Collaborative optimization scheduling method and system based on low-carbon demand response

Abstract

The invention relates to a collaborative optimization scheduling method and system based on low-carbon demand response, and the method comprises the following steps: constructing a source-network-load-carbon storage emission flow metering model, and obtaining the carbon emission data of each part during the operation of a power system; obtaining a dynamic carbon emission factor according to the carbon emission data of each part during the operation of the power system, issuing the dynamic carbon emission factor to a user side, guiding a user to change the power utilization habit, and achieving the dynamic transfer of the power utilization load; a stepped carbon transaction mechanism is formulated according to the power utilization load transfer condition of the user, and the user is stimulated to optimize the carbon emission cost through carbon quota distribution and carbon price grading; and constructing a multi-objective optimization problem with minimum carbon emission intensity and minimum total cost of the power system, and solving the multi-objective optimization problem through a near-end strategy optimization PPO algorithm to obtain an optimal scheduling strategy.

Description

Technical Field

[0001] The present application relates to the technical field of low-carbon power operation, and more specifically, to a collaborative optimization scheduling method and system based on low-carbon demand response. Background Art

[0002] Global climate change is severe, and countries are promoting carbon emission reduction policies. The power industry is facing the pressure of low-carbon transformation. In the integrated intelligent power system of source-grid-load-storage, it is crucial to optimize scheduling to achieve carbon emission control. Traditional power systems are mainly fossil fuel-based power generation with high carbon emissions. Although new energy power generation has developed rapidly, it is intermittent and volatile, which poses challenges to system stability and scheduling. Demand response is an important strategy to achieve low-carbon and intelligent power systems. It can guide users to adjust electricity consumption time, balance supply and demand, increase the proportion of renewable energy consumption, reduce fossil fuel dependence, and reduce carbon emissions. With the development of the carbon trading market, its combination with demand response will help form a multi-level low-carbon management framework, optimize the total carbon emissions, and enhance user initiative. Reinforcement learning is gradually applied in the optimization and scheduling of power systems. Its adaptive scheduling method is more flexible and real-time responsive than the traditional static optimization model. In demand response, it can guide users to respond dynamically according to carbon emission signals, improve scheduling efficiency, and achieve dual goals. The low-carbon transformation of modern power systems requires intelligent scheduling solutions that integrate multiple factors to meet low-carbon scheduling needs and promote sustainable development.

[0003] Prior art, such as the Chinese patent application with publication number "CN116362464A", discloses an energy system optimization method based on the carbon peak target, which includes the following steps: Step S1: inputting power system operation data from the power grid database to form a basic database; Step S2: constructing a carbon emission measurement model; Step S3: constructing a carbon emission peak, power supply and demand balance and operating performance optimization model; Step S4: performing multi-objective solution to design an optimization plan. The energy system optimization method based on the carbon peak target of this invention accurately calculates the carbon emission level based on the power generation and consumption structure and development goals of the power grid, effectively uses the power generation and consumption plan to digitally calculate the current power consumption carbon emissions and predict the carbon emission targets in the energy and power sector, fully combines the characteristics of power demand, reasonably optimizes the power supply layout, optimizes the power generation structure, and achieves mutual promotion at the same frequency.

[0004] The problem with the above-mentioned existing technology is that this method only considers the carbon emissions related to the source side, and does not involve the carbon emission data of the user's electricity consumption behavior, and the incentive system to guide users' electricity consumption; it is unable to integrate resources from various links to form a unified low-carbon optimization goal and achieve better overall optimization effects. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a collaborative optimization scheduling method and system based on low-carbon demand response.

[0006] The technical solution of the present invention is as follows:

[0007] The present invention proposes a collaborative optimization scheduling method based on low-carbon demand response, comprising the following steps:

[0008] Step S1, constructing a source-grid-load-storage carbon emission flow measurement model to obtain carbon emission data of each part when the power system is running;

[0009] Step S2, obtaining a dynamic carbon emission factor based on the carbon emission data of each part during the operation of the power system, publishing the dynamic carbon emission factor to the user side, guiding the user to change the electricity consumption habits, and realizing the dynamic transfer of electricity load;

[0010] Step S3: formulate a tiered carbon trading mechanism based on the user's electricity load transfer situation, and encourage users to optimize carbon emission costs through carbon quota allocation and graded carbon prices;

[0011] Step S4, construct a multi-objective optimization problem of minimizing the carbon emission intensity and the total cost of the power system, solve the multi-objective optimization problem through the proximal strategy optimization PPO algorithm, and obtain the optimal dispatching strategy.

[0012] As a preferred implementation, the source-grid-load-storage carbon emission flow metering model is constructed to obtain carbon emission data of each part when the power system is running, wherein:

[0013] The formula for calculating source-side carbon emissions is as follows:

[0014]

[0015] Where: w h is the unit power generation carbon emission intensity of thermal power unit h; a h 、b h 、c h They are the fuel consumption characteristic parameters of thermal power units; PG h represents the output power of thermal power unit h; ζh is the correction coefficient;

[0016] The calculation formula for storage-side carbon emissions is as follows:

[0017]

[0018] Where: e s (T) is the discharge carbon emission intensity when the energy storage charging ends and switches to the discharge mode at time T; Q0 and F0 are the electric energy stored in the device and the accumulated carbon emissions at the last switching moment when the energy storage device switches from discharge to charging; and are the accumulated electricity and carbon flow of the energy storage device during the charging process; η is the charging and discharging efficiency of the energy storage device;

[0019] The calculation formula for load-side carbon emissions is as follows:

[0020] R Gj→k =d jk P Lk I Gj ;

[0021] Where: R Gj→k The carbon emissions allocated from the energy injection node j of the generator set to the load node k; d jk P is the carbon emission allocation coefficient of the node j connected to the load node k; Lk is the load of node k; I Gj The carbon emission intensity of the energy injection node j for the generator set;

[0022] The calculation formula for grid-side carbon emissions is as follows:

[0023] R Gj→k-l =d jk P kl I Gj ;

[0024] Where: R Gi→k-l is the carbon emission corresponding to the node j connected to the generator set on the branch kl; P kl is the energy flow of branch kl.

