a carbon revenue sharing mechanism between operators and users, ultimately achieving the coordinated optimization of a low-carbon economy and user comfort.

[0091] In some specific embodiments, the carbon trading cost of each park is calculated based on the tiered carbon reward and penalty mechanism, including: Calculate the carbon emission rights that each park needs to trade in the carbon trading market :

[0092] in, is the actual carbon emissions, for carbon emission quotas; >0 means the park needs to purchase carbon emission rights in the carbon trading market. <0The park needs to sell carbon emission rights in the carbon trading market; Calculate the carbon trading cost of the tth park ;

[0093] in, is the function for calculating carbon trading costs; A preset threshold for carbon emission rights; is the benchmark carbon price, determined by the carbon trading market; and is the preset penalty coefficient.

[0094] By setting carbon price penalty coefficients corresponding to different excess carbon emission ranges through a tiered carbon reward and punishment mechanism and dynamically calculating carbon trading costs using piecewise functions, the problem of insufficient incentives in the traditional linear carbon price model is solved, and the revenue enhancement of low-carbon emission parks and the cost constraints of high-carbon emission parks are accurately matched.

[0095] In some embodiments, carbon emission quota The calculation formula is:

[0096] in, Additional carbon emission allowances for P2G; Initial carbon emission quota without P2G:

[0097] Initial carbon emission allowances for electricity purchased from distribution stations; Initial carbon emission allowances for gas turbines; It is the initial carbon emission quota of the cogeneration unit.

[0098] By incorporating the carbon absorption effect of P2G equipment into the calculation of carbon emission quotas, increasing the additional quota allocation corresponding to renewable energy consumption, and encouraging the application of power-to-gas technology, the problem of traditional quota models ignoring the contribution of carbon sinks is solved, and the system's carbon emission reduction flexibility is improved.

[0099] In some specific embodiments, The calculation formula is:

[0100] Where, It is the preset user revenue distribution ratio coefficient; It is the additional carbon emission quota brought about by users adjusting the indoor temperature.

[0101] In some specific embodiments, step S2 further includes calculating user satisfaction , the steps include: Calculate user satisfaction with thermal energy usage :

[0102] Calculate user satisfaction with heating costs :

[0103] Calculating user satisfaction :

[0104] in, The indoor temperature target value set by the user in time period t; The minimum indoor temperature value acceptable to the user during time period t; is the time series of indoor temperature actually detected during period t; The expected value of heating energy cost set by the user in period t; The upper limit of the maximum heat energy cost acceptable to the user in period t; The minimum heat energy cost acceptable to the user in period t; 、 is the preset weight coefficient, satisfying .

[0105] User satisfaction is quantified as a composite function of heat energy usage and cost. Heat energy usage satisfaction is related to indoor temperature uniformity, while cost satisfaction is linked to tiered pricing strategies. By defining quantitative models for user satisfaction with heat energy usage and cost, and combining the characteristics of indoor temperature uniformity with tiered pricing strategies to construct a composite satisfaction parameter, we can comprehensively reflect how user satisfaction changes during the energy scheduling process, providing quantifiable demand response indicators for scheduling models and facilitating the balanced adjustment of user comfort and energy costs.

[0106] In some specific embodiments, The calculation method is:

[0107] Where J is the preset number of historical periods, indicating the number of temperature lags considered in modeling; represents the indoor temperature during time period tj; represents the water supply temperature during time period tj; represents the outdoor temperature during time period tj; 、 、 is the weight coefficient corresponding to the historical period, which is determined by fitting historical data.

[0108] By establishing an autoregressive moving average model of indoor temperature and integrating the fitting weight coefficients of historical indoor temperature, water supply temperature and outdoor temperature data, dynamic modeling of user thermal inertia characteristics is achieved, and the matching degree between temperature prediction accuracy and heating scheduling strategy is improved.

[0109] In some specific embodiments, the constraints of step S2 include: Grid node balance constraints:

[0110] Gas network node balance constraints:

[0111] Heat network node balance constraints:

[0112] in, is the total electrical load during time period t; The exchange power of the distribution station in time period t; is the wind turbine power consumption in period t; is the photovoltaic power consumption in period t; is the power consumption of the P2G device in time period t; is the power generated by the gas turbine during period t; is the electric power of the cogeneration unit in period t; Exchange power for the storage device during time period t; is the line active power loss flowing from node i in period t; is the line active power loss injected into node j during period t; is the total gas load during period t; is the gas exchange volume of the gas distribution station in time period t; is the P2G gas production in period t; is the gas consumption of the gas turbine in period t; is the gas exchange volume of the gas storage equipment in time period t; is the gas loss from node i during period t; is the gas loss injected into node j during period t; is the total heat load during period t; is the heating power of the waste heat boiler in time period t; is the thermal power of the cogeneration unit in time period t.

