A Demand-Side Electricity-Carbon Coupling Trading Method Based on Virtual Carbon Storage

Abstract

This invention discloses a demand-side electricity-carbon coupling trading method based on virtual carbon storage, comprising the following steps: establishing a user dynamic model and objective function; users participating in the demand-side electricity-carbon coupling market and joining "virtual carbon storage," enabling users to bid on carbon emission intensity; and using a model-free, data-driven near-end strategy optimization algorithm to solve the minimization problem. This invention further releases the load adjustment potential of highly flexible users by enabling users with varying levels of flexibility to participate in the demand-side electricity-carbon coupling market and trade carbon emission responsibilities, thereby achieving a reduction in overall system carbon emissions. This invention proposes the concept of "virtual carbon storage," making the carbon emission intensity of electricity from energy storage a controllable variable, relaxing the tight coupling between carbon emission responsibility and power flow caused by the proportional sharing theorem, thus enabling flexible transfer of carbon emission responsibility among users and allowing users to flexibly bid on carbon emission intensity.

Description

Technical Field

[0001] This invention relates to power systems, and in particular to a demand-side electricity-carbon coupling trading method based on virtual carbon storage. Background Technology

[0002] A consumption pattern dominated by fossil fuels has contributed to global warming, and carbon dioxide accounts for more than half of greenhouse gases. Reducing carbon emissions has become a major challenge for economic and social development. Carbon emissions directly generated by the power industry account for more than 40% of my country's total carbon emissions, making the low-carbon transformation of the power system an inevitable path. Existing research mostly focuses on direct carbon emissions from the generation side. However, due to the "source follows load" characteristic of the power system, actual electricity demand is the root cause of carbon emissions from the generation side. Users' electricity consumption behavior affects system carbon emissions, therefore they should bear corresponding carbon emission responsibility. Demand-side management can guide users to change their electricity consumption habits through time-varying price signals. Therefore, combining demand-side management with low-carbon policies will fully unleash the carbon reduction potential of the demand side.

[0003] Existing research is largely based on carbon emission flow theory, tracking carbon emissions generated on the generation side and transferring the responsibility for carbon emissions from the generation side to the demand side. Carbon emission flows parallel to power flow in the power grid, originating from the generation side, passing through the grid, and ultimately reaching end users. By calculating the carbon emission flow, the carbon emission intensity at user nodes can be obtained. The carbon emission intensity on the generation side will be reflected in the node carbon potential on the demand side. Users will receive price signals that change in real time with system carbon emissions. Combined with demand-side management principles, this allows for guiding users to adjust their electricity consumption behavior through price signals, thereby reducing system carbon emissions.

[0004] However, existing technologies still have the following drawbacks:

[0005] 1. Based on existing research, the responsibility for carbon emissions can be shared by users. However, guiding users to adjust their electricity consumption behavior to reduce carbon emissions solely through time-varying price signals has significant limitations. This is due to the large differences in flexibility among different types of loads in the power system. Flexible users in the power system can adjust the size of their load or the duration of their usage to reduce costs based on changes in price signals. However, rigid loads have very poor flexibility. These inflexible users cannot adjust their electricity consumption strategies when electricity prices change. Therefore, inflexible users still bear a higher carbon emission responsibility during periods of high system carbon emissions, resulting in higher costs.

[0006] 2. Existing research only treats carbon emissions as an additional cost of electricity consumption for users, neglecting the impact of interactions between different types of users on system carbon emissions. In reality, flexible users have room to further adjust their electricity consumption strategies. Simply guiding users to adjust their electricity consumption behavior through time-varying price signals to reduce carbon emissions does not reach the optimization limit of demand-side carbon reduction. If market transactions could be used to transfer carbon emission responsibility from inflexible users to flexible users, flexible users would further increase load regulation, thereby reducing the total system carbon emissions and further releasing the carbon reduction potential on the demand side. Summary of the Invention

[0007] Purpose of the invention: The purpose of this invention is to provide a demand-side electricity-carbon coupling trading method based on virtual carbon storage, thereby making carbon trading more flexible and further reducing system carbon emissions.

[0008] Technical solution: The present invention provides a demand-side electricity-carbon coupling trading method based on virtual carbon storage, comprising the following steps:

[0009] (1) Establish the user dynamic model and objective function.