[0025] As a preferred implementation, the dynamic carbon emission factor is obtained according to the carbon emission data of each part when the power system is running, and the dynamic carbon emission factor is obtained by the following formula:

[0026]

[0027] Where: e C，i (t) represents the dynamic carbon emission factor of the user connected to the grid node i at time t; S C，i , P i are the corresponding carbon emissions and total electric energy injected from the source side to the user side connected to node i through the power grid; S D，i is the total carbon emissions of the user-side power generation equipment connected to node i; is the total power generation of the kth power generation equipment at the user side of node i; S C，i , S D，i All are obtained through carbon emission data.

[0028] As a preferred implementation, the dynamic carbon emission factor is published to the user side to guide the user to change the electricity consumption habits and realize the dynamic transfer of electricity load; the dynamic transfer model of electricity load is as follows:

[0029]

[0030] Where: are the electric power that can be transferred by the i-th user in time period t before and after the transfer; are the electric power transferred in and out by user i in period t respectively; is the corresponding transfer status; are the transfer coefficient of the transferable load and the upper and lower limits of its variation respectively.

[0031] As a preferred implementation, the carbon quota allocation and graded carbon prices are used to encourage users to optimize carbon emission costs, wherein:

[0032] The carbon quota allocation formula is as follows:

[0033]

[0034] In the formula, E i is the total carbon emission quota allocated to user i; are the quotas for purchasing electricity from the main grid and for gas turbine use; e , χ g are the carbon emission quota corresponding to the unit electricity purchased from the main grid and the unit gas consumption of the gas turbine unit; P buy，i , P GT，i The amount of electricity purchased and gas consumed; is the quota of user i in the carbon trading market; is the actual total carbon emissions of user i; P i represents the total electricity consumption of user i;

[0035] The formula for the graded carbon price is as follows:

[0036]

[0037] In the formula, The amount of quota used for carbon trading; is the user's carbon emission transaction cost; λ is the carbon transaction price in the market; d is the length of the carbon emission interval; α is the increase in the carbon emission price in each step.

[0038] As a preferred implementation, the multi-objective optimization problem of minimizing the carbon emission intensity and total cost of the power system is constructed, with the minimum total operating cost of each user side as the objective function, where the total cost includes the electricity purchase cost, gas purchase cost, wind and solar abandonment cost and carbon trading cost. The specific objective function is as follows:

[0039]

[0040] Where: T is the scheduling period; S is the total number of users; are the total gas purchase cost and unit gas price of user i in time period t respectively; are the total electricity purchase cost and unit electricity price of user i in time period t respectively; PV , DG are the penalty costs for unit solar and wind curtailment respectively; P PV,i , P DG,i Respectively represent the predicted output and actual output of photovoltaic and wind power of user i; C cur,i (t) is the wind and solar power abandonment cost of user i in period t; is the carbon trading cost of user i in period t.

[0041] As a preferred implementation, the multi-objective optimization problem is solved by the proximal policy optimization PPO algorithm to obtain the optimal scheduling strategy. The proximal policy optimization PPO algorithm is guided by the expert experience of behavioral cloning and correction network; behavioral cloning pre-training is added to the proximal policy optimization PPO algorithm to accelerate the convergence of the model; the pointer network is used to enhance feature extraction and enhance the accuracy of strategy generation; the actions are screened and adjusted before each round of strategy generation through expert experience to prevent high-risk decisions such as load overload or flow overload.

[0042] On the other hand, the present invention also provides a collaborative optimization scheduling system based on low-carbon demand response, comprising:

[0043] The carbon emission data collection module builds a source-grid-load-storage carbon emission flow measurement model to obtain carbon emission data of each part during the operation of the power system;

[0044] Dynamic transfer of electricity load: dynamic carbon emission factors are obtained based on the carbon emission data of each part of the power system during operation, and the dynamic carbon emission factors are published to the user side to guide users to change their electricity consumption habits and realize dynamic transfer of electricity load;

[0045] The module for building a tiered carbon trading mechanism is to formulate a tiered carbon trading mechanism based on the user's electricity load transfer, and to encourage users to optimize carbon emission costs through carbon quota allocation and graded carbon prices;

[0046] The multi-objective optimization solution module constructs a multi-objective optimization problem of minimizing the carbon emission intensity and total cost of the power system, solves the multi-objective optimization problem through the proximal strategy optimization PPO algorithm, and obtains the optimal dispatching strategy.

[0047] On the other hand, the present invention further provides an electronic device having a computer program stored thereon, wherein when the computer program is executed by a processor, the collaborative optimization scheduling method based on low-carbon demand response as described in any embodiment of the present invention is implemented.

[0048] On the other hand, the present invention also provides a computer-readable medium for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement a collaborative optimization scheduling method based on low-carbon demand response as described in any embodiment of the present invention.

[0049] The present invention has the following beneficial effects:

[0050] 1. Generate accurate low-carbon signals through dynamic carbon emission factors to guide users to adjust their electricity consumption during low-carbon periods. Compared with traditional pure price signals, it can more effectively encourage users to increase electricity consumption during low-carbon periods, significantly improving carbon emission reduction effects and the accuracy of low-carbon response.

[0051] 2. Adopt a tiered carbon trading mechanism, combined with tiered carbon prices and dynamic quotas, to provide flexible low-carbon incentives, enhance users' enthusiasm in responding to low-carbon demand, enhance the synergy between demand response and carbon emission management, and optimize system carbon emissions and operating costs.

[0052] 3. Utilize reinforcement learning algorithms to enable the dispatching system to respond to fluctuations in carbon emission intensity and load demand in real time, optimize dispatching strategies based on real-time environmental feedback, achieve a dynamic balance between low carbon and economy, enhance the flexibility and real-time adaptability of the power system, and meet the low-carbon and sustainable development needs of modern power systems.

