A day-ahead strategy generation method for coordinated bidding of an electricity-carbon market and a related device

Abstract

This application discloses a day-ahead strategy generation method and related apparatus for collaborative bidding in the electricity carbon market, relating to the field of energy management technology. The day-ahead strategy generation method includes: for wind and solar power units, quantifying the power generation fluctuation range at the target time based on the deviation between historical forecasts and actual power; constructing a 0-1 polyhedral uncertainty set characterizing discrete operational risks by combining equipment binary state variables and fault constraints; introducing a vector autoregressive model for the electricity carbon market to capture the dynamic coupling relationship between electricity and carbon market prices, thereby constructing a two-layer IGDT robust risk avoidance architecture: the upper layer aims to maximize the net revenue of the electricity carbon market, forming a bidding decision model; the lower layer aims to minimize operating costs under the worst-case uncertainty scenario, forming a scheduling model. By jointly solving the fluctuation range, uncertainty set, and market model, a day-ahead bidding strategy that balances economic efficiency and risk resistance is generated.

Description

Technical Field

[0001] This application relates to the field of energy management technology, and in particular to a method and apparatus for generating day-ahead strategies for collaborative bidding in the electricity carbon market. Background Technology

[0002] An integrated energy virtual power plant refers to a virtual entity that aggregates and coordinates geographically dispersed distributed photovoltaic, wind power, gas turbines, energy storage systems, power-to-gas equipment, electric vehicles, flexible temperature-controlled loads, and other heterogeneous energy resources through advanced information and communication technologies and intelligent control architectures, forming a virtual entity with equivalent power generation, regulation capabilities, and market participation capabilities.

[0003] With the deepening of electricity market system reform, integrated energy virtual power plants are gradually becoming qualified to participate in both the day-ahead electricity market and the carbon trading market, aiming to maximize profits through cross-market coupling optimization.

[0004] However, existing bidding strategies face severe challenges in their formulation: First, the intermittent nature of wind and solar power output leads to strong uncertainty in power output forecasting. Traditional probability distribution methods rely on a large amount of historical data and easily overlook extreme weather risks, while fuzzy optimization rules are too conservative, resulting in reduced returns. Second, frequent random failures of internal equipment and the fact that existing models often ignore the correlation logic between equipment failures lead to a disconnect between the amount of bids and available resources, resulting in high default penalties. Third, existing models struggle to balance high returns with high risks and often adopt a sequential decision-making model of "electricity first, carbon later," which often results in mismatched decisions by mis-adjusting high-carbon units during periods of high carbon prices.

[0005] How to construct a comprehensive energy virtual power plant day-ahead bidding strategy generation method that can accurately quantify the multiple uncertainties of wind and solar power output and equipment failure, balance revenue and risk, and achieve bidirectional synergistic optimization of the electricity-carbon market has become an urgent technical problem to be solved in the field of energy management technology. Summary of the Invention

[0006] To address the aforementioned issues, this application provides a method for generating day-ahead strategies for collaborative bidding in the electricity carbon market.

[0007] The embodiments of this application disclose the following technical solutions: The first aspect of this application provides a day-ahead strategy generation method for collaborative bidding in the electricity carbon market, the method being applied to an integrated energy virtual power plant, the method comprising: For the new energy units of the integrated energy virtual power plant, based on the deviation between the historical predicted power generation sequence and the corresponding historical actual power generation sequence of the new energy units, and the predicted power generation at the target sampling time, the fluctuation range of the power generation of the new energy units at the target sampling time is determined; the new energy units include wind power generation equipment and photovoltaic power generation equipment. A 0-1 polyhedral uncertainty set is constructed based on the binary state variables of each device in the integrated energy virtual power plant and the fault constraint relationship between each device. A two-layer IGDT robust risk avoidance model is constructed, which includes an upper-layer bidding decision model aimed at maximizing the net revenue of the integrated energy virtual power plant in the carbon electricity market, and a lower-layer scheduling model aimed at minimizing the operating cost of the integrated energy virtual power plant under the worst-case uncertainty scenario. A vector autoregression model for the electricity carbon market is obtained; this model is used to characterize the dynamic coupling relationship and linkage trend among electricity carbon market prices. Based on the fluctuation range of the power generation of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set, and the vector autoregressive model of the electricity carbon market, the two-layer IGDT robust risk avoidance model is solved to generate a day-ahead bidding strategy.

[0008] In one optional implementation, for the new energy generating units of the integrated energy virtual power plant, based on the deviation between the historical predicted power generation sequence and the corresponding historical actual power generation sequence of the new energy generating units, and the predicted power generation at the target sampling time, the fluctuation range of the power generation of the new energy generating units at the target sampling time is determined, including: The error sample sequence is obtained by subtracting the historical actual power generation sequence from the predicted power generation sequence element by element. Based on the error sample sequence, an empirical distribution function is constructed, and with the empirical distribution function as the center and the Wasserstein distance as the radius, the output fuzzy set corresponding to the predicted power generation of the new energy unit at the target sampling time is constructed. Based on the output fuzzy set, the fluctuation range of the power generation of the new energy unit at the target sampling time is determined.

[0009] In one alternative implementation, the upper-level bidding decision model is:

[0010]

[0011]

[0012]

[0013] in, It is the target value of the upper-level bidding decision model; It is the predicted electricity price at time t. yes t Predicting carbon prices at any given time P e,t It represents the day-ahead electricity market bid volume at time t. P c,t It represents the day-ahead carbon market bid volume at time t. F 2 It is the target value of the lower-level scheduling model; j This represents the equipment index in the aforementioned integrated energy virtual power plant. For equipment j Rated power, J cons This represents a collection of discharge devices. J gen It represents a collection of charging devices; The external power ramp-up limit per minute for the aforementioned integrated energy virtual power plant; This is the maximum purchase quota allowed per time period. It is the upper limit of the surplus quota that can be sold in a single time period, P all This represents the total transaction volume within the contract fulfillment period.

[0014] In one optional implementation, the lower-level scheduling model is:

[0015]

[0016]

[0017]

[0018] in, F 2 It is the target value of the lower-level scheduling model; U represents the feasible region of equipment failure uncertainty; y This represents the set of decision variables corresponding to the lower-level scheduling model. It is the scheduling feasible region, determined by the bid quantity. x =( P e,t , P c,t ), Equipment status u Wind and light error Common constraints J This represents the collection of all equipment in the integrated energy virtual power plant. jThis represents the equipment index in the aforementioned integrated energy virtual power plant. C op,j This represents the unit operating cost of device j. P j,t It is the output of device j at time t. C qd,j This represents the single startup cost of device j. v j,t This indicates the startup state of device j at time t. ；P grid,t This represents the real-time power deviation adjustment between the power grid and the target power at time t. in, S es,t It is the stored energy at time t. S es,t-1 yes t The stored energy at time -1; P ch,es,t It is the energy storage charging power at time t. P dis,es,t yes t The energy storage and discharge power at any given moment; η ch,es It refers to energy storage charging efficiency. η dis,es It refers to the energy storage and discharge efficiency; E gt,t It is the carbon emissions of the gas turbine at time t. E p2g,t It represents the amount of carbon absorbed by the electro-gas converter at time t. E grid,t It represents the reduction in carbon emissions at time t compared to purchasing electricity from the grid.

[0019] In one alternative implementation, the vector autoregressive model for the electricity carbon market is specifically:

[0020] in, , , c e and c c These are the intercept terms for electricity price and carbon price, respectively. p This indicates the lag order of the model. Indicates lag k The autoregressive coefficient matrix of the period, The coefficient matrix of exogenous variables, x t For exogenous variable vectors, , R xny,t for t The proportion of renewable energy power generation at any given time Pload,t Let t be the virtual power plant load. D t It's a policy event. This is the white noise error term.