[0113] By setting balance constraint equations for electricity, gas, and heat network nodes, the real-time power balance under multi-energy flow coupling is forced to be met, solving the problem of traditional single-energy network scheduling models ignoring the multi-energy complementary characteristics, and ensuring the safe and stable operation of the integrated energy system.

[0114] In some specific embodiments, the constraint condition also includes satisfying the return water temperature The dynamic equation of:

[0115] in, 、 、 is the weight coefficient of the return water temperature model.

[0116] By constructing a dynamic equation for return water temperature and correlating it with historical water supply temperature and indoor and outdoor temperature data, the temperature attenuation effect during heat network transmission is quantified, achieving accurate modeling of heating network losses and improving the efficiency of waste heat utilization and the accuracy of heating cost calculation.

[0117] Step S3: Execute the equipment dispatching instructions of each park and monitor the actual power generation data of renewable energy to obtain the actual power generation sequence of renewable energy for each park. Calculate the predicted deviation of renewable energy power generation for each park based on the predicted power generation sequence of renewable energy and the actual power generation sequence of renewable energy:

[0118] in, is the forecast deviation of renewable energy power generation in the nth park; is the actual power generation sequence of renewable energy in the nth park is the actual power generation sequence of renewable energy in the nth park, .

[0119] By calculating the absolute deviation between the actual renewable energy power generation of each park and the predicted sequence, the forecast error of renewable energy output is quantified, providing deviation data support for the subsequent improvement of the dynamic adjustment of the Shapley value algorithm. This solves the defect of traditional methods that ignore the impact of forecast fluctuations on energy interaction costs and improves the objectivity of error correlation analysis.

[0120] Step S4: Calculate the energy interaction cost and energy sharing benefit of each park based on the energy interaction strategy. Based on the forecast deviation of renewable energy power generation, energy interaction cost and energy sharing benefit, use the improved Shapley value algorithm to calculate the actual marginal contribution value of each park. The formula is:

[0121]

[0122] in, is the actual marginal contribution value of the nth park; is the confidence error adjustment coefficient; The highest value among all the forecast deviations of renewable energy power generation in the park; is the original marginal contribution value of the nth park without considering the error adjustment, calculated using the traditional Shapley value algorithm.

[0123] By introducing the confidence error adjustment coefficient and the maximum deviation normalization factor, the Shapley value algorithm is improved to calculate the actual marginal contribution value, and the prediction deviation is dynamically embedded in the profit distribution weight. This solves the problem that traditional game distribution does not consider error responsibility, and achieves a precise match between the profit incentives of high-precision prediction parks and the cost penalties of low-precision parks.

[0124] In some specific embodiments, the following method is used to calculate : Calculate the objective function:

[0125] Where N is the total number of parks in the multi-park integrated energy system, The operating cost of the nth park is:

[0126] The cost generated by the interaction between the nth park and the mth park; The cost generated by the interaction between the nth park and the mth park; The energy sharing income of the nth park is:

[0127] S is any sub-alliance that does not include park n; is the number of members of sub-alliance S; is the original cooperation value of sub-alliance S; is the probability weight of park n when joining sub-alliance S.

[0128] By defining the total cost objective function of the multi-park cooperative game and the calculation formula for the operating cost of a single park, and combining the traditional Shapley value algorithm to calculate the original marginal contribution value, a benchmark allocation basis is provided for improving the Shapley value, and the initialization modeling problem of cost sharing in multi-agent collaborative scheduling is solved.

[0129] In some specific embodiments, The generation formula is:

[0130] in, Represents a Gaussian distribution function with a mean of 0 and a variance of 1; is the confidence error of the generator; is the confidence error of the discriminator.

[0131] By generating an error adjustment coefficient based on the Gaussian distribution function of the generator and discriminator output confidence, the reward and punishment weights of the Shapley value algorithm are dynamically corrected, solving the distribution inaccuracy problem caused by model fluctuations during adversarial training and improving the correlation between the profit distribution mechanism and the stability of GAN training. Step S5: distribute the benefits of the multi-park integrated energy system according to the actual marginal contribution value of each park.

[0132] The multi-park profit distribution is carried out according to the actual marginal contribution value, and the difference in contribution of each park's output prediction accuracy to the total system cost is reflected through the dynamic weight coefficient. This solves the problem of lack of cooperation fairness caused by the fixed distribution model and ensures the long-term stability of multi-agent collaborative scheduling.