[0053] 4. Based on the expert experience of behavioral cloning and pointer network, the PPO algorithm realizes self-learning and adaptive scheduling, can continuously optimize in complex power environment, is suitable for different power system requirements, and provides technical support for future low-carbon development. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0055] Figure 1 It is a schematic diagram of the process of the present invention;

[0056] Figure 2 Flowchart of the expert experience-guided PPO algorithm based on behavior cloning and pointer networks. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0059] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0060] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0061] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0062] Embodiment 1:

[0063] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following will be combined with the specific embodiments of the present application and refer to the attached Figure 1 , clearly and completely describe the technical solution of the present invention.

[0064] To solve the problems of the prior art, the present invention provides a collaborative optimization scheduling method based on low-carbon demand response, comprising the following steps:

[0065] Step S1, constructing a source-grid-load-storage carbon emission flow measurement model to obtain carbon emission data of each part when the power system is running;

[0066] The model covers four main links in the power system: power generation sources (source side), transmission networks (grid side), energy storage devices (storage), and user loads (load side). In each link, the flow of electric energy is accompanied by the generation, transmission, and accumulation of carbon emissions. Therefore, in order to effectively reduce the carbon emissions of the overall system, these carbon emissions need to be accurately quantified and dynamically allocated. This model closely links the flow of carbon emissions with the flow of electric energy by dynamically tracking the carbon emissions of each unit of electric energy in the operation of the power system.

[0067] The source-grid-load-storage carbon emission flow metering model is constructed to obtain carbon emission data of each part when the power system is running, wherein:

[0068] Source-side carbon emissions are the main source of carbon emissions in the entire power system. In the process of power generation, traditional power generation methods that burn fossil fuels will produce a large amount of greenhouse gases such as carbon dioxide, while new energy power generation (such as wind power and photovoltaics) produces almost no direct carbon emissions during operation. Therefore, this model performs differentiated carbon emission calculations for traditional power generation units and new energy power generation units, accurately tracks the flow of carbon emissions on the source side, and provides a scientific basis for the overall low-carbon operation and scheduling of the system.

[0069] For traditional thermal power units, carbon emission intensity (carbon emissions per unit of power generation) is a key indicator in the power generation process. By modeling the fuel consumption characteristics of the unit, the formula for calculating carbon emissions from the source side of the thermal power unit can be obtained as follows:

[0070]

[0071] Where: w h is the unit power generation carbon emission intensity of thermal power unit h; a h 、b h 、c h are the fuel consumption characteristic parameters of thermal power units; P Gh represents the output power of thermal power unit h; ζ h is the correction factor; under normal operating conditions, ζ h Take 1.0; in the shutdown state, ζ h Take 0; in deep peak regulation or rapid load rise and fall state, ζ h Take 1.01.

[0072] For renewable energy generators, such as wind power and photovoltaic power, their carbon emission intensity can be considered zero because they do not use fossil fuels during operation;

[0073] This means that during the operation phase, renewable energy power generation will not directly generate carbon emissions, and is therefore an important component of achieving system low carbonization. However, it should be noted that new energy equipment may generate certain carbon emissions during the manufacturing, transportation and installation process, and these indirect carbon emissions are not specifically considered in this model.

[0074] Energy storage systems play a role in balancing supply and demand, regulating power peaks and valleys, and improving grid stability in power systems. In carbon emission management, energy storage systems dynamically adjust carbon emissions through the process of charging and discharging, thereby achieving optimized redistribution of carbon emissions in the time dimension. The carbon emission characteristics of energy storage systems depend on their different states of charging and discharging, and are affected by many factors such as energy storage type, conversion efficiency, discharge depth, and charging and discharging frequency. In order to simplify the analysis of energy storage devices in the carbon emission flow, energy storage devices are divided into three states: charging, discharging, and quasi-off-grid (no power interaction), and it is assumed that these three states can be switched quickly.

[0075] In the charging state, the energy storage device is regarded as a load of the power grid. It absorbs electric energy from the power grid for charging and accumulates and stores the carbon emissions corresponding to the total amount of charging. The carbon emission intensity of charging electricity depends on the type and proportion of power generation on the source side, so the energy storage system will accumulate corresponding carbon emissions during the charging process. These carbon emissions will be released in the subsequent discharge process.

[0076] In the discharge state, the energy storage device is regarded as a distributed power generation unit. When discharging, the energy storage system will convert the carbon emissions accumulated during the previous charging process into the carbon flow density corresponding to the unit electric power, and then inject it into the carbon emission flow calculation. The carbon emission intensity of discharge is affected by the charging carbon emissions, the depth of discharge and the conversion efficiency. The carbon emission intensity of the energy storage system can be calculated by the following method, converting the previously accumulated carbon emissions into the corresponding unit electric energy carbon emission intensity, thereby ensuring the accurate quantification of carbon emissions during the discharge process.

[0077] When the energy storage device is charged from time to time and then changes to a discharge state, the storage-side carbon emission calculation formula is as follows:

[0078]

[0079] Where: e s (T) is the discharge carbon emission intensity when the energy storage charging ends and switches to the discharge mode at time T; Q0 and F0 are the electric energy stored in the device and the accumulated carbon emissions at the last switching moment when the energy storage device switches from discharge to charging; and are the accumulated electricity and carbon flow of the energy storage device during the charging process; η is the charging and discharging efficiency of the energy storage device; in the quasi-off-grid state, there is no power interaction between the energy storage device and the grid, so there is no accumulation or release of carbon emissions in this state. This state is usually used for standby, maintenance or other scenarios that do not involve energy conversion.

[0080] The power grid is the physical carrier of electric energy transmission, and the carbon flow flows in the network with the power flow. The power flow calculation covers the mathematical model of the network, such as network nodes, branches, branch impedance, network loss and other parameters, and the carbon flow characterizes the electric energy flowing in the network based on the power flow calculation. Therefore, the carbon characterization of the grid side is reflected in the carbon flow calculation process. The total carbon emissions on the load side are calculated by the sum of the carbon flow density flowing through the adjacent nodes in each time period. Therefore, it is necessary to integrate the node carbon flow density in the required time period based on the carbon emission flow. The carbon emission calculation process on the grid side and the load side is described as follows.