[0010] (2) Users participate in the demand-side electricity-carbon coupling market and join "virtual carbon storage" so that users can bid on carbon emission intensity.

[0011] (3) Use the model-free, data-driven Proximal Policy Optimization (PPO) algorithm to solve the objective function in step (1), i.e. the cost minimization problem.

[0012] The specific steps (1) are as follows:

[0013] (1.1) First, establish the user's energy storage model. In a regional power system where users are equipped with energy storage and photovoltaic power generation, the dynamic process of user energy storage is represented as follows:

[0014]

[0015] Among them, SOC t η represents the state of charge of the stored energy at the beginning of time period t. c and η d These represent the charging and discharging efficiencies of energy storage, respectively. and Let C represent the charge and discharge amounts of energy storage during time period t; the cost of energy storage charge and discharge losses is denoted by C. bess express:

[0016] C bess,t =τ bess |SOC t -SOC t-1|(2)

[0017] Where, τ bess It is a coefficient related to the cost of energy storage charging and discharging losses, |SOC t -SOC t-1 | indicates SOC t -SOC t-1 The absolute value;

[0018] (1.2) Establish a adjustment cost model. The adjustment cost will vary greatly depending on the user type, so there are multiple forms of representation. Taking the form of a quadratic function as an example:

[0019] C a,t =e t (ΔQ load,t ) 2 +b t ΔQ load,t (3)

[0020] Among them, C a,t e represents the adjustment cost incurred by users in adjusting their electricity consumption strategies during time period t. t and b t To adjust the coefficients of the cost function, ΔQ load,t The amount of electricity adjusted by a user during time period t must satisfy the following constraints:

[0021]

[0022] in, It is the maximum adjustment amount of the load;

[0023] (1.3) Establish the objective function, where C is the cost C incurred by the user purchasing electricity during time period t. buy,t for:

[0024] C buy,t =τ e,t Q buy,t (5)

[0025] Where, τ e The predicted electricity price for time period t; the revenue C from selling electricity. sell,t for:

[0026] C sell,t =τ e,t Q sell,t (6)

[0027] The carbon emission cost C caused by the user purchasing electricity during time period t. c,t for:

[0028] C c,t =τ c (Q bess,t +Qload,t (7)

[0029] Where τ c The carbon price in the carbon market is used to determine the user's total cost C during time period t. user,t :

[0030] C user,t =C bess,t +C a,t +C c,t +C buy,t -C sell,t (8)

[0031] The ultimate optimization goal is to minimize the user's daily operating cost:

[0032]

[0033] Step (2) specifically involves:

[0034] (2.1) Users make virtual bids in the first stage of the demand-side electric carbon coupling market. The buyer's bid is a monotonically decreasing linear function, and the seller's bid is a monotonically increasing linear function.

[0035] (2.2) Demand-side electricity-carbon coupling market clearing, the clearing result is electricity price and electricity volume. In the first stage, the market only retains electricity price data and feeds it back to users for their reference.

[0036] (2.3) Establish electricity balance constraints and carbon emission balance constraints for the second phase of carbon trading; the electricity consumption of participating users in time period t must meet the balance constraints:

[0037] Q buy,t +Q pv,t -Q bess,t =Q sell,t +Q load,t (10)

[0038] Among them, Q buy,t Q represents the amount of electricity purchased by users during time period t. pv,t Q represents the photovoltaic output during time period t. bess,t Q represents the amount of energy stored during time period t. sell,t Q represents the electricity sold by users during time period t. load,t This represents the electricity consumed by the load after the user adjusts their electricity consumption strategy during time period t; the carbon emissions coupled with the above electricity consumption satisfy the following constraints:

[0039] Q buy,t ρ buy,t +Q pv,t ρ pv,t -Q bess,t ρ bess,t =Qsell,t ρ sell,t +Q load,t ρ load,t (11)

[0040] Where, ρ buy,t ρ pv,t ρ bess,t ρ sell,t and ρ load,t These are the carbon emission intensities corresponding to the electricity purchased during time period t, photovoltaic output, energy storage charging and discharging, electricity sold, and electricity consumed by the load.

