Low-carbon dispatching method for power system based on near-surface carbon dioxide disturbance control

Abstract

This invention discloses a low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control, belonging to the field of energy dispatching technology. It includes constructing an ecological coupling model with the goal of balancing economic costs and environmental disturbances. Based on the ecological coupling model, a Markov decision process for the power system to be dispatched is constructed. The Markov decision process is solved using a dual-delay deep deterministic policy gradient method to obtain the target dispatching strategy. The ecological coupling model includes a carbon flow model, a flexible carbon capture model, an atmospheric diffusion model, a near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level. Through this technical solution, deep coupling is achieved between power system operation, carbon flow regulation, and local environmental disturbance assessment. This enables adaptive optimization of continuous dispatching variables such as thermal power output, carbon capture operation, flue gas split ratio, and energy storage charging and discharging, and improves the strategy stability and dispatching robustness under complex and uncertain environments.

Description

Technical Field

[0001] This invention relates to the field of energy dispatching technology, specifically to a low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control. Background Technology

[0002] As global climate change becomes increasingly severe, the power system is a major source of carbon dioxide emissions. Traditional power system dispatching methods primarily focus on economic efficiency and reliability. In recent years, however, carbon emission cap-and-trade mechanisms have been gradually introduced, forming a low-carbon dispatching model oriented towards total emission reduction. The same total carbon emissions can have significantly different impacts on the near-surface environment under different meteorological diffusion conditions: under unfavorable diffusion conditions, carbon dioxide tends to accumulate near the surface, causing localized high-concentration disturbances that may pose a greater threat to ecosystems and human health. Therefore, managing carbon emissions solely from a total emission perspective is insufficient to accurately guide power system operation and reduce actual environmental disturbances.

[0003] Currently, in dispatching studies of new energy bases in desert and Gobi regions, the flue gas diversion ratio of carbon capture power plants is typically optimized based on internal system operating costs or carbon quota prices, lacking a dynamic response to post-emission atmospheric diffusion conditions and near-surface environmental conditions. In other words, an effective synergistic mechanism has not yet been established between carbon capture dispatching and environmental disturbance control. This results in an inability to proactively enhance capture to alleviate near-surface accumulation under unfavorable meteorological conditions, while potentially leading to over-capture and unnecessary economic costs under favorable diffusion conditions. Furthermore, power system dispatching faces significant uncertainties in wind power, photovoltaic output, and meteorological conditions. Traditional optimization methods often suffer from high computational complexity, sensitivity to model parameters, and difficulty in online adaptive adjustment when dealing with high-dimensional continuous action spaces, cross-temporal coupling, and multi-source uncertainties. In recent years, deep reinforcement learning methods have shown potential in the field of power system dispatching, but existing applications are mostly concentrated on economic dispatching or voltage control, which is insufficient to meet the stability and robustness requirements under complex and uncertain environments. Summary of the Invention

[0004] The purpose of this invention is to provide a low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control, so as to solve the technical problems existing in related technologies.

[0005] To achieve the above objectives, the present invention provides a low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control, comprising:

[0006] An ecological coupling model for the power system to be dispatched is constructed. The ecological coupling model includes a carbon flow model, a flexible carbon capture model, an atmospheric diffusion model, a near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level. The carbon flow model is used to calculate the carbon dioxide emissions generated by thermal power units in the power system. The flexible carbon capture model is used to calculate the carbon dioxide emissions generated by carbon capture units in the power system. The atmospheric diffusion model is used to calculate the near-surface carbon dioxide concentration caused by plume diffusion in the power system. The near-surface net disturbance concentration field model is used to calculate the sum of the near-surface carbon dioxide concentration changes caused by the plume and the near-surface carbon dioxide concentration changes caused by the ecosystem.

[0007] With the goal of balancing economic costs and environmental disturbances, a Markov decision process for the power system to be dispatched is constructed based on an ecological coupling model. The Markov decision process includes the state space, action space, and reward function of the power system to be dispatched.