[0133] In a specific embodiment, a low-carbon scheduling method for a multi-park integrated energy system is used to schedule a multi-park integrated energy system (PIESs). The architecture diagram of the multi-park integrated energy system is as follows: Figure 2 As shown in Figure 1, there are three park-level integrated energy systems in the multi-park integrated energy system, namely PIES A, PIES B, and PIES C. The steps of the low-carbon scheduling method include: Step S1: Obtain historical renewable energy output data and historical meteorological parameters for each park in the multi-park integrated energy system, and generate a corresponding renewable energy power generation prediction sequence for each park using a pre-trained generative adversarial network. The generator generates the prediction sequence based on the historical power generation data and meteorological parameters of each park, and the discriminator verifies the authenticity of the prediction sequence through adversarial training. Among them, the generator uses the LSTM network structure to generate the prediction sequence, and its output function for:

[0134] in, is the nonlinear mapping of the LSTM unit; is the weight matrix of the generator; is the bias term; is a noise vector that obeys the normal distribution; The historical output and historical meteorological parameters; The discriminator uses convolutional neural network to extract data features, and its feature extraction function for:

[0135] in, is the feature map of the convolutional neural network; is the convolution kernel weight; is the bias term; To generate data or real data series; The loss function of the generator is:

[0136] in, For noisy data Normal distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the generated data, ranging from 0 to 1; The loss function of the discriminator is:

[0137] in, It is real data obey the distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the real data, ranging from 0 to 1; The training process of the generative adversarial network is: The generator and the discriminator perform adversarial training by alternately optimizing the objective function to approach the Nash equilibrium and obtain a pre-trained generative adversarial network. The objective function is:

[0138] in, is the adversarial loss of the generative adversarial network.

[0139] Step S2: Construct a comprehensive energy optimization model. Input the renewable energy power generation prediction sequence of each park into the comprehensive energy optimization model. The optimization goal is to minimize the total system cost while satisfying the constraints. Output the equipment scheduling instructions and energy interaction strategy for each park. The total system cost includes energy interaction cost and carbon trading cost considering user satisfaction. The objective function of the comprehensive energy optimization model is:

[0140] in, is the total cost of individual operation of the nth park; T is the total number of time periods; is the carbon trading cost of the park in the tth period considering user satisfaction:

[0141] is the carbon trading cost of the park in period t, The cost of adapting the peak load benefits to users in this park; The calculation formula is:

[0142] It is the preset user revenue distribution ratio coefficient; It is the additional carbon emission quota brought by users adjusting indoor temperature; The cost of electricity purchased from the distribution station in the park during the tth period; The electricity sales revenue of the distribution station in the park in the tth period; is the gas purchase cost of the gas distribution station in the park during the tth period; is the gas sales revenue of the gas distribution station in the park during the tth period; is the PT operating cost of the park in the tth period; is the operating cost of the P2G equipment in the park during the tth period; Calculate the carbon trading costs of each park based on the tiered carbon reward and penalty mechanism, including: Calculate the carbon emission rights that each park needs to trade in the carbon trading market :

[0143] in, is the actual carbon emissions, for carbon emission quotas; >0 means the park needs to purchase carbon emission rights in the carbon trading market. <0The park needs to sell carbon emission rights in the carbon trading market; Calculate the carbon trading cost of the tth park ;

[0144] in, is the function for calculating carbon trading costs; A preset threshold for carbon emission rights; is the benchmark carbon price, determined by the carbon trading market; and is the preset penalty coefficient; Carbon emission quota The calculation formula is:

[0145] in, Additional carbon emission allowances for P2G; Initial carbon emission quota without P2G:

[0146] Initial carbon emission allowances for electricity purchased from distribution stations; Initial carbon emission allowances for gas turbines; Initial carbon emission allowances for the combined heat and power units; Constraints include: Grid node balance constraints:

[0147] Gas network node balance constraints:

[0148] Heat network node balance constraints:

[0149] in, is the total electrical load during time period t; The exchange power of the distribution station in time period t; is the wind turbine power consumption in period t; is the photovoltaic power consumption in period t; is the power consumption of the P2G device in time period t; is the power generated by the gas turbine during period t; is the electric power of the cogeneration unit in period t; Exchange power for the storage device during time period t; is the line active power loss flowing from node i in period t; is the line active power loss injected into node j during period t; is the total gas load during period t; is the gas exchange volume of the gas distribution station in time period t; is the P2G gas production in period t; is the gas consumption of the gas turbine in period t; is the gas exchange volume of the gas storage equipment in time period t; is the gas loss from node i during period t; is the gas loss injected into node j during period t; is the total heat load during period t; is the heating power of the waste heat boiler in time period t; is the thermal power of the cogeneration unit in time period t; The constraints also include satisfying the return water temperature The dynamic equation of:

[0150] in, 、 、 is the weight coefficient of the return water temperature model; Calculating user satisfaction , the steps include: Calculate user satisfaction with thermal energy usage :

[0151] Calculate user satisfaction with heating costs :

[0152] Calculating user satisfaction :

[0153] in, The indoor temperature target value set by the user in time period t; The minimum indoor temperature value acceptable to the user during time period t; is the time series of indoor temperature actually detected during period t; The expected value of heating energy cost set by the user in period t; The upper limit of the maximum heat energy cost acceptable to the user in period t; The minimum heat energy cost acceptable to the user in period t; 、 is the preset weight coefficient, satisfying ; The calculation method is:

[0154] Where J is the preset number of historical periods, indicating the number of temperature lags considered in modeling; represents the indoor temperature during time period tj; represents the water supply temperature during time period tj; represents the outdoor temperature during time period tj; 、 、 is the weight coefficient corresponding to the historical period, which is determined by fitting historical data.

[0155] Step S3: Execute the equipment dispatching instructions of each park and monitor the actual power generation data of renewable energy to obtain the actual power generation sequence of renewable energy for each park. Calculate the predicted deviation of renewable energy power generation for each park based on the predicted power generation sequence of renewable energy and the actual power generation sequence of renewable energy:

[0156] in, is the forecast deviation of renewable energy power generation in the nth park; is the actual power generation sequence of renewable energy in the nth park is the actual power generation sequence of renewable energy in the nth park, .

[0157] Step S4: Calculate the energy interaction cost and energy sharing benefit of each park based on the energy interaction strategy. Based on the forecast deviation of renewable energy power generation, energy interaction cost and energy sharing benefit, use the improved Shapley value algorithm to calculate the actual marginal contribution value of each park. The formula is:

[0158]

[0159] in, is the actual marginal contribution value of the nth park; is the confidence error adjustment coefficient; The highest value among all the forecast deviations of renewable energy power generation in the park; is the original marginal contribution value of the nth park without considering the error adjustment, which is calculated using the following method : Calculate the objective function:

[0160] Where N is the total number of parks in the multi-park integrated energy system, The operating cost of the nth park is:

[0161] The cost generated by the interaction between the nth park and the mth park; The cost generated by the interaction between the nth park and the mth park; The energy sharing income of the nth park is:

[0162] S is any sub-alliance that does not contain park n; is the number of members of sub-alliance S; is the original cooperation value of sub-alliance S; is the probability weight of park n when joining sub-alliance S; The generation formula is:

[0163] in, Represents a Gaussian distribution function with a mean of 0 and a variance of 1; is the confidence error of the generator; is the confidence error of the discriminator.

[0164] Step S5: distribute the benefits of the multi-park integrated energy system according to the actual marginal contribution value of each park.

[0165] Figure 3 This is a diagram of the dynamic reward and punishment mechanism based on GAN error classification in this embodiment. Figure 3 In the figure, (a) is the wind turbine prediction and error diagram of GAN, (b) is the wind turbine reward and punishment coefficient diagram under different errors, (c) is the photovoltaic prediction and error diagram of GAN, and (d) is the photovoltaic reward and punishment coefficient diagram under different errors.

[0166] It can be seen that the reward and punishment mechanism adopts: Wind power (WT) error classification: <0.05MW reward, 0.05-0.1MW light penalty, >0.2MW severe punishment ( Figure 3 (b)); Photovoltaic (PV) error management: The overall approach is similar to the traditional approach, but a 1.3x penalty is applied to the peak error at 13 / 15-16 ( Figure 3 (d)); Intelligent control of confidence: rewards when G / D fluctuation is <±5%, moderate penalties when it is ±5-10%, and enhanced correction when it is >±15% ( Figure 3 (a) Figure 3 (c)).

[0167] Figure 4 This is the result diagram of renewable energy output scenario generated in multiple parks. Figure 4 In the figure, (a) is the WT1 image generated by GAN in PIES A, (b) is the WT2 image generated by GAN in PIES A, (c) is the PV1 image generated by GAN in PIES A, (d) is the PV1 image generated by GAN in PIES B, (e) is the WT1 image generated by GAN in PIES C, and (f) is the PV1 image generated by GAN in PIES C.

[0168] from Figure 4 It can be seen that the three park-level integrated energy systems in this example exhibit spatiotemporal complementarity: PIES A (2WT+1PV) is dominated by wind power at night, PIES B (1PV) is dominated by photovoltaic power during the day, and PIES C (1WT+1PV) exhibits mixed characteristics.