[0021] In one optional implementation, the step of solving the two-layer IGDT robust risk aversion model based on the fluctuation range of the power generation of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set, and the electric carbon market vector autoregressive model to generate a day-ahead bidding strategy includes: Based on preset risk avoidance parameters, the radius of the fluctuation range of the power generation and the fault tolerance upper limit of the 0-1 polyhedral uncertainty set are adjusted respectively to define the uncertainty boundary. Based on the strong duality theorem, the minimum operating cost objective function of the lower-level scheduling model and the upper-level bidding decision model are jointly dualized under the constraint of the uncertainty boundary to obtain a single-layer robust optimization model. An adaptive chaotic quantum particle swarm optimization algorithm is used, with the output of the electric carbon market vector autoregressive model as the price signal, to optimize the single-layer robust optimization model under the constraint of the uncertainty boundary, thereby generating the day-ahead bidding strategy.

[0022] In one optional implementation, the process of employing an adaptive chaotic quantum particle swarm optimization algorithm, using the output of the electric carbon market vector autoregressive model as the price signal, and optimizing the single-layer robust optimization model under the constraints of the uncertainty boundary to generate the day-ahead bidding strategy includes: Calculate the variance of the fitness of all particles in the current population at the current iteration round, and define the variance as the fitness variance, which is used to measure the degree of aggregation of the population; The number of iterations in which the global optimal position remains unchanged is counted, and this number of iterations is defined as the evolutionary stagnation degree; the evolutionary stagnation degree is used to measure the convergence state of the algorithm. When the fitness variance is lower than a preset first threshold or the evolutionary stagnation exceeds a preset second threshold, it is determined that the current population is trapped in a local optimum, and the contraction and expansion coefficient is automatically increased to force the particles to perform a large-scale search. When the fitness variance is higher than the first threshold and the evolutionary stagnation is lower than the second threshold, the current population is determined to be in a normal search state, and the contraction-expansion coefficient is maintained or reduced to achieve precise search.

[0023] In one optional implementation, the process of employing an adaptive chaotic quantum particle swarm optimization algorithm, using the output of the electric carbon market vector autoregressive model as the price signal, and optimizing the single-layer robust optimization model under the constraints of the uncertainty boundary to generate the day-ahead bidding strategy includes: Based on a preset ratio, the current population in the current iteration round is divided into an exploration group and a development group; The exploration swarm is controlled to perform quantum position updates with a contraction-expansion coefficient greater than 1, so as to make large-scale jumps in the solution space and maintain population diversity; The development group is controlled to perform quantum position updates with a contraction and expansion coefficient of less than 1 in order to perform a fine search in the neighborhood of the current optimal solution and accelerate the convergence of the algorithm. Every preset exchange cycle, the elite particles with the best fitness in the exploration group are migrated to the development group, and the global optimal position of the development group is fed back to the exploration group to achieve complementary evolution between the two groups.

[0024] A second aspect of this application provides a day-ahead strategy generation device for collaborative bidding in the electricity carbon market, applied to an integrated energy virtual power plant, comprising: The wind and solar uncertainty determination module is used to determine the fluctuation range of the power generation of the new energy units in the integrated energy virtual power plant, based on the deviation between the historical predicted power generation sequence and the corresponding historical actual power generation sequence of the new energy units, and the predicted power generation at the target sampling time; the new energy units include wind power generation equipment and photovoltaic power generation equipment. The polyhedral uncertainty set determination module is used to construct a 0-1 polyhedral uncertainty set based on the binary state variables of each device in the integrated energy virtual power plant and the fault constraint relationship between the devices. The risk aversion model construction module is used to construct a two-layer IGDT robust risk aversion model. The model includes an upper-layer bidding decision model with the goal of maximizing the net revenue of the integrated energy virtual power plant in the electricity carbon market, and a lower-layer scheduling model with the goal of minimizing the operating cost of the integrated energy virtual power plant under the worst-case uncertainty scenario. The module for determining the relationship in the electricity carbon market is used to obtain a vector autoregressive model of the electricity carbon market; the vector autoregressive model of the electricity carbon market is used to characterize the dynamic coupling relationship and linkage trend among electricity carbon market prices. The day-ahead bidding strategy generation module is used to solve the two-layer IGDT robust risk avoidance model based on the fluctuation range of the power generation of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set, and the electric carbon market vector autoregressive model, and generate the day-ahead bidding strategy.

[0025] A third aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any implementation of the first aspect.

[0026] A fourth aspect of this application provides an electronic device, comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method described in any implementation of the first aspect.

[0027] Compared with the prior art, this application has the following beneficial effects: This application determines the power generation fluctuation range at the target sampling time by combining the deviation between the historical predicted power generation sequence and the historical actual power generation sequence of new energy units with the predicted power generation at the target sampling time. Based on the binary state variables of each device and the fault constraint relationship between devices, a 0-1 polyhedral uncertainty set is constructed, thereby organically integrating the continuous uncertainty of new energy output with discrete equipment fault logic. On this basis, the two-layer IGDT robust risk avoidance model constructed in this application makes bidding decisions with the goal of maximizing the net profit of the integrated energy virtual power plant in the electricity carbon market at the upper layer, and optimizes scheduling with the goal of minimizing the operating cost under the worst uncertainty scenario at the lower layer. Combined with the electricity carbon market vector autoregression model to represent the dynamic coupling relationship and linkage trend between electricity carbon market prices, it realizes the electricity carbon collaborative optimization that can still ensure stable returns under extremely unfavorable operating conditions, effectively balancing market profit-seeking and system operation safety.

[0028] Furthermore, this application integrates the solution of the two-layer IGDT robust risk avoidance model based on the power generation fluctuation range, the 0-1 polyhedral uncertainty set, and the vector autoregression model of the electricity carbon market. It optimizes the strategy within the defined uncertainty boundary, avoiding decision-making errors caused by the probability distribution assumption bias of traditional stochastic optimization. This significantly improves the robustness and reliability of the bidding strategy in the face of complex and volatile market environments and physical failure risks, ensuring the engineering practical value of the day-ahead bidding strategy. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A schematic diagram of an integrated energy virtual power plant energy topology provided in this application embodiment; Figure 2 A flowchart illustrating a day-ahead strategy generation method for collaborative bidding in the electric carbon market, provided as an embodiment of this application; Figure 3 A schematic diagram of the structure of a two-layer IGDT robust risk avoidance model provided in this application embodiment; Figure 4 A schematic diagram of a day-ahead strategy generation device for collaborative bidding in the electric carbon market is provided for an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a strategy output display module provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of a master control storage module provided in an embodiment of this application. Detailed Implementation

[0031] With the development of new electricity markets, virtual power plants play a crucial role in enhancing the resilience and economic efficiency of power systems. However, virtual power plants face numerous challenges in bidding for electricity carbon markets, specifically: First, the uncertainty in wind and solar power output is difficult to quantify, leading to insufficient strategy robustness. Integrated energy virtual power plants are highly dependent on distributed renewable energy, but their output is highly intermittent. Existing mainstream processing methods have significant shortcomings: stochastic optimization methods based on precise probability distributions heavily rely on the accuracy of historical data. Once encountering unprecedented extreme weather (such as typhoons or extreme cold), the predicted distribution will shift, resulting in actual output being far lower than the bid amount and facing huge bias penalties; while optimization methods based on intervals or fuzzy sets do not rely on distributions, but often construct overly conservative confidence intervals, excessively compressing the bidding space, preventing virtual power plants from capturing excess profits during peak electricity price periods, resulting in potential revenue loss.

[0032] Secondly, oversimplification of equipment failure models and neglect of interconnected logic lead to performance risks. Existing bidding strategies often assume equipment is fully reliable or use only independent Poisson distributions to describe single-equipment failures, completely ignoring the complex physical coupling and logical interlocking relationships within a virtual power plant. For example, the operation of power-to-gas conversion equipment depends on the gas pressure support of the gas storage equipment and the carbon source supply of the gas turbine; the energy storage system and the electric vehicle cluster share the same reverse discharge protection logic on the same busbar. This simplified assumption of independent failures means that in worst-case scenarios involving multiple interconnected equipment failures or sudden shutdowns of critical support equipment, the actual adjustable resources are far less than the bid amount, directly leading to market contract defaults.