[0081] In the power system, when electric energy is transmitted in the transmission network, certain power losses will be generated. These losses are closely related to the resistance, voltage level, line length and load distribution of the power grid. In order to accurately measure carbon emissions, carbon emissions need to be reasonably allocated to the source side and the load side to achieve fairness. First, through the P and Q power decomposition method, the active power and reactive power injection of each node in the power grid can be decomposed, as shown below:

[0082]

[0083] Where: P j,t , Q j,t Respectively represent the active power and reactive power injected by node j in the tth period; U j,t , U k,t Respectively represent the voltage of node j and node k in the tth period; G jk , B jk Respectively represent the conductance and susceptance of nodes j and k; θ jk,t It represents the phase difference between nodes j and k in the tth period. Through the above formula, we can get the active and reactive power flow distribution of each node in the system, so as to further determine the power situation of each node.

[0084] The extended association matrix between nodes in the power network is denoted as A = [a jk ] n×n ,in:

[0085]

[0086] P Sj =P mj +P Gj ;

[0087] Where: P jk (>0) represents the energy flow of branch jk (from node j to node k), P Sj Represents the energy flow injected into node j, and its value is the injected power P of each node unit Gj With load power P mj The sum of P ljRepresents the power of load m on node j. If the node is not connected to a load, the value is 0; P Gj Represents the energy flow of energy injection into node j. If the node is not connected to an energy injection device, its value is 0.

[0088] According to the construction of the extended correlation matrix and Kirchhoff's first law, the sum of the elements in the kth row of matrix A is equal to the load power in the kth row, that is, AE = P L , where E is a unit column vector whose dimension is equal to the number of network nodes; P L is the matrix of network node loads. We also have E = A -1 P L .

[0089] According to existing research, the node carbon footprint represents the carbon emissions corresponding to the energy flow through the node or network branch per unit time, and its calculation formula is:

[0090] R j =P j I j ;

[0091] Among them, R j is the carbon emission corresponding to the energy flow of node or network branch j, I j and P j are the carbon emission intensity and energy flow of node or network branch j, respectively.

[0092] We express the output symmetric matrix of the energy flow injected by the generator set into the node as:

[0093] P GG =diag(P G1 , P G2 , P G3 , ..., P Gn );

[0094] Carbon intensity matrix I of the energy injected into the grid by the generator set G It is expressed as:

[0095] I G =[I G1 , I G2 , I G3 , ..., I Gn ] T ;

[0096] Among them, I Gn is the carbon emission intensity of the nth power generation unit. According to the previous discussion, for thermal power units, I Gn =w i ; For new energy units (wind turbines, photovoltaics), I Gn=0; for energy storage devices, they are regarded as loads when they are charged, and I Gn =e s (T), in the quasi-off-grid state, I Gn =0.

[0097] The energy flow injected into the node by the unit is P G =P GG E, so the carbon emission of the unit injection node is R G =P G I G , we can deduce that:

[0098] R G =P GG EI G =P GG A -1 P L I G ;

[0099] The above formula reveals the relationship between the carbon emissions of the unit injection node and the carbon emissions of the load node, which can realize the tracking of the network carbon emission flow.

[0100] We will P GG A -1 Defined as the carbon emission distribution coefficient matrix, denoted as D = [d jk ]=P GG A -1 Based on this, we can allocate the carbon emissions from the power generation side to the load side:

[0101]

[0102] R Gj→k =d jk P Lk I Gj ;

[0103] In the formula, R Gj and P Lk They represent the carbon emission of node j and the load of node k respectively; R Gj→k It represents the carbon emissions allocated from the energy injection node j to the load node k.

[0104] Likewise, it is possible to allocate carbon emissions from generation to the electricity network:

[0105] R Gj→k-l =d jk P kl I Gj ;

[0106] In the formula, R Gi→k-l It represents the carbon emission corresponding to the node j connected to the generator set on the branch kl.

[0107] The calculation formula for load-side carbon emissions is as follows:

[0108] R Gj→k =d jk P Lk I Gj ;

[0109] Where: R Gj→k The carbon emissions allocated from the energy injection node j of the generator set to the load node k; d jk P is the carbon emission allocation coefficient of the node j connected to the load node k; Lk is the load of node k; I Gj The carbon emission intensity of the energy injection node j for the generator set;

[0110] The calculation formula for grid-side carbon emissions is as follows:

[0111] R Gj→k-l =d jk P kl I Gj ;

[0112] Where: R Gi→k-l is the carbon emission corresponding to the node j connected to the generator set on the branch kl; P kl is the energy flow of branch kl.

[0113] Step S2, obtaining a dynamic carbon emission factor based on the carbon emission data of each part during the operation of the power system, publishing the dynamic carbon emission factor to the user side, guiding the user to change the electricity consumption habits, and realizing the dynamic transfer of electricity load;

[0114] The dynamic carbon emission factor is obtained according to the carbon emission data of each part when the power system is running. The dynamic carbon emission factor is obtained by the following formula:

[0115]

[0116] Where: e C，i (t) represents the dynamic carbon emission factor of the user connected to the grid node i at time t; S C，i , P i are the corresponding carbon emissions and total electric energy injected from the source side to the user side connected to node i through the power grid; S D，i is the total carbon emissions of the user-side power generation equipment connected to node i; is the total power generation of the kth power generation equipment at the user side of node i; S C，i , S D，i All are obtained through carbon emission data.

[0117] The dynamic carbon emission factor is the core guiding signal in the demand response mechanism. It is generated in real time based on the source-grid-load-storage carbon emission flow metering model to reflect the carbon emission intensity per unit of electricity in the system in a specific time period. The system dynamically adjusts the factor value according to the load status, power generation structure and distribution of carbon emission flows in the power grid, and publishes it to the user side in advance so that users can perceive the carbon emission level per unit of electricity in different time periods and arrange electricity consumption reasonably.