[0169] The weighted average of the new energy generation error of GAN in each PIES in this embodiment is shown in Table 1: Table 1 Weighted average of GAN new energy generation errors

[0170] As shown in Table 1, the GAN error ranking for renewable energy (RE) output for each PIES is: PIES A (7.32%) > PIESC (6.15%) > PIES B (5.64%). PIES A has the largest error due to its low photovoltaic (PV) output and primary reliance on wind power (WT). PIES B, equipped only with a PV system, has the highest error accuracy. PIES C, with its reduced WT error and increased PV output, has an intermediate error value.

[0171] Figure 5 This is the power interaction and output power diagram of each park. Figure 5 In the figure, (a) is the power interaction diagram between the parks, and (b) is the output power diagram of each park.

[0172] Figure 6 This is a map of natural gas and heat output for each park. Figure 6 In the figure, (a) is the natural gas output diagram of each park, and (b) is the heat energy output diagram of each park.

[0173] The coordinated operation of multiple PIES can achieve "peak load shifting and valley filling" through power mutual assistance. Figure 5 As shown in the figure, during the nighttime period of 7:00 PM to 7:00 AM, PIES A's high WT output supplies power to PIES B and C, which lack wind turbines, with PIES B receiving the most power. During the daytime period of 9:00 AM to 6:00 PM, PIES B's high PV output, along with PIES C's, feeds back to PIES A, which has insufficient PV / WT output, with power transmission from PIES B dominating.

[0174] The energy interaction characteristics are as follows: the gas grid system is strongly coupled with the power grid, and the interaction curve fluctuations are synchronized with the power ( Figure 6 (a)); The heat network is subject to transmission loss and cost, and the maximum interactive power is only 0.2MW ( Figure 6 (b)).

[0175] from Figure 5-Figure 6 The energy interaction patterns of the three park-level integrated energy systems in this example can be seen: power mutual assistance enables daytime photovoltaic feedback (PIES B→A reaches 1.5MW) and nighttime wind power support (PIES A→B / C reaches 2.2MW). The gas-grid linkage synchronization rate reaches 92%, and the heat network is limited by losses and has a maximum interaction of only 0.2MW.

[0176] By integrating the dynamic Shapley value method and the GAN reward and punishment mechanism, it is possible to reduce carbon emissions by 14.7% while maintaining economy, providing an innovative solution for the coordinated "carbon-energy-cost" triple goals of multi-energy systems.

[0177] In order to verify the effectiveness of this embodiment, three simulation scenarios (Scenario 1-Scenario 3) are set, and the low-carbon scheduling method of this embodiment is used to schedule the three simulation scenarios respectively. The corresponding new energy consumption rate, CAN generation error, cost, and optimized scheduling results are calculated for each simulation scenario.

[0178] The three simulation scenarios are: Scenario 1: Integrated energy system considering PIES of ETS, with user satisfaction set to ≥0; Scenario 2: Integrated energy system considering PIES of ETS, with user satisfaction set to ≥10%; Scenario 3: Integrated energy system considering PIES of ETS, with user satisfaction set to ≥20%.

[0179] The cost statistics under the three simulation scenarios are shown in Table 2: Table 2 Cost statistics under different simulation scenarios

[0180] The user satisfaction statistics under each simulation scenario are shown in Table 3: Table 3 Statistics of user satisfaction in different simulation scenarios

[0181] Figure 7 This is a comparison chart of system operation indicators in different scenarios. Figure 7 In the figure, (a) is a comparison of electricity purchase costs, (b) is a comparison of GT carbon emissions, (c) is a comparison of PT carbon emissions, and (d) is a comparison of indoor temperatures. Scenario 1 corresponds to Case 3.1, Scenario 2 corresponds to Case 3.2, and Scenario 3 corresponds to Case 3.3.

[0182] It can be seen that by redistributing ETS profits to encourage users to participate in carbon emission reduction, the total system cost in scenarios 2 and 3 was reduced by 3.07% (38,470 yuan) and 0.68% (8,274 yuan), respectively. At the same time, carbon trading revenue achieved a step-by-step increase of 635.38% (8,495 yuan) and 95.74% (9,413 yuan). The cost optimization path showed significant differences: in the scenario 1→2 stage, the reduction of 23,601 yuan in electricity purchase cost became the main driving force, and the multi-energy balance was maintained by increasing the output of GT and PT ( Figure 7 (b)- Figure 7 (c) ), achieving an improvement in thermal energy satisfaction under the premise of a slight increase of 0.8% in carbon emissions; in the 2→3 stage of scenarios, the profit contribution of carbon trading reached 72%, and carbon emission quotas were released by reducing GT / PT output. Although the electricity purchase cost rebounded by 1.2% and the reduction in heat network supply led to a decrease in satisfaction, it verified the key role of the carbon market mechanism.