[0033] Third, the decision-making process in the electricity and carbon markets is fragmented, lacking dynamic coordination and risk hedging mechanisms. Existing optimization models mostly employ a sequential decision-making model, determining electricity volume first and then calculating carbon emissions, failing to consider the dynamic coupling between the electricity and carbon markets in their price formation mechanisms. This fragmentation leads to decision-making blind spots: when electricity prices are high but carbon prices are soaring simultaneously, strategies may still blindly deploy high-carbon units to profit, only to have profits eroded by the surge in carbon costs; conversely, failing to promptly allocate low-carbon resources when carbon prices are high can result in missed opportunities for carbon asset appreciation. Furthermore, existing models are mostly single-objective optimizations, unable to provide flexible adjustment mechanisms between pursuing high returns and mitigating uncertainty risks.

[0034] Therefore, how to construct a comprehensive energy virtual power plant day-ahead bidding strategy generation method that can accurately quantify the multiple uncertainties of wind and solar power output and equipment failure, balance revenue and risk, and achieve bidirectional synergistic optimization of the electricity-carbon market has become an urgent technical problem to be solved in the field of energy management technology.

[0035] Based on this, this application provides a day-ahead strategy generation method for collaborative bidding in the electricity carbon market. By integrating historical power deviation analysis and equipment physical interlock logic, it accurately defines multiple risk boundaries including power generation fluctuation range and 0-1 polyhedral fault uncertainty set. It also introduces an electricity carbon market vector autoregression model to capture price linkage trends, constructs a two-layer IGDT robust risk avoidance model that takes into account both electricity carbon revenue and extreme defense, and solves the model in an integrated manner. This achieves a quantitative balance between market profit and operational safety in a complex and uncertain environment, effectively overcoming the problem of overly aggressive or conservative strategies caused by ignoring equipment-level cascading faults and market interaction effects in traditional methods, and significantly improving the robustness and engineering practicality of day-ahead bidding strategies.

[0036] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0037] Figure 1 This is a schematic diagram of an integrated energy virtual power plant energy topology provided for an embodiment of this application. For example... Figure 1 As shown, this topology constructs a deeply coupled network covering multiple energy forms such as electricity, heat, natural gas, and hydrogen.

[0038] On the energy supply side, distributed photovoltaic and wind turbines are connected to the top of the topology to provide clean electricity, while power support is obtained through connection to the external power grid and fuel supply is provided by natural gas. On the energy conversion and hub side, gas turbines and power-to-gas (EPG) equipment are deployed in the central core area. The gas turbines achieve efficient gas-electricity-heat conversion by burning natural gas, while the EPG equipment integrates an electrolyzer and a methanation reactor, which can convert surplus electricity into hydrogen or synthesize natural gas from hydrogen and carbon dioxide, thereby realizing the storage of electrical energy into chemical energy and the recycling of carbon resources. On the energy storage and load side, the topology is equipped with energy storage and gas storage devices to smooth fluctuations in the electricity and gas networks, respectively. The right-side terminal connects to diversified electricity loads, including electric vehicles, heat loads, and natural gas loads. In the coupling relationship of various energy networks, the power network is composed of the interconnection of external power grid and distributed power sources, the natural gas network connects gas sources, gas storage equipment and gas turbines, the heat network recovers waste heat from gas turbines to supply heat loads, and the hydrogen energy flow runs through the electricity-to-gas reaction process, ultimately forming an energy internet architecture in which multiple energy flows of electricity, heat, gas and hydrogen interact synergistically.

[0039] for Figure 1 The integrated energy virtual power plant topology shown uses a hierarchical heterogeneous communication architecture and intelligent sensing terminals deployed on the field side to collect raw monitoring data from the power generation side, energy conversion side, energy storage side, load side, and carbon metering process.

[0040] The raw monitoring data of the integrated energy virtual power plant includes: generation side data, energy conversion side data, energy storage side data, load side data, and carbon metering data.

[0041] The data includes: power generation data (real-time output power, operating status, and power generation efficiency of distributed photovoltaic and wind turbines); energy conversion data (fuel consumption, power generation, and waste heat output of gas turbines, as well as operating parameters and conversion efficiency of electrolyzers and methanation reactors in power-to-gas conversion equipment); energy storage data (state of charge and charging / discharging power of energy storage equipment, as well as tank pressure and storage capacity of gas storage equipment); load data (real-time consumption of electricity load, heat load, and natural gas load, as well as grid connection status and charging / discharging power of electric vehicles); and carbon measurement data (carbon emission intensity, cumulative carbon emissions, and carbon capture at each stage).

[0042] After collecting the raw monitoring data associated with the preset target sampling time, it is transmitted to the edge computing node through the Modbus TCP industrial-grade protocol conversion gateway. The data is preprocessed, missing values ​​are filled and time sequence is aligned to generate a standardized running dataset corresponding to the preset target sampling time, and then transmitted to the core processing unit for other modules to use.

[0043] Among them, the preset target sampling time refers to discretizing the trading cycle of the day-ahead electricity market and carbon trading market (usually from 00:00 to 24:00 the next day) into several consecutive scheduling periods (such as 24 periods), and determining several scheduling periods as preset target sampling times.

[0044] Figure 2 A flowchart illustrating a day-ahead strategy generation method for collaborative bidding in the electric carbon market, provided as an embodiment of this application. (Combined with...) Figure 2 As shown, the day-ahead strategy generation method for collaborative bidding in the electricity carbon market provided in this application includes: S201, for the new energy generating units of the integrated energy virtual power plant, based on the deviation between the historical predicted power generation sequence of the new energy generating units and the corresponding historical actual power generation sequence, and the predicted power generation at the target sampling time, the fluctuation range of the power generation of the new energy generating units at the target sampling time is determined.

[0045] The new energy sources in this application include wind power generation equipment and photovoltaic power generation equipment.

[0046] The historical actual power generation sequence of new energy units in this application refers to a time-series dataset consisting of the measured power generation values ​​of new energy unit power generation equipment collected at fixed sampling intervals (such as every hour) within a preset time period in the past, arranged in chronological order of sampling time.

[0047] The historical predicted power generation sequence of new energy generating units in this application refers to a time series dataset consisting of predicted power generation values ​​for the same time points, obtained under the same sampling period and sampling interval as the aforementioned historical actual power generation sequence, generated by external meteorological forecasting agencies or internal prediction models.

[0048] The target sampling time in this application refers to a specific future scheduling period (such as any hourly time point between 00:00 and 24:00 the following day) before the market opens, during which a bidding strategy needs to be formulated.

[0049] It should be noted that the calculation method for the fluctuation range of power generation of wind power generation equipment at the target sampling time in this application is the same as the calculation method for the fluctuation range of power generation of photovoltaic power generation equipment at the target sampling time.

[0050] The following section uses the calculation method of the power generation fluctuation range of wind power generation equipment at the target sampling time as an example to explain in detail the method for determining the power generation fluctuation range of new energy units at the target sampling time.

[0051] First, for wind power generation equipment, the error sample sequence is obtained by subtracting the historical actual power generation sequence and the predicted power generation sequence of the wind power generation equipment element by element.

[0052] This application adopts This represents the historical predicted power generation sequence of wind power generation equipment, using... If the historical actual power generation sequence of wind power generation equipment is represented, then the corresponding error sample sequence of wind power generation equipment is... .

[0053] Secondly, an empirical distribution function is constructed based on the error sample sequence, and with the empirical distribution function as the center and the Wasserstein distance as the radius, the output fuzzy set corresponding to the predicted power generation of the new energy unit at the target sampling time is constructed.

[0054] based on N The empirical distribution is formed by the historical error samples, and the empirical distribution function is obtained.