[0008] The Markov decision process is solved by using the double-delay deep deterministic policy gradient method to obtain the target scheduling policy.

[0009] Optionally, the ecological coupling model is constructed using the following method:

[0010] Calculate the output of thermal power units in the power system to be dispatched, and construct a carbon flow model based on the output of thermal power units;

[0011] Calculate the total generating power of carbon capture units in the power system to be dispatched, and construct a flexible carbon capture model based on the total generating power;

[0012] The carbon dioxide emission rate is calculated based on the carbon flow model and the flexible carbon capture model, and an atmospheric diffusion model is constructed based on the carbon dioxide emission rate.

[0013] Calculate the near-surface carbon dioxide concentration caused by the power system, and construct a near-surface net disturbance concentration field model based on carbon dioxide concentration and atmospheric diffusion model;

[0014] The upper tail disturbance intensity and average exposure level were constructed based on the near-surface net disturbance concentration field model.

[0015] The carbon flow model, flexible carbon capture model, atmospheric diffusion model, near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level were identified as the ecological coupling model.

[0016] Optionally, the construction of the Markov decision process for the power system to be dispatched based on the ecological coupling model includes:

[0017] Construct a state space based on the operating parameters of the power system to be dispatched;

[0018] The action space is determined based on the state space and preset exploration noise;

[0019] Based on the upper tail disturbance intensity and average exposure level in the ecological coupling model, an environmental objective function is constructed, and an economic objective function is constructed based on the operating costs of thermal power, carbon trading, carbon capture operation, renewable energy curtailment, and energy storage system in the power system to be dispatched. The reward function is then calculated based on the environmental and economic objective functions.

[0020] The state space, action space, and reward function are defined as a Markov decision process.

[0021] Optionally, the state space includes wind turbine output, photovoltaic unit output, total load of the power system to be dispatched, energy storage of the energy storage system, solvent volume of the rich liquid storage tank, carbon dioxide exchange flux, solvent volume of the amine liquid storage tank in the carbon capture system, total output of the thermal power unit, and boundary layer height. The operational space includes the output of thermal power units, flue gas split ratio, net output power of carbon capture units, operating energy consumption of carbon capture units, and output of energy storage systems. The reward function is expressed by the following formula:

[0022] ;

[0023] in, For the power system to be dispatched in the first Rewards obtained by the agent through interaction with the environment during a given time period. As a scalar, To balance the preference parameters between economic and environmental goals, Let the economic objective function be... This is the environmental objective function.

[0024] Optionally, the environmental objective function is obtained by the following method: The first value is obtained by processing the upper tail disturbance intensity of the scheduling system by a linear normalization operator based on the reference value; The second value is obtained by processing the average exposure level of the system to be scheduled using a linear normalization operator based on the reference value; Construct an environmental objective function based on the first and second values; The economic objective function is obtained through the following method: Determine the operating costs of thermal power plants, carbon trading costs, carbon capture operation costs, renewable energy curtailment costs, and energy storage system operation costs in the system to be dispatched; The total operating cost is calculated based on the operating costs of thermal power plants, carbon trading costs, carbon capture operation costs, renewable energy curtailment costs, and energy storage system operation costs. The total operating cost is then processed using a linear normalization operator based on reference values ​​to obtain the economic objective function.

[0025] Optionally, the intensity of the upper tail perturbation is expressed by the following formula:

[0026] ;

[0027] in, The intensity of the upper tail perturbation. To schedule the current time period Internal and effective diffusion domain The values ​​at the 95th percentile after sorting all grid points by net disturbance concentration values ​​from smallest to largest. For time period The effective diffusion domain within;

[0028] The average exposure level is expressed by the following formula:

[0029] ;

[0030] in, For average exposure levels, This represents the number of grid cells within the effective diffusion domain.

[0031] Optionally, the constraints of the ecological coupling model include energy storage system constraints, energy conservation constraints, thermal power unit constraints, carbon capture system constraints, and power balance constraints.