[0183] In order to analyze how the improved Shapley value method can further optimize the low-carbon economic dispatch of the cooperative game between PIES while considering the scenario errors generated by GAN, two simulation scenarios (Scenario 4 and Scenario 5) are set. The two simulation scenarios are: Scenario 4: Use the traditional Shapley value method to analyze the carbon emissions and costs of PIES participating in the cooperative game in Scenario 4; Scenario 5: The improved Shapley value method adopted in this embodiment is used to analyze the carbon emissions and costs of PIES participating in the cooperative game in Scenario 5.

[0184] In scenarios 4 and 5, the revenue and costs of each PIES' different cooperation methods are the same, as shown in Table 4: Table 4 Statistics of income and costs of different cooperation modes of each PIES

[0185] In Scenario 5, the additional benefits brought by the reduction of total system operating costs due to cooperative game are taken into account, and the corresponding improved Shapley value method is used to redistribute the benefits of each PIE. Table 5 shows the carbon trading income and cooperative benefits of each PIE in Scenario 4-5: Table 5 Carbon trading income and cooperation benefits of each PIES in scenario 4-5

[0186] It can be seen that among the three PIES, PIES B had the lowest carbon trading income when operating independently, at only RMB 3,472. This is because PIES B is a commercial area, and its heat network DR capacity is much lower than that of PIES A and C residential areas.

[0187] Compared with Scenario 4, in terms of carbon trading revenue, Scenario 5 shows that PIES A's revenue flows to PIES B and PIES C. Specifically, PIES A reduces its carbon trading revenue by RMB 2,084, a decrease of 6.69%, PIES B increases it by RMB 945, an increase of 19.97%, and PIES C increases it by RMB 1,139, an increase of 3.973%.

[0188] Therefore, when using the distribution method of this embodiment, namely, the Shapley value-based benefit allocation method that considers GAN errors, more revenue flows to PIES B and PIES C, which have smaller errors. Although, as mentioned above, PIES A's revenue will decrease due to the higher GAN error rate, larger errors correspond to larger penalties. In the next iteration cycle of the GAN in this system, it will optimize the adversarial generation process, forcing the system to fluctuate around the Nash equilibrium point while gradually approaching it, thereby improving the accuracy of generated RE scenarios. This, in turn, will rationally allocate resources among PIES and achieve low-carbon economic operation.

[0189] The following is an embodiment of the low-carbon scheduling system provided in the embodiments of the present application. The low-carbon scheduling system of the multi-park integrated energy system and the low-carbon scheduling methods of the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiments of the low-carbon scheduling system of the multi-park integrated energy system, please refer to the embodiments of the low-carbon scheduling method of the multi-park integrated energy system.

[0190] like Figure 8 As shown in Figure 1, the low-carbon dispatch system of the multi-park integrated energy system includes: The data acquisition module is used to obtain the historical renewable energy output data and historical meteorological parameters of each park in the multi-park integrated energy system; Renewable energy generation power prediction module, which is used to generate a corresponding renewable energy generation power prediction sequence for each park through a pre-trained generative adversarial network; The integrated energy optimization module is used to build an integrated energy optimization model, input the renewable energy power generation prediction sequence of each park into the integrated energy optimization model, and output the equipment scheduling instructions and energy interaction strategies of each park; The instruction execution and deviation calculation module is used to execute the equipment scheduling instructions of each park and monitor the actual power generation data of renewable energy, obtain the actual power generation sequence of renewable energy in each park, and calculate the predicted deviation of renewable energy power generation in each park based on the predicted power generation sequence of renewable energy and the actual power generation sequence of renewable energy; The marginal contribution evaluation module is used to calculate the energy interaction cost and energy sharing benefits of each park based on the energy interaction strategy. Based on the forecast deviation of renewable energy power generation, energy interaction cost and energy sharing benefits, the improved Shapley value algorithm is used to calculate the actual marginal contribution value of each park; The profit distribution module is used to distribute the profits of the multi-park integrated energy system according to the actual marginal contribution value of each park.

[0191] The low-carbon scheduling system of this embodiment is used to implement a low-carbon scheduling method for a multi-park integrated energy system.