[0055] The empirical distribution function is as follows: ;

[0056] in, Let P represent the empirical distribution function, where P is the true theoretical distribution. for P and The 1-Wasserstein distance between two probability distributions; N is the sample size, and i is the sample index. It is the error of the i-th sample. To concentrate on The Dirac measure, where π is the joint probability distribution. For all P and It is the set of joint probability distributions of marginal distributions. denoted as , where is the distance between two error samples; E is the entire range of power deviation values, also known as the support set.

[0057] Secondly, based on empirical distribution Centered on the Wasserstein distance r Using radius , construct a fuzzy set containing all possible true distributions.

[0058] The expression F for the fuzzy set is: .

[0059] in, P ( E ) for all defined in the support set E The set of probability distributions on, r The larger the value, the stronger the distribution uncertainty, the more robust the optimization results, and the more quantifiable the degree of skepticism decision-makers have about the credibility of historical data.

[0060] Finally, based on the output fuzzy set, the fluctuation range of the power generation of the new energy unit at the target sampling time is determined.

[0061] Assuming the target sampling time is time t, then the predicted value of the wind power generation equipment at time t is: Their uncertain actual outputs are respectively .in, This represents the future prediction error, varying within the fuzzy set F of the true distribution, centered on the predicted point value, i.e., the empirical distribution. The sample represents the error surrounding the ideal prediction.

[0062] S202, Based on the binary state variables of each device in the integrated energy virtual power plant and the fault constraint relationship between each device, a 0-1 polyhedral uncertain set is constructed.

[0063] In this application, each device in the integrated energy virtual power plant is modeled as a binary decision variable to represent its start-up, shutdown, and operating states. Linear inequality constraints are then established to define a feasible set of all possible values ​​for the binary variables. This feasible set describes both the failure scenarios of individual devices and the interconnected failure logic between devices. The uncertainty parameters of device failures can vary arbitrarily within a 0-1 polyhedron set.

[0064] Binary state variables in this application u j ∈{0,1}, where u j =1 indicates the device j Normal operation u j =0 indicates the device j Fault.

[0065] Table 1. Schematic representation of device state variables

[0066] The set of uncertain faulty equipment rows corresponding to each piece of equipment in the integrated energy virtual power plant in this application is as follows:

[0067] The meanings of each letter in the uncertainty set are as follows: U represents the feasible region of equipment failure uncertainty. u Indicates the device status; where A For the constraint matrix, b For vectors, the meanings of the remaining letters are shown in Table 1.

[0068] The constraints involved in the fault uncertainty set are as follows: ① Overall reliability level constraints: That is, the total number of devices that fail simultaneously does not exceed α , α For risk aversion parameters; ② Electro-pneumatic coupling constraint: & That is, the output of the power-to-gas equipment depends on the storage of gas in the gas storage equipment and the carbon dioxide provided by the gas turbine; ③ Power supply support constraints: This means that at least one of the energy storage devices and the gas turbine must be online as a supporting power source; ④ Renewable energy linkage constraints: This means that simultaneous failure of wind and solar power must be prohibited; ⑤ Reverse discharge protection constraints: This means preventing energy storage devices and electric vehicles from reverse discharging on the same busbar; ⑥ Minimum power supply constraints for the operation of electro-gas conversion equipment: This means that at least one power source must be guaranteed to supply power to it.

[0069] In summary, the uncertainty described by the 0-1 polyhedron is as follows:

[0070] The 0-1 polyhedron set in this application can systematically describe the functional dependencies, physical interlocks, and operating rules among equipment within a virtual power plant, transforming complex engineering common sense into precise mathematical constraints and effectively describing equipment fault constraint relationships.

[0071] S203, construct a two-layer IGDT robust risk aversion model.

[0072] The robust risk aversion model in this application consists of two layers: the upper-layer bidding decision model focuses on the market bidding decisions of the integrated energy virtual power plant (hereinafter referred to as the virtual power plant), aiming to determine the bidding volume and quota trading volume in the day-ahead electricity market and carbon market to maximize the overall revenue of the virtual power plant; the lower-layer dispatching model focuses on internal operation optimization, using the IGDT (Information Gap Decision Theory) strategy to comprehensively consider the uncertainties of wind and solar power output and equipment failure, minimizing internal dispatching costs and uncertainty risks under constraints such as power balance, equipment physical operation, and carbon flow balance. The two-layer model iteratively interacts with the bidding capacity and minimum operating cost to form a closed-loop optimization.

[0073] Figure 3 This is a schematic diagram of the structure of a two-layer IGDT robust risk avoidance model provided in an embodiment of this application. Combined with... Figure 3As shown, the upper-level bidding decision model (i.e., the upper-level model) is a model constructed with the net benefit of virtual power plants participating in the carbon market as the optimization objective, the bidding volume and carbon quota trading volume as decision variables, and the constraints of power market bidding capacity, bidding volume ramp-up, and carbon market trading as constraints. The lower-level dispatch model (i.e., the lower-level model) is a model constructed with the minimum operating cost of virtual power plants as the optimization objective, using the IGDT robust model, the uncertainty of wind and solar uncertainties, the output of each device, and the energy stored in energy storage as decision variables, and the constraints of power balance, physical operation of equipment, and carbon-sulfur balance as constraints.

[0074] In one alternative implementation, the upper-level bidding decision model is as follows:

[0075]

[0076]

[0077]

[0078] in, It is the target value of the upper-level bidding decision model; It is the predicted electricity price at time t. yes t Predicting carbon prices at any given time P e,t It represents the day-ahead electricity market bid volume at time t. P c,t It represents the day-ahead carbon market bid volume at time t. F 2 It is the target value of the lower-level scheduling model; j This represents the equipment index in the aforementioned integrated energy virtual power plant. For equipment j Rated power, J cons This represents a collection of discharge devices. J gen It represents a collection of charging devices; The external power ramp-up limit per minute for the aforementioned integrated energy virtual power plant; This is the maximum purchase quota allowed per time period. It is the upper limit of the surplus quota that can be sold in a single time period, P allThe first formula represents the total transaction volume within the compliance period; the second formula in the upper-level bidding decision model represents the electricity market bidding capacity constraint, and the electricity market bidding volume depends on the rated power of the virtual power plant's internal equipment; the third formula in the upper-level bidding decision model represents the bidding electricity ramp-up constraint, that is, the limit on the rate of change of the overall external power output of the virtual power plant; the fourth formula in the upper-level bidding decision model represents the carbon market trading constraint, including the virtual power plant's own carbon quota limit and compliance period rules.

[0079] In one alternative implementation, the lower-level scheduling model is as follows:

[0080]

[0081]

[0082]

[0083] in, F 2 It is the target value of the lower-level scheduling model; U represents the feasible region of equipment failure uncertainty; y This represents the set of decision variables corresponding to the lower-level scheduling model. It is the scheduling feasible region, determined by the bid quantity. x =( P e,t , P c,t ), Equipment status u Wind and light error Common constraints J This represents the collection of all equipment in the integrated energy virtual power plant. j This represents the equipment index in the aforementioned integrated energy virtual power plant. C op,j This represents the unit operating cost of device j. P j,t It is the output of device j at time t. C qd,j This represents the single startup cost of device j. v j,t This indicates the startup state of device j at time t; P grid,t This represents the real-time power deviation adjustment between the power grid and the target power at time t. in, S es,t It is the stored energy at time t. S es,t-1 yes t The stored energy at time -1; P ch,es,tIt is the energy storage charging power at time t. P dis,es,t yes t The energy storage and discharge power at any given moment; η ch,es It refers to energy storage charging efficiency. η dis,es It refers to the energy storage and discharge efficiency; E gt,t It is the carbon emissions of the gas turbine at time t. E p2g,t It represents the amount of carbon absorbed by the electro-gas converter at time t. E grid,t It represents the reduction in carbon emissions at time t compared to purchasing electricity from the grid; Among them, the second formula in the lower-level scheduling model represents the power balance constraint, that is, the power generation and consumption of the virtual power plant must be balanced with the power exchanged in the bidding; the third formula in the lower-level scheduling model represents the energy storage charging and discharging constraint, that is, to ensure the balance of charging and discharging energy; the fourth formula in the lower-level scheduling model represents the carbon flow balance constraint, that is, to ensure that the total amount of carbon market trading and the amount of carbon inside the virtual power plant are balanced.