[0041] (2.4) The carbon emission intensity corresponding to the electricity consumption of the user load at this time can be derived from equations (10) and (11):

[0042]

[0043] When users sell electricity, Q buy,t =0, and the carbon emission intensity of electricity sold by users is equal to the carbon emission intensity of electricity consumed by user loads. Therefore, the carbon emission intensity of electricity sold is:

[0044]

[0045] (2.5) Establish a carbon emission flow model for user energy storage; the carbon emission intensity corresponding to the electricity stored in the energy storage is expressed by the following formula:

[0046]

[0047] in, It is the amount of electricity stored in the energy storage system during time period t. It represents the carbon emission intensity corresponding to the electricity stored in the energy storage during time period t; the change in carbon emission intensity during energy storage charging depends on the carbon emission intensity of the charging source and the current carbon emission intensity within the energy storage, and the same applies during energy storage discharging.

[0048] (2.6) Add "virtual carbon storage" to enable users to bid on carbon emission intensity; according to formula (13), ρ at this time sell,t Depends only on variable Q bess,t This means that the carbon emission intensity corresponding to the electricity sold by users is fixed, making it difficult to flexibly transfer carbon emission responsibility between users and thus restricting carbon emission trading. To improve the flexibility of carbon trading, "virtual carbon storage" is introduced, starting with energy storage. "Virtual carbon storage" refers to increasing the carbon emission intensity of energy storage while maintaining the carbon emission responsibility transferred from the generation side to the demand side unchanged. bess,t Becoming a controllable variable breaks the constraints imposed by the proportional sharing theorem; ρ sell,t Flexibility is greatly improved; the carbon intensity of the electricity stored in energy storage is significantly reduced. When energy storage becomes a controllable variable, it can temporarily store carbon emission liability. To prevent users from storing carbon emission intensity exceeding their affordability, the carbon emission intensity of the remaining electricity in the energy storage must meet the following constraints:

[0049]

[0050] (2.7) Thanks to the “virtual carbon storage” introduced in step (2.6), users will bid on electricity and carbon emission intensity in the second phase of the demand-side electricity-carbon coupling market. The buyer’s bid is a monotonically decreasing linear function, and the seller’s bid is a monotonically increasing linear function.

[0051] Step (3) specifically involves:

[0052] (3.1) First, set the status to:

[0053]

[0054] Where, τ e,t Indicates electricity price, μ t σ represents the basic carbon intensity of electricity in the bid price. in,t and σ out,t μ represents the variable coefficient describing carbon emission intensity when buying and selling electricity during time period t. t σ in,t and σ out,t All depend on the external market.

[0055] (3.2) Set the action space, as follows:

[0056]

[0057] in, This represents the amount of electricity purchased by users in the market during time period t. This represents the user's load adjustment amount during time period t. This represents the amount of energy stored during time period t. This indicates the decision made by users regarding their carbon emission intensity bids when selling electricity.

[0058] (3.3) Set the reward, reward function r t This can be represented as the user's benefit:

[0059]

[0060] Where, α t (s t+1 ) is the energy storage SOC over-limit penalty function, if from state s t Updated to s t+1 If the state of charge of the stored energy exceeds the upper limit or falls below the lower limit, a penalty value will be accumulated. It is the adjustment cost offset penalty function, in state s t If condition (3) is not met, a penalty value will be accumulated once. It is the power imbalance penalty function, in state s t If condition (10) is not met, a penalty value is accumulated. To limit the impact of future states on current rewards, future rewards need to be multiplied by a discount factor. The accumulated reward after discount is:

[0061]

[0062] Where γ is the discount factor.

[0063] (3.4) Input the environmental information into the new actor network to obtain the mean μ and variance σ of the two-dimensional normal distribution. Sample an action a, and then input it into the environment to obtain the reward r and the state of the next step. Then store the above results, and then... Input into the network and repeat step (3.2).

[0064] (3.5) Repeat step (3.4) until the last step is completed. The input is fed into the critic-NN network to obtain the state. The value is then calculated, and the discount reward in equation (19) is calculated. All stored combinations of s are input into the critic-NN network to obtain the values ​​of all states. Value, calculate the loss c loss =mean(square(A) t Then, backpropagate to update the critic-NN network.