[0192] This application also provides an electronic device for implementing each embodiment of this application. Figure 9 A hardware structure diagram of an electronic device for implementing various embodiments of the present application is shown in FIG. Figure 9 As shown, the electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0193] Those skilled in the art will understand that the electronic device structure involved in the embodiments of the present application does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0194] In the embodiments of the present application, electronic devices include, but are not limited to, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or claimed herein.

[0195] In the embodiment of the present application, the processor can be implemented by using at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a processor, a controller, a microcontroller, a microprocessor, and an electronic unit designed to perform the functions described herein. In some cases, such an embodiment can be implemented in a controller. For software implementation, an embodiment such as a process or function can be implemented with a separate software module that allows the execution of at least one function or operation. The software code can be implemented by a software application (or program) written in any appropriate programming language, and the software code can be stored in a memory and executed by a controller.

[0196] In addition, the electronic device includes some functional modules not shown, which will not be described here.

[0197] Those skilled in the art will appreciate that various aspects of the electronic device provided herein may be implemented as a system, method, or program product. Therefore, various aspects of the present application may be specifically implemented in the following forms: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to herein as "circuits," "modules," or "systems."

[0198] The present application also provides a storage medium storing a program product capable of implementing a low-carbon scheduling method for a multi-park integrated energy system. In some possible implementations, various aspects of the present application may also be implemented in the form of a program product comprising program code that, when executed on a terminal device, causes the terminal device to execute the steps described in the "Exemplary Methods" section above of this specification according to various exemplary implementations of the present application.

[0199] The storage medium can be any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0200] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A low-carbon scheduling method for a multi-park integrated energy system, characterized in that: include: S1. Obtain historical renewable energy output data and historical meteorological parameters for each park in a multi-park integrated energy system. Use a pre-trained generative adversarial network to generate a corresponding renewable energy power generation prediction sequence for each park. The generator generates the prediction sequence based on each park's historical power generation data and meteorological parameters, and the discriminator verifies the authenticity of the prediction sequence through adversarial training. S2. Build a comprehensive energy optimization model. Input the forecasted renewable energy generation power series for each park into the model. The optimization goal is to minimize the total system cost while satisfying the constraints. Output the equipment scheduling instructions and energy interaction strategy for each park. The total system cost includes energy interaction costs and carbon trading costs that take into account user satisfaction. S3. Execute the equipment dispatch instructions for each park and monitor the actual renewable energy power generation data to obtain the actual renewable energy power generation sequence for each park. Calculate the predicted renewable energy power generation deviation for each park based on the predicted renewable energy power generation sequence and the actual renewable energy power generation sequence: in, is the forecast deviation of renewable energy power generation in the nth park; is the actual power generation sequence of renewable energy in the nth park; is the actual power generation sequence of renewable energy in the nth park, ; S4. Calculate the energy interaction costs and energy sharing benefits of each park based on the energy interaction strategy. Use the improved Shapley value algorithm to calculate the actual marginal contribution of each park based on the forecast deviation of renewable energy power generation, energy interaction costs, and energy sharing benefits. The formula is: in, is the actual marginal contribution value of the nth park; is the confidence error adjustment coefficient; The highest value among all the forecast deviations of renewable energy power generation in the park; is the original marginal contribution value of the nth park without considering the error adjustment, calculated using the traditional Shapley value algorithm; S5. Allocate the benefits of the multi-park integrated energy system based on the actual marginal contribution of each park.

2. The low-carbon scheduling method according to claim 1, characterized in that: In step S1, the loss function of the generator is: in, For noisy data Normal distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the generated data, ranging from 0 to 1; The loss function of the discriminator is: in, It is real data obey the distribution; It's about Expectation of a distribution; is the probability of the discriminator successfully distinguishing the real data, ranging from 0 to 1; The training process of the generative adversarial network is: The generator and the discriminator perform adversarial training by alternately optimizing the objective function to approach the Nash equilibrium and obtain a pre-trained generative adversarial network. The objective function is: in, is the adversarial loss of the generative adversarial network.

3. The low-carbon scheduling method according to claim 2, characterized in that: The generator uses the LSTM network structure to generate the prediction sequence, and its output function for: in, is the nonlinear mapping of the LSTM unit; is the weight matrix of the generator; is the bias term; is a noise vector that obeys the normal distribution; The historical output and historical meteorological parameters; The discriminator uses convolutional neural network to extract data features, and its feature extraction function for: in, is the feature map of the convolutional neural network; is the convolution kernel weight; is the bias term; is generated data or real data series.