[0084] S204, Obtain the vector autoregressive model of the carbon electricity market.

[0085] The vector autoregression model for the electricity carbon market in this application is used to characterize the dynamic coupling relationship and linkage trend among electricity carbon market prices.

[0086] In one alternative implementation, the vector autoregressive model for the electricity carbon market is as follows:

[0087] in, , , c e and c c These are the intercept terms for electricity price and carbon price, respectively. p This indicates the lag order of the model. Indicates lag k The autoregressive coefficient matrix of the period, The coefficient matrix of exogenous variables, x t For exogenous variable vectors, , R xny,t Let t be the percentage of renewable energy generation. P load,t Let t be the virtual power plant load. D t It's a policy event. This is the white noise error term.

[0088] The vector autoregression model for the electricity carbon market in this application takes into account new energy power generation, virtual power plant load, policies, and related errors, effectively describing the mutual influence between electricity carbon markets, and laying a stable and effective foundation for the formulation of collaborative bidding strategies for virtual power plants in the dual markets.

[0089] S205, based on the fluctuation range of the power generation of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set, and the electric carbon market vector autoregressive model, the two-layer IGDT robust risk avoidance model is solved to generate a day-ahead bidding strategy.

[0090] S205 specifically includes the following steps: A1, based on preset risk avoidance parameters, adjust the radius of the fluctuation range of the power generation and the fault tolerance upper limit of the 0-1 polyhedral uncertainty set respectively to define the uncertainty boundary.

[0091] The IGDT framework quantifies the size of the uncertainty set using the risk aversion parameter α, where the wind and solar uncertainty set and the equipment failure set can be represented as:

[0092] in, This represents the set of uncertainties in wind and solar power output (i.e., the set of uncertainties in wind and solar power output mentioned above). Let represent the wind and solar prediction error vector, and p represent the order of the p-norm. As the baseline fluctuation radius, This represents a risk aversion parameter. Let represent the set of uncertainties related to equipment failures, u represent the state vector of a single device (i.e., a single device state vector), M represent the total number of devices, and A represent the constraint matrix. This represents the state vector of the j-th device. represents the maximum number of allowed equipment failures; b is a vector.

[0093] A2. Based on the strong duality theorem, the minimum operating cost objective function of the lower-level scheduling model and the upper-level bidding decision model are jointly dualized under the constraint of the uncertainty boundary to obtain a single-layer robust optimization model.

[0094] The upper-level bidding decision model (i.e., the upper-level model) in the two-layer IGDT robust risk aversion model takes the virtual power plant as the leader and maximizes its trading revenue under the worst-case uncertainty scenario in the electricity-carbon market; the lower-level model, as the follower, minimizes its internal operating costs under the worst-case scenario. This model essentially constitutes a Stackelberg game, and its core challenge lies in handling the min-max structure of the lower layer, that is, finding the worst-case uncertainty scenario that maximizes the operating costs of the virtual power plant.

[0095] In this application, the inner-level min problem is transformed into an equivalent max form based on the strong duality theorem, thereby reconstructing the lower-level objective into a convex optimization problem.

[0096]

[0097] in, It is the feasible region of the variable. x =( P e,t , P c,t () represents the model of the upper-level model, device status. u Wind and light error Common constraints It is the scheduling feasible region, and y represents the lower-level scheduling variable.

[0098] Its dual objective can be expressed as:

[0099] in, μ Constraints for the dual variables and their corresponding original lower-level objectives. This is the right-hand item of the original lower-level target. B This is the coefficient matrix of the original upper-level objective constraint. c The cost coefficient is the original upper-level objective constraint.

[0100] Therefore, the lower-level objective of the min-max approach is transformed into a joint maximization objective:

[0101] Therefore, the lower-level objective of the min-max approach is transformed into a joint maximization objective:

[0102] A3 employs an adaptive chaotic quantum particle swarm optimization algorithm, using the output of the electric carbon market vector autoregressive model as the price signal, to optimize the single-layer robust optimization model under the constraint of the uncertainty boundary, thereby generating the day-ahead bidding strategy.

[0103] Among them, the Adaptive Chaotic Quantum Particle Swarm Optimization (AQSO) is an intelligent optimization algorithm that integrates chaotic mapping, adaptive parameter control, and dual-population evolution mechanism. It is used to efficiently solve robust bidding strategies in complex non-convex solution spaces.

[0104] Because the aforementioned single-layer robust optimization model exhibits strong nonlinearity and multiple extrema under uncertain boundary constraints, the traditional particle swarm optimization algorithm is prone to getting trapped in local optima or premature convergence. Therefore, this application makes targeted improvements to the algorithm during execution, mainly including the following two points: First, an adaptive adjustment mechanism for the contraction and expansion coefficients based on population state.

[0105] The algorithm calculates the variance of the fitness of all particles in the current population at the current iteration round and defines this variance as the fitness variance. It also counts the number of iterations in which the global optimal position remains unchanged and defines this number as the evolutionary stagnation degree. The fitness variance measures the degree of population aggregation, while the evolutionary stagnation degree measures the convergence state of the algorithm.

[0106] When the fitness variance is lower than a preset first threshold or the evolutionary stagnation exceeds a preset second threshold, the current population is determined to be trapped in a local optimum, and the contraction-expansion coefficient is automatically increased to force particles to perform a wide-range search; when the fitness variance is higher than the first threshold and the evolutionary stagnation is lower than the second threshold, the current population is determined to be in a normal search state, and the contraction-expansion coefficient is maintained or decreased to achieve precise search.

[0107] Second, a quantum position update mechanism based on dual-population co-evolution.

[0108] Based on a preset ratio, the current population in the current iteration round is divided into an exploration group and a development group. The exploration group is controlled to perform quantum position updates with a contraction and expansion coefficient greater than 1, so as to make large-scale jumps in the solution space and maintain population diversity. The development group is controlled to perform quantum position updates with a contraction and expansion coefficient less than 1, so as to perform fine search in the neighborhood of the current optimal solution and accelerate algorithm convergence. Every preset exchange cycle, the elite particles with the best fitness in the exploration group are migrated to the development group, and the global optimal position of the development group is fed back to the exploration group to achieve complementary evolution of the two populations.

[0109] In one alternative implementation, an adaptive chaotic quantum particle swarm optimization algorithm is used, with the output of the electric carbon market vector autoregressive model as the price signal. Under the constraint of the uncertainty boundary, the single-layer robust optimization model is optimized to generate the day-ahead bidding strategy, specifically including the following steps: The first step is to unify the continuous and discrete decision variables of the virtual power plant into particle position vectors through hybrid encoding and chaotic mapping initialization, and generate a uniformly distributed initial population in the solution space.

[0110] For example, with a maximum iteration count of 500 and a risk avoidance parameter of 0.05, the variables are encoded as shown in Table 2 below, resulting in the particle position vector. .

[0111] Table 2 Coding Table

[0112] in, This represents the decision variable vector (or solution vector), where the subscript i usually refers to the i-th particle (in the context of particle swarm optimization) or the i-th sample. This indicates the amount of electricity tendered in the electricity market recently; This indicates the current day's carbon market trading volume; Represents the wind and solar forecasting error vector; The set of real numbers is represented by T and M, where both T and M represent dimensions. The meanings of the other letters are described in the previous embodiments and will not be repeated here.

[0113] Furthermore, we construct the quantum particle swarm optimization equation, where the particle positions follow a quantum potential well distribution, and the position update formula is:

[0114] in, i Represents particles, N This represents the particle size, taken as 100. j Representing dimension, The random weight coefficients follow the order [0,1]. express t The optimal position of an individual particle at any given moment. Indicates the globally optimal position. Indicates the coefficient of contraction and expansion. u This represents a random number belonging to the range [0,1].