[0065] (3.6) Input all stored s combinations into the original actor network (actor-old) and actor-new network to obtain normal distribution 1 and normal distribution 2 respectively. Input all stored action combinations into normal distribution 1 and 2 respectively to obtain the probabilities P1 and P2 corresponding to each action. Then divide P2 by P1 to obtain the importance weight. Calculate the loss function.

[0066]

[0067] in This is the advantage estimate of the advantage function, where ε is a hyperparameter, and the clip function represents the probability of target clipping. Then, backpropagation is performed to update the actor-new network.

[0068] (3.7) Repeat step (3.6) until the number of steps set according to the actual scenario is reached, and then the loop ends. Use the weights of the actor-new network to update the weights of the actor-old network. Repeat step (3.4) to step (3.6) until the loss function is less than the accuracy ω set according to the actual scenario.

[0069] (3.8) Use the PPO algorithm from steps (3.1) to (3.7) to determine the user's adjusted electricity consumption strategy in the second stage, and use [Q] buy,t Q sell,t ,ΔQ load,t Q bess,t The composition is as follows: Specifically, in the second phase of the demand-side electricity-carbon coupling market, users bid for carbon emission intensity and electricity volume. After the second phase market clears, users obtain carbon emission intensity, electricity price, and electricity volume, and their electricity purchase and sale strategies [Q]. buy,t Q sell,t This leads to the conclusion.

[0070] A computer storage medium storing a computer program that, when executed by a processor, implements the aforementioned demand-side electric carbon coupling trading method based on virtual carbon storage.

[0071] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned demand-side electric carbon coupling trading method based on virtual carbon storage.

[0072] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0073] 1. This invention enables different types of users participating in the electricity-carbon coupling market to bid on carbon emission intensity, thereby reducing the overall carbon emissions of the system through demand-side trading, and allowing users with different levels of flexibility to benefit from demand-side carbon trading.

[0074] 2. This invention proposes the concept of "virtual carbon storage," which makes the carbon emission intensity of energy storage a controllable variable, loosening the tight coupling between carbon emission responsibility and tidal current brought about by the proportional sharing theorem, thereby realizing the flexible transfer of carbon emission responsibility among users and allowing users to bid on carbon emission intensity. Attached Figure Description

[0075] Figure 1 This is a flowchart of the steps of the method described in this invention;

[0076] Figure 2 This outlines the operational sequence of the demand-side electro-carbon coupling market. Detailed Implementation

[0077] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0078] like Figure 1 As shown, a demand-side electricity-carbon coupling trading method based on virtual carbon storage includes the following steps:

[0079] (1) Establish the user dynamic model and objective function.

[0080] (2) Users participate in the demand-side electricity-carbon coupling market and join "virtual carbon storage" so that users can bid on carbon emission intensity.

[0081] (3) Use the model-free, data-driven Proximal Policy Optimization (PPO) algorithm to solve the objective function in step (1), i.e. the cost minimization problem.

[0082] The specific steps (1) are as follows:

[0083] (1.1) First, establish the user's energy storage model. In a regional power system where users are equipped with energy storage and photovoltaic power generation, the dynamic process of user energy storage is represented as follows:

[0084]

[0085] Among them, SOC t η represents the state of charge of the stored energy at the beginning of time period t. c and η d These represent the charging and discharging efficiencies of energy storage, respectively. and Let C represent the charge and discharge amounts of energy storage during time period t; the cost of energy storage charge and discharge losses is denoted by C. bess express:

[0086] C bess,t =τ bess |SOC t -SOC t-1 | (2)

[0087] Where, τ bess It is a coefficient related to the cost of energy storage charging and discharging losses, |SOC t -SOC t-1 | indicates SOC t -SOC t-1 The absolute value;

[0088] (1.2) Establish a adjustment cost model. The adjustment cost will vary greatly depending on the user type, so there are multiple forms of representation. Taking the form of a quadratic function as an example:

[0089] C a,t =et (ΔQ load,t ) 2 +b t ΔQ load,t (3)

[0090] Among them, C a,t e represents the adjustment cost incurred by users in adjusting their electricity consumption strategies during time period t. t and b t To adjust the coefficients of the cost function, ΔQ load,t The amount of electricity adjusted by a user during time period t must satisfy the following constraints:

[0091]

[0092] in, It is the maximum adjustment amount of the load;

[0093] (1.3) Establish the objective function, where C is the cost C incurred by the user purchasing electricity during time period t. buy,t for:

[0094] C buy,t =τ e,t Q buy,t (5)

[0095] Where, τ e The predicted electricity price for time period t; the revenue C from selling electricity. sell,t for:

[0096] C sell,t =τ e,t Q sell,t (6)

[0097] The carbon emission cost C caused by the user purchasing electricity during time period t. c,t for:

[0098] C c,t =τ c (Q bess,t +Q load,t (7)

[0099] Where τ c The carbon price in the carbon market is used to determine the user's total cost C during time period t. user,t :

[0100] C user,t =C bess,t +C a,t +C c,t +C buy,t -C sell,t (8)

[0101] The ultimate optimization goal is to minimize the user's daily operating cost:

[0102]

[0103] Step (2) specifically involves:

[0104] (2.1) Users make virtual bids in the first stage of the demand-side electric carbon coupling market. The buyer's bid is a monotonically decreasing linear function, and the seller's bid is a monotonically increasing linear function.

[0105] (2.2) Demand-side electricity-carbon coupling market clearing, the clearing result is electricity price and electricity volume, such as Figure 2 As shown, in the first phase, the market only retains electricity price data and provides it to users for their reference.

[0106] (2.3) Establish electricity balance constraints and carbon emission balance constraints for the second phase of carbon trading; the electricity consumption of participating users in time period t must meet the balance constraints:

[0107] Q buy,t +Q pv,t -Q bess,t =Q sell,t +Q load,t (10)

[0108] Among them, Q buy,t Q represents the amount of electricity purchased by users during time period t. pv,t Q represents the photovoltaic output during time period t. bess,t Q represents the amount of energy stored during time period t. sell,t Q represents the electricity sold by users during time period t. load,t This represents the electricity consumed by the load after the user adjusts their electricity consumption strategy during time period t; the carbon emissions coupled with the above electricity consumption satisfy the following constraints:

[0109] Q buy,t ρ buy,t +Q pv,t ρ pv,t -Q bess,t ρ bess,t =Q sell,t ρ sell,t +Q load,t ρ load,t (11)

[0110] Where, ρ buy,t ρ pv,t ρ bess,t ρ sell,t and ρ load,t These are the carbon emission intensities corresponding to the electricity purchased during time period t, photovoltaic output, energy storage charging and discharging, electricity sold, and electricity consumed by the load.

[0111] (2.4) The carbon emission intensity corresponding to the electricity consumption of the user load at this time can be derived from equations (10) and (11):

[0112]

[0113] When users sell electricity, Q buy,t =0, and the carbon emission intensity of electricity sold by users is equal to the carbon emission intensity of electricity consumed by user loads. Therefore, the carbon emission intensity of electricity sold is:

[0114]

[0115] (2.5) Establish a carbon emission flow model for user energy storage; the carbon emission intensity corresponding to the electricity stored in the energy storage is expressed by the following formula:

[0116]

[0117] in, It is the amount of electricity stored in the energy storage system during time period t. It represents the carbon emission intensity corresponding to the electricity stored in the energy storage during time period t; the change in carbon emission intensity during energy storage charging depends on the carbon emission intensity of the charging source and the current carbon emission intensity within the energy storage, and the same applies during energy storage discharging.

[0118] (2.6) Add "virtual carbon storage" to enable users to bid on carbon emission intensity; according to formula (13), ρ at this time sell,t Depends only on variable Q bess,t This means that the carbon emission intensity corresponding to the electricity sold by users is fixed, making it difficult to flexibly transfer carbon emission responsibility between users and thus restricting carbon emission trading. To improve the flexibility of carbon trading, "virtual carbon storage" is introduced, starting with energy storage. "Virtual carbon storage" refers to increasing the carbon emission intensity of energy storage while maintaining the carbon emission responsibility transferred from the generation side to the demand side unchanged. bess,t Becoming a controllable variable breaks the constraints imposed by the proportional sharing theorem; ρ sell,t Flexibility is greatly improved; the carbon intensity of the electricity stored in energy storage is significantly reduced. When energy storage becomes a controllable variable, it can temporarily store carbon emission liability. To prevent users from storing carbon emission intensity exceeding their affordability, the carbon emission intensity of the remaining electricity in the energy storage must meet the following constraints:

[0119]

[0120] (2.7) Benefiting from the "virtual carbon storage" introduced in step (2.6), users will bid on electricity and carbon emission intensity in the second phase of the demand-side electricity-carbon coupling market. For example... Figure 2 As shown, the buyer's quote is a monotonically decreasing linear function, and the seller's quote is a monotonically increasing linear function.