[0032] The above technical solution aims to balance economic costs and environmental disturbances. Based on the ecological coupling model of the power system to be dispatched, a Markov decision process for the power system is constructed. The Markov decision process is then solved using a dual-delay deep deterministic policy gradient method to obtain the target dispatch policy. This ecological coupling model includes a carbon flow model, a flexible carbon capture model, an atmospheric diffusion model, a near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level. It enables deep coupling of power system operation, carbon flow regulation, and local environmental disturbance assessment. This allows for adaptive optimization of continuous dispatch variables such as thermal power output, carbon capture operation, flue gas split ratio, and energy storage charging and discharging, and improves policy stability and dispatch robustness under complex and uncertain environments.

[0033] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0034] Figure 1This is a schematic diagram illustrating a low-carbon dispatching method for a power system based on near-surface carbon dioxide disturbance control according to an exemplary embodiment of the present invention.

[0035] Figure 2 This is a schematic diagram of the carbon cycle according to an exemplary embodiment of the present invention.

[0036] Figure 3 This is a schematic diagram illustrating the solution based on a dual-delay deep deterministic strategy gradient algorithm according to an exemplary embodiment of the present invention.

[0037] Figure 4 This is a comparison chart illustrating the training effect of a scheduling control model according to an exemplary embodiment of the present invention.

[0038] Figure 5 This is a schematic diagram illustrating the power balance results under a calm wind stable scenario according to an exemplary embodiment of the present invention.

[0039] Figure 6 This is a schematic diagram illustrating the results of two indicators under a wind condition change scenario according to an exemplary embodiment of the present invention. Detailed Implementation

[0040] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, so as to provide a better understanding of the concept of the present invention, the technical problem solved, the technical features constituting the technical solution, and the technical effects brought about.

[0041] like Figure 1 As shown, Figure 1 This is a schematic diagram illustrating a low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control according to an exemplary embodiment of the present invention. (Refer to...) Figure 1 The method includes;

[0042] S101: Construct an ecological coupling model for the power system to be dispatched. The ecological coupling model includes a carbon flow model, a flexible carbon capture model, an atmospheric diffusion model, a near-surface net disturbance concentration field model, an upper-tail disturbance intensity, and an average exposure level. The carbon flow model is used to calculate the carbon dioxide emissions generated by thermal power units in the power system. The flexible carbon capture model is used to calculate the carbon dioxide emissions generated by carbon capture units in the power system. The atmospheric diffusion model is used to calculate the near-surface carbon dioxide concentration caused by plume diffusion in the power system. The near-surface net disturbance concentration field model is used to characterize the sum of the near-surface carbon dioxide concentration changes caused by the plume and the near-surface carbon dioxide concentration changes caused by the ecosystem.

[0043] S102: With the goal of balancing economic costs and environmental disturbances, a Markov decision process for the power system to be dispatched is constructed based on an ecological coupling model. The Markov decision process includes the state space, action space, and reward function of the power system to be dispatched.

[0044] S103: The Markov decision process is solved by the double-delay deep deterministic policy gradient method to obtain the target scheduling policy.

[0045] The above technical solution aims to balance economic costs and environmental disturbances. Based on the ecological coupling model of the power system to be dispatched, a Markov decision process for the power system is constructed. The Markov decision process is then solved using a dual-delay deep deterministic policy gradient method to obtain the target dispatch policy. This ecological coupling model includes a carbon flow model, a flexible carbon capture model, an atmospheric diffusion model, a near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level. It enables deep coupling of power system operation, carbon flow regulation, and local environmental disturbance assessment. This allows for adaptive optimization of continuous dispatch variables such as thermal power output, carbon capture operation, flue gas split ratio, and energy storage charging and discharging, and improves policy stability and dispatch robustness under complex and uncertain environments.

[0046] To enable those skilled in the art to better understand the low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control provided by this invention, the above steps are illustrated in detail below.