[0115] The second step is to initialize the chaotic population, generating the initial population and particle initialization positions.

[0116]

[0117] in, c k The first chaotic sequence represents the... k One value; μ This indicates a control parameter, set to 4; UB j Indicates the first j The upper limit of a dimensional variable; LB j Indicates the first j The lower bound of a dimension variable; c i,j This represents a chaotic number, which is a random number between [0,1].

[0118] The third step is to initialize the population fitness assessment.

[0119] Bidding strategy variables, wind and solar uncertainty variables, and equipment failure variables are extracted from particle positions. It is then verified whether the uncertainty variables are within the uncertainty set, and finally, the underlying operating cost problem is solved, taking the negative value. F 1As a fitness metric, if the fitness is better than the current global best position, then the global best position is updated.

[0120] The fourth step is to calculate the average optimal position and begin the iteration.

[0121]

[0122] in, mest j express j The average optimal position of the dimension; the meanings of the other letters are explained in the description of the previous embodiments, and will not be repeated here.

[0123] Step 5: Adaptively update the shrinkage and expansion coefficients.

[0124]

[0125] in, and Represent The minimum and maximum values ​​are 0.5 and 1, respectively. This represents the attenuation adjustment coefficient, which is set to 2. Indicates individual fitness; Indicates average fitness; Indicates the global optimal fitness; This represents the historical global optimal fitness; L indicates that the window length is 10. Fitness variance represents population diversity. The degree of evolutionary stagnation indicates the rate of convergence.

[0126] Step 6: Two-population partitioning and parallel update.

[0127] Two populations were constructed as shown in Table 3. The quantum position of the exploration population was updated according to the position update formula, with a shrinkage coefficient of 1.2 for the exploration population. The shrinkage coefficient of the development group is 0.6. .

[0128] Table 3 Population Division

[0129] Step 7: Boundary handling.

[0130] Perform a sigmoid mapping on the device state variables, if u j Values ​​≥0.5 are set to 1, otherwise 0. The wind and light error is assessed by determining if it lies within the Wasserstein distance; if not, it is projected onto its boundary. Other variables are then... deal with.

[0131] Step 8: Fitness assessment of the new population.

[0132] Update the individual optimal and global optimal values ​​according to the fitness rule.

[0133] Step 9: Chaotic perturbation check.

[0134] When the global optimal solution remains unchanged for 15 consecutive generations, a chaotic perturbation is triggered. At this point, the current best particle is retained, and other particles undergo chaotic mutation:

[0135]

[0136] In the formula, These represent the positions before and after the disturbance, respectively. This represents the disturbance intensity coefficient, which belongs to [0,1]. Take 0.5.

[0137] Step 10: Information exchange between the two populations.

[0138] Information is exchanged every 5 generations. The best 10% of particles are selected from the exploration group to replace the worst 10% of particles in the development group. The global optimal solution of the development group is passed to the exploration group as additional guiding direction to ensure that the excellent strategies in the exploration group can guide the development group. The fine optimization results of the development group are fed back to the exploration group.

[0139] This application employs an adaptive chaotic quantum particle swarm optimization algorithm that follows an initialization, iterative evolution, adaptive adjustment, and convergence output process. First, initial values ​​for the number of iterations and risk avoidance parameters are set. The bidding strategy, wind and solar errors, and equipment status of the virtual power plant are hybridized and encoded, mapping continuous and discrete decision variables to particle position vectors. The population is initialized through chaotic mapping to ensure uniform particle distribution in the solution space. Then, an iterative loop begins: in each generation, the average optimal position of the population is calculated to guide the distribution of quantum potential wells for the particles. Next, the contraction and expansion coefficients are dynamically updated based on the population's current fitness variance and evolutionary stagnation, enabling automatic switching from global exploration to local development. The population is then divided into an exploration group and a development group, with parallel quantum position updates. The exploration group uses a larger contraction coefficient to expand the search range, while the development group uses a smaller contraction coefficient to refine the search near the optimal solution. The updated particles undergo boundary processing, mapping the equipment status to 0 / 1 integers using the Sigmoid function and projecting the wind and solar errors into a Wasserstein sphere to ensure all particles satisfy the uncertainty set constraints. The particle fitness is then reassessed, and the individual optimal and global optimal positions are updated based on the lower-level scheduling cost calculation results. If the global optimal position remains unchanged for several consecutive generations, a chaotic perturbation is triggered, applying chaotic mutation to non-optimal particles to force the population out of the stagnation region. Simultaneously, bi-population information exchange is performed every fixed number of generations, migrating elite particles from the exploration group to the development group and feeding back the global optimal position from the development group to the exploration group, achieving co-evolution of the two populations. Finally, this process is repeated until the maximum number of iterations is reached, ultimately outputting the optimal strategy, which is the day-ahead strategy for collaborative bidding in the electric carbon market.

[0140] In one alternative implementation, after obtaining the day-ahead strategy for coordinated bidding in the electricity carbon market, the information in the day-ahead strategy can be displayed on the interface. The specific content that can be displayed is shown in Table 4. The advantage of this approach is that it transforms complex optimization results into an understandable and operable visual interface, supporting efficient decision-making by operators.

[0141] Compared with existing technologies, the method for generating day-ahead bidding strategies for electro-carbon synergy provided in this application has at least the following beneficial effects: By integrating the Wasserstein-DRO model with 0-1 polyhedron sets, the uncertainties of wind and solar power output and equipment failures are characterized respectively. This solves the problems of large bias in distribution assumptions and lack of associated failure logic in traditional methods, and significantly enhances the robustness of the model.

[0142] Based on the IGDT framework, risk aversion parameters are introduced to construct a complete spectrum from risk neutral to extreme risk aversion, giving decision-makers the ability to dynamically adjust the weights of returns and risks.

[0143] A vector autoregression model for the electricity carbon market was established to accurately quantify the dynamic coupling relationship of electricity carbon prices, thereby achieving coordinated allocation of resources across markets and maximizing benefits.

[0144] An adaptive chaotic quantum particle swarm optimization algorithm is adopted, which utilizes chaotic ergodicity and quantum behavior mechanism to effectively avoid premature convergence of the algorithm and ensure the solution quality under complex constraints.

[0145] The generated bidding strategy is immune to worst-case uncertainty scenarios. Even under extreme conditions such as a sharp drop in wind and solar power or multiple equipment failures, it can still maintain an acceptable level of returns, significantly improving the market resilience of the integrated energy virtual power plant.

[0146] Based on the same concept, this application also provides a day-ahead strategy generation device for collaborative bidding in the electric carbon market. Figure 4 This is a schematic diagram of a day-ahead strategy generation device for collaborative bidding in the electric carbon market, provided as an embodiment of this application. (Combined with...) Figure 4 As shown, the current strategy generation device 400 includes: The wind and solar uncertainty determination module 401 is used to determine the fluctuation range of the power generation of the new energy units in the integrated energy virtual power plant based on the deviation between the historical predicted power generation sequence and the corresponding historical actual power generation sequence of the new energy units, and the predicted power generation at the target sampling time; the new energy units include wind power generation equipment and photovoltaic power generation equipment. The polyhedral uncertainty set determination module 402 is used to construct a 0-1 polyhedral uncertainty set based on the binary state variables of each device in the integrated energy virtual power plant and the fault constraint relationship between each device; The risk aversion model construction module 403 is used to construct a two-layer IGDT robust risk aversion model. The model includes an upper-layer bidding decision model with the goal of maximizing the net revenue of the integrated energy virtual power plant in the electricity carbon market, and a lower-layer scheduling model with the goal of minimizing the operating cost of the integrated energy virtual power plant under the worst-case uncertainty scenario. The module 404 for determining the relationship in the electricity carbon market is used to obtain a vector autoregressive model of the electricity carbon market; the vector autoregressive model of the electricity carbon market is used to characterize the dynamic coupling relationship and linkage trend among electricity carbon market prices. The day-ahead bidding strategy generation module 405 is used to solve the two-layer IGDT robust risk avoidance model based on the fluctuation range of the power generation of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set and the electric carbon market vector autoregression model, and generate the day-ahead bidding strategy.