[0121] Step (3) specifically involves:

[0122] (3.1) First, set the status to:

[0123]

[0124] Where, τ e,t Indicates electricity price, μ t σ represents the basic carbon intensity of electricity in the bid price. in,t and σ out,t μ represents the variable coefficient describing carbon emission intensity when buying and selling electricity during time period t. t σ in,t and σ out,t All depend on the external market.

[0125] (3.2) Set the action space, as follows:

[0126]

[0127] in, This represents the amount of electricity purchased by users in the market during time period t. This represents the user's load adjustment amount during time period t. This represents the amount of energy stored during time period t. This indicates the decision made by users regarding their carbon emission intensity bids when selling electricity.

[0128] (3.3) Set the reward, reward function r t This can be represented as the user's benefit:

[0129]

[0130] Where, α t (s t+1 ) is the energy storage SOC over-limit penalty function, if from state s t Updated to s t+1 If the state of charge of the stored energy exceeds the upper limit or falls below the lower limit, a penalty value will be accumulated. It is the adjustment cost offset penalty function, in state s t If condition (3) is not met, a penalty value will be accumulated once. It is the power imbalance penalty function, in state s t If condition (10) is not met, a penalty value is accumulated. To limit the impact of future states on current rewards, future rewards need to be multiplied by a discount factor. The accumulated reward after discount is:

[0131]

[0132] Where γ is the discount factor.

[0133] (3.4) Input the environmental information into the new actor network to obtain the mean μ and variance σ of the two-dimensional normal distribution. Sample an action a, and then input it into the environment to obtain the reward r and the state of the next step. Then store the above results, and then... Input into the network and repeat step (3.2).

[0134] (3.5) Repeat step (3.4) until the last step is completed. The input is fed into the critic-NN network to obtain the state. The value is then calculated, and the discount reward in equation (19) is calculated. All stored combinations of s are input into the critic-NN network to obtain the values ​​of all states. Value, calculate the loss c loss =mean(square(A) t Then, backpropagate to update the critic-NN network.

[0135] (3.6) Input all stored s combinations into the original actor network (actor-old) and actor-new network to obtain normal distribution 1 and normal distribution 2 respectively. Input all stored action combinations into normal distribution 1 and 2 respectively to obtain the probabilities P1 and P2 corresponding to each action. Then divide P2 by P1 to obtain the importance weight. Calculate the loss function.

[0136]

[0137] in This is the advantage estimate of the advantage function, where ε is a hyperparameter, and the clip function represents the probability of target clipping. Then, backpropagation is performed to update the actor-new network.

[0138] (3.7) Repeat step (3.6) until the number of steps set according to the actual scenario is reached, and then the loop ends. Use the weights of the actor-new network to update the weights of the actor-old network. Repeat step (3.4) to step (3.6) until the loss function is less than the accuracy ω set according to the actual scenario.

[0139] (3.8) Use the PPO algorithm from steps (3.1) to (3.7) to determine the user's adjusted electricity consumption strategy in the second stage, and use [Q] buy,t Q sell,t ,ΔQ load,t Q bess,tThe composition is as follows: Specifically, in the second phase of the demand-side electricity-carbon coupling market, users bid for carbon emission intensity and electricity volume. After the second phase market clears, users obtain carbon emission intensity, electricity price, and electricity volume, and their electricity purchase and sale strategies [Q]. buy,t Q sell,t This leads to the conclusion.