4. The low-carbon scheduling method according to claim 1, characterized in that: In step S2, the objective function of the comprehensive energy optimization model is: in, is the total cost of individual operation of the nth park; T is the total number of time periods; is the carbon trading cost of the park in the tth period considering user satisfaction: is the carbon trading cost of the park in period t, The cost of adapting the peak load benefits to users in this park; The cost of electricity purchased from the distribution station in the park during the tth period; The electricity sales revenue of the distribution station in the park in the tth period; is the gas purchase cost of the gas distribution station in the park during the tth period; is the gas sales revenue of the gas distribution station in the park during the tth period; is the PT operating cost of the park in the tth period; is the operating cost of the P2G equipment in the park during the tth period.

5. The low-carbon scheduling method according to claim 4, characterized in that: Calculate the carbon trading costs of each park based on the tiered carbon reward and penalty mechanism, including: Calculate the carbon emission rights that each park needs to trade in the carbon trading market : in, is the actual carbon emissions, for carbon emission quotas; >0 means the park needs to purchase carbon emission rights in the carbon trading market. <0The park needs to sell carbon emission rights in the carbon trading market; Calculate the carbon trading cost of the tth park ; in, is the function for calculating carbon trading costs; A preset threshold for carbon emission rights; is the benchmark carbon price, determined by the carbon trading market; and is the preset penalty coefficient.

6. The low-carbon scheduling method according to claim 5, characterized in that: The calculation formula is: Where, It is the preset user benefit distribution ratio coefficient; It is the additional carbon emission quota brought about by users adjusting the indoor temperature.

7. The low-carbon scheduling method according to claim 1, characterized in that: The constraints of step S2 include: Grid node balance constraints: Gas network node balance constraints: Heat network node balance constraints: in, is the total electrical load during time period t; The exchange power of the distribution station in time period t; is the wind turbine power consumption in period t; is the photovoltaic power consumption in period t; is the power consumption of the P2G device in time period t; is the power generated by the gas turbine during period t; is the electric power of the cogeneration unit in period t; Exchange power for the storage device during time period t; is the line active power loss flowing from node i in period t; is the line active power loss injected into node j during period t; is the total gas load during period t; is the gas exchange volume of the gas distribution station in time period t; is the P2G gas production in period t; is the gas consumption of the gas turbine in period t; is the gas exchange volume of the gas storage equipment in time period t; is the gas loss from node i during period t; is the gas loss injected into node j during period t; is the total heat load during period t; is the heating power of the waste heat boiler in time period t; is the thermal power of the cogeneration unit in time period t.

8. The low-carbon scheduling method according to claim 1, characterized in that: In step S4, the following method is used to calculate : Calculate the objective function: Where N is the total number of parks in the multi-park integrated energy system, The operating cost of the nth park is: The cost generated by the interaction between the nth park and the mth park; The cost generated by the interaction between the nth park and the mth park; The energy sharing income of the nth park is: S is any sub-alliance that does not contain park n; is the number of members of sub-alliance S; is the original cooperation value of sub-alliance S; is the probability weight of park n when joining sub-alliance S.

9. The low-carbon scheduling method according to claim 1, characterized in that: The generation formula is: in, Represents a Gaussian distribution function with a mean of 0 and a variance of 1; is the confidence error of the generator; is the confidence error of the discriminator.

10. A low-carbon dispatching system for a multi-park integrated energy system, characterized in that: A method for implementing the low-carbon scheduling method according to any one of claims 1 to 9, comprising: The data acquisition module is used to obtain the historical renewable energy output data and historical meteorological parameters of each park in the multi-park integrated energy system; Renewable energy generation power prediction module, which is used to generate a corresponding renewable energy generation power prediction sequence for each park through a pre-trained generative adversarial network; The integrated energy optimization module is used to build an integrated energy optimization model, input the renewable energy power generation prediction sequence of each park into the integrated energy optimization model, and output the equipment scheduling instructions and energy interaction strategies of each park; The instruction execution and deviation calculation module is used to execute the equipment scheduling instructions of each park and monitor the actual power generation data of renewable energy, obtain the actual power generation sequence of renewable energy in each park, and calculate the predicted deviation of renewable energy power generation in each park based on the predicted power generation sequence of renewable energy and the actual power generation sequence of renewable energy; The marginal contribution evaluation module is used to calculate the energy interaction cost and energy sharing benefits of each park based on the energy interaction strategy. Based on the forecast deviation of renewable energy power generation, energy interaction cost and energy sharing benefits, the improved Shapley value algorithm is used to calculate the actual marginal contribution value of each park; The profit distribution module is used to distribute the profits of the multi-park integrated energy system according to the actual marginal contribution value of each park.

Citation Information

Cited By

  • Carbon emission scheduling method, device, equipment and medium

    CN122088998A

  • Climbing demand node mapping method based on multi-source contribution degree distribution

    CN122115015A