[0147] In one alternative implementation, the wind and solar uncertainty determination module 401 includes: An error sample sequence determination unit is used to subtract the historical actual power generation sequence and the predicted power generation sequence element by element to obtain an error sample sequence. The output fuzzy set determination unit is used to construct an empirical distribution function based on the error sample sequence, and construct the output fuzzy set corresponding to the predicted power generation of the new energy unit at the target sampling time with the empirical distribution function as the center and the Wasserstein distance as the radius. The power generation range determination unit is used to determine the fluctuation range of the power generation of the new energy unit at the target sampling time based on the output fuzzy set.

[0148] In one alternative implementation, the day-ahead bidding strategy generation module 405 includes: The uncertainty boundary determination unit is used to define the uncertainty boundary by adjusting the radius of the fluctuation range of the power generation and the fault tolerance upper limit of the 0-1 polyhedral uncertainty set based on preset risk avoidance parameters. A single-layer robust optimization model determination unit is used to perform joint dual transformation between the minimum operating cost objective function of the lower-layer scheduling model and the upper-layer bidding decision model under the constraint of the uncertainty boundary, based on the strong duality theorem, to obtain a single-layer robust optimization model. The day-ahead bidding strategy determination unit is used to optimize the single-layer robust optimization model under the constraints of the uncertainty boundary by using an adaptive chaotic quantum particle swarm optimization algorithm with the output of the electric carbon market vector autoregressive model as the price signal, and to generate the day-ahead bidding strategy.

[0149] In one alternative implementation, the day-ahead bidding strategy determination unit includes: The fitness variance determination subunit is used to calculate the variance of the fitness of all particles in the current population under the current iteration round, and defines the variance as the fitness variance, which is used to measure the degree of aggregation of the population. The evolutionary stagnation determination subunit is used to count the number of iterations in which the global optimal position remains unchanged, and the number of iterations is defined as the evolutionary stagnation degree; the evolutionary stagnation degree is used to measure the convergence state of the algorithm; The first search determination subunit is used to determine that the current population is trapped in a local optimum when the fitness variance is lower than a preset first threshold or the evolutionary stagnation exceeds a preset second threshold, and automatically increase the contraction and expansion coefficient to force the particles to perform a large-scale search. The second search determination subunit is used to determine that the current population is in a normal search state when the fitness variance is higher than the first threshold and the evolutionary stagnation is lower than the second threshold, and to maintain or reduce the contraction and expansion coefficient to achieve precise search.

[0150] In one alternative implementation, the day-ahead bidding strategy determination unit includes: The population determination subunit is used to divide the current population in the current iteration round into an exploration group and a development group based on a preset ratio. The first position update subunit is used to control the exploration group to perform quantum position updates with a contraction and expansion coefficient greater than 1, so as to make large-scale jumps in the solution space and maintain population diversity. The second position update subunit is used to control the development group to perform quantum position updates with a contraction and expansion coefficient of less than 1, so as to perform a fine search in the neighborhood of the current optimal solution and accelerate the convergence of the algorithm. The species exploration subunit is used to migrate the elite particles with the best fitness in the exploration group to the development group at preset exchange cycles, and to feed back the global optimal position of the development group to the exploration group, so as to achieve complementary evolution between the two species.

[0151] In one alternative implementation, the day-ahead policy generation device 400 also includes a policy output display module. Combined with... Figure 5 As shown, the strategy output display module in this application specifically includes: a bidding strategy visualization submodule, an equipment scheduling display submodule, a risk indicator display submodule, a human-computer interaction submodule, a strategy export submodule, and a historical query submodule.

[0152] The bidding strategy visualization submodule is used to visually display the combination of volume and price bids and the distribution of expected returns at different time periods in the form of line charts, bar charts, or heat maps; the equipment scheduling display submodule is used to simultaneously present the start-up and shutdown status, output plan, and operating conditions of generator sets, energy storage equipment, and flexible loads within the previous 24 hours; the risk indicator display submodule focuses on graphical early warnings of key data such as volatility, deviation penalty, and confidence interval; the human-computer interaction submodule is responsible for receiving user instructions for fine-tuning strategy parameters and refreshing the display content in real time; the strategy export submodule supports converting the finalized strategy scheme into a standard format file for the scheduling system to execute; and the historical query submodule is used to compare and analyze the execution effect of past strategies and conduct retrospective reviews, thereby forming a closed-loop feedback for strategy optimization.

[0153] Table 4 is a detailed correspondence table of specific display indicators, content, and presentation formats for each sub-module of the strategy output display module. Figure 5 The specific implementation details of the module architecture shown are explained below.

[0154] Table 4 Strategy Output Display Table

[0155] In one alternative implementation, the day-ahead policy generation device 400 also includes a main control storage module. Combined with... Figure 6 As shown, the main control storage module specifically includes: a main control scheduling unit, a data storage unit, a clock synchronization unit, a log management unit, and a security and access control unit.

[0156] Among them, the main control and scheduling unit, as the core brain of the module, is responsible for coordinating the task execution and resource allocation of the lower-level units, ensuring the timing and logical consistency of storage operations; the data storage unit is responsible for the persistent storage of massive amounts of raw data, intermediate calculation results, and final strategy solutions during device operation, providing a solid data foundation for the system; the clock synchronization unit ensures that the time base of all nodes and records within the device is highly consistent, meeting the stringent requirements of the power system for time-series data; the log management unit comprehensively records the device's operating status, operation traces, and abnormal events, providing a basis for maintenance, troubleshooting, and audit tracking; and the security and access management unit strictly protects the confidentiality and integrity of sensitive data and core strategies through identity authentication and access control mechanisms.

[0157] The five units work together, each corresponding to core functions such as data storage, caching services, and operation log recording, to ensure that the entire device is coordinated, reliable, stable, and secure.

[0158] Based on the aforementioned embodiments of the day-ahead strategy generation method and apparatus for collaborative bidding in the electric carbon market, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements some or all of the steps in the configuration migration method mentioned above.

[0159] Based on the day-ahead strategy generation method and apparatus for collaborative bidding in the electric carbon market provided in the foregoing embodiments, this application also provides an electronic device, including: A memory on which computer programs are stored; A processor is configured to execute the computer program in the memory to implement some or all of the steps in the configuration migration method provided in the foregoing embodiments.

[0160] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separate. The components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment solution according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0161] The above description is merely one specific embodiment of this application, but the scope of protection of this application 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 this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating day-ahead strategies for collaborative bidding in the electricity carbon market, characterized in that, The method is applied to an integrated energy virtual power plant, and the method includes: For the new energy units of the integrated energy virtual power plant, based on the deviation between the historical predicted power generation sequence and the corresponding historical actual power generation sequence of the new energy units, and the predicted power generation at the target sampling time, the fluctuation range of the power generation of the new energy units at the target sampling time is determined; the new energy units include wind power generation equipment and photovoltaic power generation equipment. A 0-1 polyhedral uncertainty set is constructed based on the binary state variables of each device in the integrated energy virtual power plant and the fault constraint relationship between each device. A two-layer IGDT robust risk avoidance model is constructed, which includes an upper-layer bidding decision model aimed at maximizing the net revenue of the integrated energy virtual power plant in the carbon electricity market, and a lower-layer scheduling model aimed at minimizing the operating cost of the integrated energy virtual power plant under the worst-case uncertainty scenario. A vector autoregression model for the electricity carbon market is obtained; this model is used to characterize the dynamic coupling relationship and linkage trend among electricity carbon market prices. Based on the fluctuation range of the power generation of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set, and the vector autoregressive model of the electricity carbon market, the two-layer IGDT robust risk avoidance model is solved to generate a day-ahead bidding strategy.