Claims

1. A virtual carbon storage based demand side electricity-carbon coupling transaction method, characterized in that, The method comprises the following steps: (1) establishing a user dynamic model and a target function; (2) a user participates in a demand-side electricity-carbon coupling market and joins a "virtual carbon storage", so that the user can bid for carbon emission intensity; (3) a model-free, data-driven proximal policy optimization algorithm is used to solve the target function in step (1); Step (1) is specifically: (1.1) first, a user's energy storage model is established; in a regional power system where users are equipped with energy storage and photovoltaic power generation, the dynamic process of user energy storage is represented as: where SOC t denotes the state of charge of the energy storage at the beginning of the time period t, η c and η d denote the charging and discharging efficiency of the energy storage, respectively, and denote the charging and discharging amount of the energy storage in the time period t, respectively; the charging and discharging loss cost of the energy storage is denoted by C bess,t : C bess,t = τ bess | SOC t - SOC t-1 | (2) Wherein, τ bess is a coefficient related to the cost of energy storage charge and discharge loss, |SOC t -SOC t-1 | represents the absolute value of SOC t -SOC t-1 ; (1.2) a regulation cost model is established: C a,t = e t (ΔQ load,t ) 2 + b t ΔQ load,t (3) Wherein, C a,t represents the adjustment cost of the user adjusting the power strategy in the t period, e t and b t are the coefficients of the adjustment cost function, ΔQ load,t represents the size of the user's adjustment power in the t period, and the following constraints are met: wherein is the maximum adjustment amount of the load; (1.3) Establishing the objective function; cost C resulting from user t period electricity purchase buy,t is: C buy,t = τ e,t Q buy,t (5) where τ e,t represents the predicted electricity price for the time period t; the revenue C sell,t is given by: C sell,t = τ e,t Q sell,t (6) Carbon emission cost C caused by user buying electricity at time period t c,t is: C c,t = τ c (Q bess,t + Q load,t ) (7) Where, τ c is the carbon price of the carbon market; the total cost C user,t of the user in the t period is finally obtained C user,t = C bess,t + C a,t + C c,t + C buy,t - C sell,t (8) The final optimization target is to minimize the user's daily operation cost: Step (2) is specifically: (2.1) the user makes a virtual bid in the first stage of the demand-side electricity-carbon coupling market, the buyer's bid is a monotonically decreasing linear function, and the seller's bid is a monotonically increasing linear function; (2.2) the demand-side electricity-carbon coupling market is cleared, and the clearing result is the electricity price and the electricity quantity; only the electricity price data is retained in the first stage market and fed back to the user for reference; (2.3) an electricity quantity balance constraint and a carbon emission balance constraint of the second stage carbon transaction are established; the electricity quantity of the user participating in the market at time t satisfies the balance constraint: Q buy,t +Q pv,t -Q bess,t =Q sell,t +Q load,t (10) wherein Q buy,t represents the electricity purchase amount of the user in the t period, Q pv,t represents the photovoltaic output in the t period, Q bess,t represents the charging and discharging amount of the energy storage in the t period, Q sell,t represents the electricity sale amount of the user in the t period, Q load,t represents the electricity amount consumed by the load of the user after adjusting the electricity use strategy in the t period; and the carbon emission amount coupled with the above electricity amount satisfies the following constraint: Q buy ,tρ buy,t +Q pv ,tρ pv,t -Q bess ,tρ bess,t =Q sell ,tρ sell,t +Q load ,tρ load,t (11) wherein, p buy,t , p pv,t , p bess,t , p sell,t and p load,t are the carbon emission intensities corresponding to the t-period electricity purchase amount, the photovoltaic output, the energy storage charge and discharge amount, the electricity sale amount and the load consumption electricity amount, respectively. (2.4) the carbon emission intensity corresponding to the electricity consumed by the user load at this time is derived from formula (10) and formula (11): When users sell electricity, Q buy,t =0, and the carbon emission intensity of electricity sold by users is equal to the carbon emission intensity of electricity consumed by user loads. Therefore, the carbon emission intensity of electricity sold is: (2.5) a carbon emission flow model of the user's energy storage is established; the carbon emission intensity corresponding to the electricity stored in the energy storage is represented by the following formula: wherein, is the amount of electricity stored in the energy storage at time t, is the carbon intensity of the electricity stored in the energy storage at time t; the change in carbon intensity of the energy storage when charging depends on the carbon intensity of the source of the charge and the current carbon intensity in the energy storage, and the same for discharging. (2.6) Adding "virtual carbon storage", users can bid for carbon intensity; according to formula (13), at this time ρ sell,t Only depends on the variable Q bess,t , that is, the corresponding carbon intensity of the user selling electricity is fixed, resulting