[0118] The user side can understand the carbon emission intensity of each period based on the dynamic carbon emission factor and reasonably allocate electricity demand. For example, when the carbon emission factor value is low, the user can appropriately increase the electricity demand; when the factor value is high, the load can be reduced or transferred, thereby effectively reducing the carbon emission pressure during peak hours.

[0119] In the demand response mechanism, the load transfer mechanism is an important means to achieve low-carbon goals. This mechanism uses the signal provided by the dynamic carbon emission factor to help users transfer part of the transferable load from the high-carbon emission period to the low-carbon emission period, thereby achieving the effect of peak shaving and valley filling, balancing the system load and reducing total carbon emissions, but the total energy demand remains unchanged.

[0120] The dynamic carbon emission factor is published to the user side to guide the user to change the electricity consumption habits and realize the dynamic transfer of electricity load; the dynamic transfer model of electricity load is as follows:

[0121]

[0122] Where: are the electric power that can be transferred by the i-th user in time period t before and after the transfer; are the electric power transferred in and out by user i in period t respectively; is the corresponding transfer status; are the transfer coefficient of the transferable load and the upper and lower limits of its variation respectively.

[0123] Step S3: formulate a tiered carbon trading mechanism based on the user's electricity load transfer situation, and encourage users to optimize carbon emission costs through carbon quota allocation and graded carbon prices;

[0124] In order to further enhance the incentive effect of demand response, this patent proposes a tiered carbon trading mechanism, which provides users with reasonable incentives for low-carbon electricity consumption through carbon quota allocation and market-based transactions. This mechanism not only helps to control the overall carbon emissions of the system, but also effectively constrains users' high-carbon emission behaviors through tiered carbon prices, helping users achieve cost optimization within different carbon emission ranges.

[0125] The carbon trading mechanism is an effective policy for managing carbon emissions using market-based means. At present, when the system's carbon emissions are lower than the allocated quota, users can sell the excess quota in the market to obtain income; if the carbon emissions exceed the quota, they need to purchase additional quotas in the market to avoid penalties or pay higher carbon emission costs. At present, the carbon quota method adopted in my country is a free quota system, that is, the system allocates carbon quotas in accordance with the user's electricity demand and the energy structure of the unit, in accordance with the combination of the baseline method and the pre-allocation method. This quota allocation method takes into account factors such as the user's electricity purchase and natural gas consumption to ensure that the carbon quota is consistent with the user's energy consumption behavior.

[0126] The carbon quota allocation and graded carbon prices are used to encourage users to optimize carbon emission costs, including:

[0127] The carbon quota allocation formula is as follows:

[0128]

[0129] In the formula, E i is the total carbon emission quota allocated to user i; are the quotas for purchasing electricity from the main grid and for gas turbine use; e , χ g are the carbon emission quota corresponding to the unit electricity purchased from the main grid and the unit gas consumption of the gas turbine unit; P buy,i , P GT,i The amount of electricity purchased and gas consumed; is the quota of user i in the carbon trading market; is the actual total carbon emissions of user i; P i Represents the total electricity consumption of user i.

[0130] The tiered carbon trading mechanism helps users optimize carbon emission costs through economic incentives by setting tiered carbon prices, thus forming a strong market-based constraint. In this mechanism, the system divides the user's carbon emissions into several tiers, and sets a gradually increasing carbon price for each tier. When the user's carbon emissions reach a higher range, they will have to bear a higher carbon cost, thereby limiting high carbon emission behavior;

[0131] The formula for the graded carbon price is as follows:

[0132]

[0133] In the formula, The amount of quota used for carbon trading; is the user's carbon emission transaction cost; λ is the carbon transaction price in the market; d is the length of the carbon emission interval; α is the increase in the carbon emission price in each step.

[0134] When carbon emissions are low, users pay a lower carbon price; when emissions gradually rise to a higher level, the carbon price rises accordingly. This step-by-step carbon price structure allows users to use electricity during low-carbon emission periods under the guidance of demand response, avoiding the high costs of high-carbon emission periods. In this way, the step-by-step carbon price not only helps users save carbon emission costs, but also forms a market-based incentive to curb high-carbon emission behaviors.

[0135] Step S4, construct a multi-objective optimization problem of minimizing the carbon emission intensity and the total cost of the power system, solve the multi-objective optimization problem through the proximal strategy optimization PPO algorithm, and obtain the optimal dispatching strategy.

[0136] Comprehensively consider the operating costs and carbon emission costs on the user side to ensure the optimal overall benefits of the system. The optimization scheduling module mainly relies on the carbon emission data provided by the carbon emission measurement model, the carbon intensity signal of the time period provided by the dynamic carbon emission factor, the load adjustment brought by the demand response mechanism, and the carbon price incentive set by the step-by-step carbon trading mechanism. Construct a multi-objective optimization problem with the minimum carbon emission intensity and the minimum total cost of the power system;

[0137] The specific objective function is as follows:

[0138]

[0139] Where: T is the scheduling period; S is the total number of users; are the total gas purchase cost and unit gas price of user i in time period t respectively; are the total electricity purchase cost and unit electricity price of user i in time period t respectively; PV , DG are the penalty costs for unit solar and wind curtailment respectively; P PV,i , P DG,i Respectively represent the predicted output and actual output of photovoltaic and wind power of user i; C cur,i (t) is the wind and solar power abandonment cost of user i in period t; is the carbon trading cost of user i in period t.