2. The method according to claim 1, characterized in that, For the new energy generating units in the integrated energy virtual power plant, based on the deviation between the historical predicted power generation sequence and the corresponding historical actual power generation sequence of the new energy generating units, and the predicted power generation at the target sampling time, the fluctuation range of the power generation of the new energy generating units at the target sampling time is determined, including: The error sample sequence is obtained by subtracting the historical actual power generation sequence from the predicted power generation sequence element by element. Based on the error sample sequence, an empirical distribution function is constructed, and with the empirical distribution function as the center and the Wasserstein distance as the radius, the output fuzzy set corresponding to the predicted power generation of the new energy unit at the target sampling time is constructed. Based on the output fuzzy set, the fluctuation range of the power generation of the new energy unit at the target sampling time is determined.

3. The method according to claim 1, characterized in that, The upper-level bidding decision model is as follows: in, It is the target value of the upper-level bidding decision model; It is the predicted electricity price at time t. yes t Predicting carbon prices at any given time P e,t It represents the day-ahead electricity market bid volume at time t. P c,t It represents the day-ahead carbon market bid volume at time t. F 2 It is the target value of the lower-level scheduling model; j This represents the equipment index in the aforementioned integrated energy virtual power plant. For equipment j Rated power, J cons This represents a collection of discharge devices. J gen It represents a collection of charging devices; The external power ramp-up limit per minute for the aforementioned integrated energy virtual power plant; This is the maximum purchase quota allowed per time period. It is the upper limit of the surplus quota that can be sold in a single time period, P all This represents the total transaction volume within the contract fulfillment period.

4. The method according to claim 3, characterized in that, The lower-level scheduling model is as follows: in, F 2 It is the target value of the lower-level scheduling model; U represents the feasible region of equipment failure uncertainty; y This represents the set of decision variables corresponding to the lower-level scheduling model. It is the scheduling feasible region, determined by the bid quantity. x =( P e,t , P c,t ), Equipment status u Wind and light error Common constraints J This represents the collection of all equipment in the integrated energy virtual power plant. j This represents the equipment index in the aforementioned integrated energy virtual power plant. C op,j This represents the unit operating cost of device j. P j,t It is the output of device j at time t. C qd,j This represents the single startup cost of device j. v j,t This indicates the startup state of device j at time t. ；P grid,t This represents the real-time power deviation adjustment between the power grid and the target power at time t. in, S es,t It is the stored energy at time t. S es,t-1 yes t The stored energy at time -1; P ch,es,t It is the energy storage charging power at time t. P dis,es,t yes t The energy storage and discharge power at any given moment; η ch,es It refers to energy storage charging efficiency. η dis,es It refers to the energy storage and discharge efficiency; E gt,t It is the carbon emissions of the gas turbine at time t. E p2g,t It represents the amount of carbon absorbed by the electro-gas converter at time t. E grid,t It represents the reduction in carbon emissions at time t compared to purchasing electricity from the grid.

5. The method according to claim 1, characterized in that, The vector autoregressive model for the electricity carbon market is as follows: in, , , c e and c c These are the intercept terms for electricity price and carbon price, respectively. p This indicates the lag order of the model. Indicates lag k The autoregressive coefficient matrix of the period, The coefficient matrix of exogenous variables, x t For exogenous variable vectors, , R xny,t for t The proportion of renewable energy power generation at any given time P load,t Let t be the virtual power plant load. D t It's a policy event. This is the white noise error term.

6. The method according to claim 1, characterized in that, The method involves solving the two-layer IGDT robust risk aversion model based on the power generation fluctuation range of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set, and the vector autoregressive model of the electricity carbon market, to generate a day-ahead bidding strategy, including: Based on preset risk avoidance parameters, the radius of the fluctuation range of the power generation and the fault tolerance upper limit of the 0-1 polyhedral uncertainty set are adjusted respectively to define the uncertainty boundary. Based on the strong duality theorem, the minimum operating cost objective function of the lower-level scheduling model and the upper-level bidding decision model are jointly dualized under the constraint of the uncertainty boundary to obtain a single-layer robust optimization model. An adaptive chaotic quantum particle swarm optimization algorithm is used, with the output of the electric carbon market vector autoregressive model as the price signal, to optimize the single-layer robust optimization model under the constraint of the uncertainty boundary, thereby generating the day-ahead bidding strategy.

7. The method according to claim 6, characterized in that, The process of using an adaptive chaotic quantum particle swarm optimization algorithm, with the output of the electric carbon market vector autoregressive model as the price signal, to optimize the single-layer robust optimization model under the constraints of the uncertainty boundary, and generating the day-ahead bidding strategy includes: Calculate the variance of the fitness of all particles in the current population at the current iteration round, and define the variance as the fitness variance, which is used to measure the degree of aggregation of the population; The number of iterations in which the global optimal position remains unchanged is counted, and this number of iterations is defined as the evolutionary stagnation degree; the evolutionary stagnation degree is used to measure the convergence state of the algorithm. When the fitness variance is lower than a preset first threshold or the evolutionary stagnation exceeds a preset second threshold, it is determined that the current population is trapped in a local optimum, and the contraction and expansion coefficient is automatically increased to force the particles to perform a large-scale search. When the fitness variance is higher than the first threshold and the evolutionary stagnation is lower than the second threshold, the current population is determined to be in a normal search state, and the contraction-expansion coefficient is maintained or reduced to achieve precise search.

8. The method according to claim 7, characterized in that, The process of using an adaptive chaotic quantum particle swarm optimization algorithm, with the output of the electric carbon market vector autoregressive model as the price signal, to optimize the single-layer robust optimization model under the constraints of the uncertainty boundary, and generating the day-ahead bidding strategy includes: Based on a preset ratio, the current population in the current iteration round is divided into an exploration group and a development group; The exploration swarm is controlled to perform quantum position updates with a contraction-expansion coefficient greater than 1, so as to make large-scale jumps in the solution space and maintain population diversity; The development group is controlled to perform quantum position updates with a contraction and expansion coefficient of less than 1 in order to perform a fine search in the neighborhood of the current optimal solution and accelerate the convergence of the algorithm. Every preset exchange cycle, the elite particles with the best fitness in the exploration group are migrated to the development group, and the global optimal position of the development group is fed back to the exploration group to achieve complementary evolution between the two groups.

9. A day-ahead strategy generation device for collaborative bidding in the electricity carbon market, applied to an integrated energy virtual power plant, characterized in that, The device includes: The wind and solar uncertainty determination module is used to determine the fluctuation range of the power generation of the new energy units in the integrated energy virtual power plant, based on the deviation between the historical predicted power generation sequence and the corresponding historical actual power generation sequence of the new energy units, and the predicted power generation at the target sampling time; the new energy units include wind power generation equipment and photovoltaic power generation equipment. The polyhedral uncertainty set determination module is used to construct a 0-1 polyhedral uncertainty set based on the binary state variables of each device in the integrated energy virtual power plant and the fault constraint relationship between the devices. The risk aversion model construction module is used to construct a two-layer IGDT robust risk aversion model. The model includes an upper-layer bidding decision model with the goal of maximizing the net revenue of the integrated energy virtual power plant in the electricity carbon market, and a lower-layer scheduling model with the goal of minimizing the operating cost of the integrated energy virtual power plant under the worst-case uncertainty scenario. The module for determining the relationship in the electricity carbon market is used to obtain a vector autoregressive model of the electricity carbon market; the vector autoregressive model of the electricity carbon market is used to characterize the dynamic coupling relationship and linkage trend among electricity carbon market prices. The day-ahead bidding strategy generation module is used to solve the two-layer IGDT robust risk avoidance model based on the fluctuation range of the power generation of the new energy unit at multiple target sampling times, the 0-1 polyhedral uncertainty set, and the electric carbon market vector autoregressive model, and generate the day-ahead bidding strategy.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the method described in any one of claims 1-8.

Images