in the inability to flexibly transfer carbon emission responsibility among users, limiting carbon emission trading; In order to improve the flexibility of carbon trading, "virtual carbon storage" is introduced from energy storage, which means that under the condition of keeping the carbon emission responsibility transferred from the power generation side to the demand side unchanged, the ρ bess,t of the energy storage becomes a controllable variable, breaking the constraints caused by the proportional sharing theorem; ρ sell,t Flexibility is greatly improved; the carbon intensity of the stored electricity in the energy storage becomes a controllable variable, at which time the energy storage can temporarily store carbon emission responsibility; In order to prevent users from storing carbon intensity beyond their payment capacity, the carbon intensity of the remaining electricity in the energy storage needs to meet the following constraints: (2.7) thanks to the "virtual carbon storage" introduced in step (2.6), the user will bid for electricity quantity and carbon emission intensity in the second stage of the demand-side electricity-carbon coupling market; the buyer's bid is a monotonically decreasing linear function, and the seller's bid is a monotonically increasing linear function. 2.The virtual carbon storage based demand side electric carbon coupling transaction method according to claim 1, characterized in that, The step (3) is specifically: (3.1) first, the state is set as: where τ e,t denotes the predicted electricity price for period t, μ t denotes the basic carbon intensity of electricity in the bid offer, σ in,t and σ out,t denote the variable coefficients describing the carbon intensity of electricity for buying and selling, respectively, in period t, μ t , σ in,t and σ out,t are all dependent on external markets; (3.2) the action space is set, which is represented as: wherein, represents the electricity buying amount of the user in the market at time t, represents the load adjustment amount of the user at time t, represents the charging and discharging amount of the energy storage at time t, represents the decision of the user's bidding price for carbon emission intensity when selling electricity. (3.3) Set a reward, reward function r t represents the user's revenue: where, is the energy storage SOC out-of-limit penalty function, if the state-of-charge of the energy storage exceeds the upper limit or is below the lower limit when updating from state s t to s t+1 ; is the regulation cost offset penalty function, if (3) is not satisfied at state s t ; is the power imbalance penalty function, if (10) is not satisfied at state s t ; in order to limit the influence of future states on the current reward, the future reward needs to be multiplied by a discount factor, and the accumulated reward after discounting is: Wherein, γ is the discount factor; (3.4) input the environment information into the new actor network to obtain the mean μ and variance σ of the two-dimensional normal distribution, and sample to obtain an action a t , and input it into the environment to obtain the reward r and the next state Then store the above results, and input into the actor network, and loop step (3.2); (3.5) The resulting values from the last step of the loop in step (3.4) are input into the critic-NN to obtain the values for all states, and then the discounted rewards in equation (19) are computed, all the stored s are input into the critic-NN to obtain the values for all states, and the loss c loss = mean(square(A t )) is computed, and then the critic-NN is updated by backpropagation. (3.6) combine all the stored s into the original actor network actor-old and actor-new network, respectively, to get normal distribution 1 and normal distribution 2, and input all the stored actions into normal distribution 1 and 2 respectively to get the probability P1 and P2 corresponding to each action, and then divide P2 by P1 to get the importance weight; calculate the loss function where, is the advantage estimate of the advantage function, ε is a hyperparameter, and the clip function represents the probability of target clipping; then backpropagation is performed and the actor-new network is updated; (3.7) loop step (3.6), after reaching the number of steps set according to the actual scene, the loop ends, the weight of the actor-old network is updated with the weight of the actor-new network, and steps (3.4) to (3.6) are looped until the loss function is less than the precision ω set according to the actual scene. (3.8) The PPO algorithm in steps (3.1) to (3.7) is used to obtain the user's power consumption strategy after adjustment in the second stage, which is composed of [Q buy,t , Q sell,t , ΔQ load,t , Q bess,t ]; Specifically, the user's carbon emission intensity and power price in the second stage of the demand-side electricity-carbon coupling market, and the carbon emission intensity, power price and power obtained after the two-stage market clearing, the user's power purchase and sale strategy [Q buy,t , Q sell,t ] are obtained.

3. A computer storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the demand-side electricity-carbon coupling transaction method based on virtual carbon storage according to any one of claims 1-2.

4. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the demand-side electricity-carbon coupling transaction method based on virtual carbon storage according to any one of claims 1-2.

Images