[0140] In the process of optimizing the dispatching, a series of constraints must be met to ensure the safety of the system and the stability of operation. The main constraints include wind and solar output constraints, energy storage charging and discharging constraints, gas turbine output constraints and load balance constraints. The specific formula is as follows:

[0141] Wind and solar power unit output constraints:

[0142]

[0143] Energy storage charging and discharging constraints:

[0144]

[0145] In the formula, They represent the charging and discharging power of energy storage device i in time period t respectively; The maximum power of a single charge and discharge of the energy storage device; are 0-1 variables, representing the state variables of charging and discharging of the energy storage device in time period t. is the actual output power of the energy storage device in time period t, Respectively represent the charging and discharging efficiency of the energy storage device; S i (t) is the capacity percentage of energy storage device i in time period t; is the rated capacity of energy storage device i. and Respectively represent the minimum and maximum capacities of the energy storage device.

[0146] Gas turbine output constraints:

[0147]

[0148] Where P GT,i (t) and G buy,i (t) represent the power generation and natural gas consumption of the gas turbine of user i in time period t; η EB,i is the gas turbine conversion efficiency; and The upper and lower limits of gas turbine power output.

[0149] Load balancing constraints:

[0150]

[0151] Where: P L,i (t) represents the load power of user i in time period t.

[0152] The multi-objective optimization problem is solved by the proximal policy optimization PPO algorithm to obtain the optimal scheduling strategy. The proximal policy optimization PPO algorithm combines the expert experience of behavioral cloning and correction network to guide the proximal policy optimization PPO algorithm; behavioral cloning pre-training is added to the proximal policy optimization PPO algorithm to accelerate the convergence speed of the model; the pointer network is used to enhance feature extraction and enhance the accuracy of strategy generation; the actions are screened and adjusted before each round of strategy generation through expert experience to prevent high-risk decisions such as load overload or flow overload.

[0153] The multi-objective optimization problem is solved by the proximal policy optimization PPO algorithm to obtain the optimal scheduling strategy. The proximal policy optimization PPO algorithm combines the expert experience of behavioral cloning and correction network to guide the proximal policy optimization PPO algorithm; behavioral cloning pre-training is added to the proximal policy optimization PPO algorithm to accelerate the convergence speed of the model; the pointer network is used to enhance feature extraction and enhance the accuracy of strategy generation; the actions are screened and adjusted before each round of strategy generation through expert experience to prevent high-risk decisions such as load overload or flow overload.

[0154] Power system dispatch is a complex multi-constraint, multi-objective optimization problem involving multiple dynamic factors such as load demand, power generation, energy storage status, carbon emission intensity, etc. In such a complex environment, traditional optimization methods have limitations in adaptability and real-time response, while reinforcement learning can more efficiently cope with dynamic changes and nonlinear constraints in the environment by continuously optimizing strategies through the interaction between the agent and the environment.

[0155] The following key modules are introduced on the basis of the PPO algorithm: 1. Behavior cloning pre-training: By imitating expert scheduling data, behavior cloning provides an initial strategy for the intelligent agent, enabling it to make reasonable decisions at an early stage, significantly reducing the blindness of random exploration and system resource consumption, and accelerating the convergence of the model. 2. Pointer network enhanced feature extraction: Since power scheduling involves multi-dimensional and complex feature information, direct input may lead to information redundancy and feature mixing. This scheme combines the pointer network and dynamically focuses on key features (such as load demand and carbon emission intensity) through the attention mechanism, thereby extracting the most relevant information for the current decision at each time step and enhancing the accuracy of strategy generation. 3. Real-time supervision of the expert experience module: To ensure the security of strategy execution and the robustness of the system, this scheme introduces the expert experience module to screen and adjust the actions before each round of strategy generation to prevent high-risk decisions such as load overload or flow overload. In addition, the rule base is continuously updated through execution feedback, and the expert experience module has dynamic adaptability and can be flexibly adjusted in practical applications.

[0156] By comprehensively applying the PPO algorithm of reinforcement learning, behavior cloning pre-training, pointer network feature extraction and expert experience module, this solution can not only achieve effective low-carbon dispatch optimization in a dynamic and complex power system environment, but also ensure the safety and stability of system operation. Specifically, it includes the following steps:

[0157] 1. Initialization:

[0158] Expert data collection: Extract state-action pairs from historical scheduling records as expert demonstration data for behavior cloning pre-training.

[0159] Rule base construction: Build a rule base of expert experience modules based on the safety standards of power grid dispatch, including load rate restrictions, flow constraints and power allocation.

[0160] Environment Definition: The power system is modeled as a Markov decision process (MDP) for reinforcement learning, and the state space, action space, and reward function are defined to provide feedback during the dispatch optimization process.

[0161] 2. Behavior cloning pre-training:

[0162] In the behavior cloning pre-training stage, by imitating the strategy of expert demonstration data, a reasonable starting point is provided for subsequent PPO training, reducing the risk of initial exploration.

[0163] Input expert demonstration data: The collected expert demonstration data (i.e., state-action pairs) are input into the behavior cloning model to train the agent to imitate the expert’s scheduling behavior.

[0164] Calculate loss and optimize: Use the cross entropy loss function to calculate the difference between the agent output and the expert strategy, and back-propagate to optimize the policy network so that its output actions match the expert strategy.

[0165] Generate an initial policy network: After completing pre-training, a preliminary policy network is obtained to provide a basic strategy for the PPO optimization stage.

[0166] 3. Experience sampling and storage:

[0167] State Inputs t : At each time step t, receive the current state s of the system t ={D t ,P t ,E t ,C t ,P c,t The status information includes key information such as load demand, power generation, energy storage status, carbon emission intensity and carbon trading price, providing basic data for decision-making for the intelligent agent.

[0168] Pointer network feature extraction:

[0169] Feature encoding: converting state information into hidden vectors through encoders

[0170] Attention weight calculation: The pointer network calculates the attention weight of the hidden vector and dynamically focuses on the key features of the current decision (such as load demand and carbon emission intensity).

[0171] Weighted feature output: The weighted key feature vector is input into the policy network to generate a more accurate scheduling action distribution.

[0172] Action selection a t: According to the action probability distribution π of the policy network θ (a|s), the agent is in the current state s t Downsample an action a t Action a t These actions, including power generation adjustment, energy storage charging and discharging, demand response and carbon trading decisions, constitute the dispatching strategy at the current moment.

[0173] Real-time strategy screening and adjustment of expert experience module:

[0174] Strategy screening: In action a t Before execution, the expert experience module screens the action according to the rule base to check whether it meets the safety requirements of the power system.

[0175] Real-time adjustment: If a t If the safety constraints in the rule base are not met, the expert experience module will make real-time adjustments to ensure the safety of the action.

[0176] Environmental interaction and instant feedback:

[0177] The agent performs the adjusted action a t , and interact with the environment. The environment returns an immediate reward r t and the next state s t+1 , instant reward t It is used to measure the effect of the current action, and s t+1 Reflects action a t The impact on the system state. The purpose of this step is to verify the scheduling strategy of the intelligent agent through environmental feedback. t , a t , r t ,s t+1 ) is stored in the experience pool for subsequent PPO training.

[0178] Determine whether the experience pool is full: If the experience pool is not full, return to the state input and continue to collect experience; if the experience pool is full, enter the next step of the PPO training process.

[0179] 4. PPO training process:

[0180] Sample batches from the experience pool:

[0181] Batch sample interaction data from the experience pool, including (s t , a t , r t ,s t+1 ), used for updating the strategy and value network.

[0182] Calculate the advantage function Advantage function in PPO algorithm It is used to measure the pros and cons of the current action relative to the average level. The advantage function is calculated based on the immediate reward r t and the value of the next state V φ (s t+1 ):

[0183]

[0184] Where γ is a discount factor that measures the impact of future rewards on current decisions. Advantage function The positive or negative value of determines whether the current action should be preferred by the policy network to help the agent optimize the strategy.

[0185] Policy network update: The policy network update is implemented through the PPO clipping policy gradient method to ensure the stability of the update. The updated objective function is:

[0186]

[0187] Probability ratio calculation: Indicates that the current strategy and the old strategy are in action a t The probability ratio on indicates the optimization degree of the current strategy.

[0188] Clip operation: clip(r t (θ), 1-∈, 1+∈) limits the amplitude of policy updates to between 1-∈ and 1+∈ to avoid excessive policy updates and ensure the stability of the policy network. The clipping mechanism can prevent oscillation problems caused by excessive policy gradients.

[0189] Value network update: The value network update uses the mean square error minimization method to improve the accuracy of state value estimation. The loss function is:

[0190] L value (φ)=E t [(r t +γV φ (s t+1 )-V φ (s t )) 2 ];

[0191] This loss function calculates the error between the actual reward and the estimated value. By minimizing this error, the value network can more accurately estimate the long-term return of each state and provide effective value feedback to the policy network.

[0192] Determine whether the training has converged:

[0193] If the training has not converged (the loss function is still higher than the set threshold), return to the experience sampling and storage module to continue collecting new samples; if the training converges, enter the rule base dynamic update step.

[0194] 5. Dynamic update of rule base:

[0195] Execution feedback record: After each round of strategy execution, record the feedback data of the scheduling results, including load rate, power flow fluctuation, etc.

[0196] Rule base update: If feedback shows that certain strategies cause system instability, the system will add this feedback information to the rule base and dynamically adjust the safety constraints to ensure that the rule base can adapt to different grid conditions.

[0197] 6. Final strategy output:

[0198] After training convergence, the optimized policy network is output and used for low-carbon demand response scheduling in actual power grids to meet the goal of source-grid-load-storage coordinated optimization.

[0199] Embodiment 2:

[0200] This embodiment provides a collaborative optimization scheduling system based on low-carbon demand response, including:

[0201] The carbon emission data collection module builds a source-grid-load-storage carbon emission flow measurement model to obtain carbon emission data of each part during the operation of the power system;

[0202] Dynamic transfer of electricity load: dynamic carbon emission factors are obtained based on the carbon emission data of each part of the power system during operation, and the dynamic carbon emission factors are published to the user side to guide users to change their electricity consumption habits and realize dynamic transfer of electricity load;

[0203] The module for building a tiered carbon trading mechanism is to formulate a tiered carbon trading mechanism based on the user's electricity load transfer, and to encourage users to optimize carbon emission costs through carbon quota allocation and graded carbon prices;

[0204] The multi-objective optimization solution module constructs a multi-objective optimization problem of minimizing the carbon emission intensity and total cost of the power system, solves the multi-objective optimization problem through the proximal strategy optimization PPO algorithm, and obtains the optimal dispatching strategy.

[0205] Embodiment three:

[0206] This embodiment provides an electronic device having a computer program stored thereon, and when the computer program is executed by a processor, a collaborative optimization scheduling method based on low-carbon demand response as described in any embodiment of the present invention is implemented.

[0207] Embodiment 4:

[0208] This embodiment provides a computer-readable medium for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement a collaborative optimization scheduling method based on low-carbon demand response as described in any embodiment of the present invention.

[0209] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0210] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0211] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0212] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0213] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A collaborative optimization scheduling method based on low-carbon demand response, characterized in that: The following steps are involved: Step S1, constructing a source-grid-load-storage carbon emission flow measurement model to obtain carbon emission data of each part when the power system is running; Step S2, obtaining a dynamic carbon emission factor based on the carbon emission data of each part during the operation of the power system, publishing the dynamic carbon emission factor to the user side, guiding the user to change the electricity consumption habits, and realizing the dynamic transfer of electricity load; Step S3: formulate a tiered carbon trading mechanism based on the user's electricity load transfer situation, and encourage users to optimize carbon emission costs through carbon quota allocation and graded carbon prices; Step S4, construct a multi-objective optimization problem of minimizing the carbon emission intensity and the total cost of the power system, solve the multi-objective optimization problem through the proximal strategy optimization PPO algorithm, and obtain the optimal dispatching strategy.

2. The collaborative optimization scheduling method based on low-carbon demand response according to claim 1 is characterized in that: The source-grid-load-storage carbon emission flow metering model is constructed to obtain carbon emission data of each part when the power system is running, wherein: The formula for calculating source-side carbon emissions is as follows: Where: w h is the unit power generation carbon emission intensity of thermal power unit h; a h , b h 、c h are the fuel consumption characteristic parameters of thermal power units; P Gh represents the output power of thermal power unit h; ζ h is the correction factor; The calculation formula for storage-side carbon emissions is as follows: Where: e s (T) is the discharge carbon emission intensity when the energy storage charging ends and switches to the discharge mode at time T; Q0 and F0 are the electric energy stored in the device and the accumulated carbon emissions at the last switching moment when the energy storage device switches from discharge to charging; and are the accumulated electricity and carbon flow of the energy storage device during the charging process; η is the charging and discharging efficiency of the energy storage device; The calculation formula for load-side carbon emissions is as follows: R Gj→k =d jk P Lk I Gj ; Where: R Gi→k The carbon emissions allocated from the energy injection node j of the generator set to the load node k; d jk P is the carbon emission allocation coefficient of the node j connected to the load node k; Lk is the load of node k; I Gj The carbon emission intensity of the energy injection node j for the generator set; The calculation formula for grid-side carbon emissions is as follows: R Gj→k-1 =d jk P kl I Gj ; Where: R Gi→k-l is the carbon emission corresponding to the node j connected to the generator set on the branch kl; P kl is the energy flow of branch kl.

3. The collaborative optimization scheduling method based on low-carbon demand response according to claim 1 is characterized in that: The dynamic carbon emission factor is obtained according to the carbon emission data of each part when the power system is running. The dynamic carbon emission factor is obtained by the following formula: Where: e C，i (t) represents the dynamic carbon emission factor of the user connected to the grid node i at time t; S C，i , P i are the corresponding carbon emissions and total electric energy injected from the source side to the user side connected to node i through the power grid; S D，i is the total carbon emissions of the user-side power generation equipment connected to node i; is the total power generation of the kth power generation equipment at the user side of node i; S C，i , S D，i All are obtained through carbon emission data.

4. The collaborative optimization scheduling method based on low-carbon demand response according to claim 1 is characterized in that: The dynamic carbon emission factor is published to the user side to guide the user to change the electricity consumption habits and realize the dynamic transfer of electricity load; the dynamic transfer model of electricity load is as follows: Where: are the electric power that can be transferred by the i-th user in time period t before and after the transfer; are the electric power transferred in and out by user i in period t respectively; is the corresponding transfer status; are the transfer coefficient of the transferable load and the upper and lower limits of its variation respectively.

5. The collaborative optimization scheduling method based on low-carbon demand response according to claim 1 is characterized in that: The carbon quota allocation and graded carbon prices are used to encourage users to optimize carbon emission costs, including: The carbon quota allocation formula is as follows: In the formula, E i is the total carbon emission quota allocated to user i; are the quotas for purchasing electricity from the main grid and for gas turbine use; e , χ g are the carbon emission quota corresponding to the unit electricity purchased from the main grid and the unit gas consumption of the gas turbine unit; P buy,i , P GT,i The amount of electricity purchased and gas consumed; is the quota of user i in the carbon trading market; is the actual total carbon emissions of user i; P i represents the total electricity consumption of user i; The formula for the graded carbon price is as follows: In the formula, The amount of quota used for carbon trading; is the user's carbon emission transaction cost; λ is the carbon transaction price in the market; d is the length of the carbon emission interval; α is the increase in the carbon emission price in each step.

6. The collaborative optimization scheduling method based on low-carbon demand response according to claim 1 is characterized in that: The multi-objective optimization problem of minimizing the carbon emission intensity and total cost of the power system is constructed, and the minimum total operating cost of each user side is taken as the objective function, where the total cost includes the electricity purchase cost, gas purchase cost, wind and solar abandonment cost and carbon trading cost. The specific objective function is as follows: Where: T is the scheduling period; S is the total number of users; are the total gas purchase cost and unit gas price of user i in time period t respectively; are the total electricity purchase cost and unit electricity price of user i in time period t respectively; PV , DG are the penalty costs for unit solar and wind curtailment respectively; P PV,i , P DG,i Respectively represent the predicted output and actual output of photovoltaic and wind power of user i; C cur,i (t) is the wind and solar power abandonment cost of user i in period t; is the carbon trading cost of user i in period t.

7. The collaborative optimization scheduling method based on low-carbon demand response according to claim 1 is characterized in that: The multi-objective optimization problem is solved by the proximal policy optimization PPO algorithm to obtain the optimal scheduling strategy. The proximal policy optimization PPO algorithm combines the expert experience of behavioral cloning and correction network to guide the proximal policy optimization PPO algorithm; behavioral cloning pre-training is added to the proximal policy optimization PPO algorithm to accelerate the convergence speed of the model; the pointer network is used to enhance feature extraction and enhance the accuracy of strategy generation; the actions are screened and adjusted before each round of strategy generation through expert experience to prevent high-risk decisions such as load overload or flow overload.

8. A collaborative optimization scheduling system based on low-carbon demand response, characterized in that: include: The carbon emission data collection module builds a source-grid-load-storage carbon emission flow measurement model to obtain carbon emission data of each part during the operation of the power system; Dynamic transfer of electricity load: dynamic carbon emission factors are obtained based on the carbon emission data of each part of the power system during operation, and the dynamic carbon emission factors are published to the user side to guide users to change their electricity consumption habits and realize dynamic transfer of electricity load; The module for building a tiered carbon trading mechanism is to formulate a tiered carbon trading mechanism based on the user's electricity load transfer, and to encourage users to optimize carbon emission costs through carbon quota allocation and graded carbon prices; The multi-objective optimization solution module constructs a multi-objective optimization problem of minimizing the carbon emission intensity and total cost of the power system, solves the multi-objective optimization problem through the proximal strategy optimization PPO algorithm, and obtains the optimal dispatch strategy.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the collaborative optimization scheduling method based on low-carbon demand response as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a collaborative optimization scheduling method based on low-carbon demand response as described in any one of claims 1 to 7 is implemented.

Images