[0047] For example, the power system to be dispatched can be an energy renewable system, which may include generating units such as thermal power units, carbon capture units, wind power units, and photovoltaic units. An ecological coupling model can be used to characterize the carbon cycle in the power system to be dispatched, specifically as follows: Figure 2 As shown. Figure 2 In this process, carbon emissions occur during the operation of the power system to be dispatched. The total carbon emissions are calculated using a carbon flow model, while a flexible carbon capture model is used to calculate how much carbon is captured by the carbon capture device. The remaining emitted carbon diffuses directly into the surrounding environment of the power system to be dispatched. The near-surface carbon dioxide concentration caused by the plume diffusion can be calculated using an atmospheric diffusion model. The concentration increment caused by atmospheric diffusion and the equivalent concentration change caused by ecosystem exchange are linearly superimposed at the same spatial coordinate point to obtain the final net near-surface carbon dioxide disturbance concentration field.

[0048] In one possible manner, the ecological coupling model is constructed by the following method:

[0049] Calculate the output of thermal power units in the power system to be dispatched, and construct a carbon flow model based on the output of thermal power units;

[0050] Calculate the total generating power of carbon capture units in the power system to be dispatched, and construct a flexible carbon capture model based on the total generating power;

[0051] The carbon dioxide emission rate is calculated based on the carbon flow model and the flexible carbon capture model, and an atmospheric diffusion model is constructed based on the carbon dioxide emission rate.

[0052] Calculate the near-surface carbon dioxide concentration caused by the power system, and construct a near-surface net disturbance concentration field model based on carbon dioxide concentration and atmospheric diffusion model;

[0053] The upper tail disturbance intensity and average exposure level were constructed based on the near-surface net disturbance concentration field model.

[0054] The carbon flow model, flexible carbon capture model, atmospheric diffusion model, near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level were identified as the ecological coupling model.

[0055] It should be understood that when constructing the carbon flow model, it can be based on the output of thermal power units in the power system to be dispatched. Specifically, considering only direct CO2 emissions on the dispatch timescale, thermal power units are therefore regarded as the sole carbon source in the carbon flow model. The carbon flow model is expressed by the following formula:

[0056] ;

[0057] in, For the first Taiwan thermal power units during the period Carbon dioxide emissions within the region. This is the carbon dioxide emission intensity coefficient. For thermal power units in the system to be dispatched In the Efforts during specific time periods.

[0058] When constructing a flexible carbon capture model, it can be based on the total power generation of the carbon capture units. Specifically, the flexible carbon capture model is expressed by the following formula:

[0059] ;

[0060] in, For the system to be scheduled in the first Energy consumption of carbon capture unit during time period. For carbon capture units in the first Total power generation during the period For the system to be scheduled in the first Net output power of the carbon capture unit during the time period. This refers to the fixed energy consumption of the carbon capture unit.

[0061] The overall carbon emission level of carbon capture units is related to the different power generation outputs of the carbon capture units, as shown in the following formula:

[0062] ;

[0063] in, For carbon capture units in the first Carbon dioxide emissions generated during the period. This is the carbon dioxide emission intensity coefficient of the carbon capture unit.

[0064] The expression that distinguishes between the carbon dioxide being processed by the absorption tower and the carbon dioxide being processed by the regeneration tower is:

[0065]

[0066] in, Energy consumption per unit of carbon dioxide capture. For the system to be scheduled in the first The amount of carbon dioxide emissions processed during the period. For carbon dioxide capture efficiency. The flue gas split ratio, For the carbon capture units in the system to be dispatched in the first Carbon dioxide emissions during a given period.

[0067] Carbon dioxide exists in the form of a compound in an alcoholamine solution, and the following relationship holds:

[0068] ;

[0069] in, for Carbon dioxide supplied by the periodic rich liquid storage tank. In the first The solvent volume that shifts from a rich solvent to a poor solvent during a given period. This represents the molar mass of the ethanolamine solution. Here is the molar mass of carbon dioxide. For the regeneration resolution coefficient, This represents the mass percentage concentration of ethanolamine. This is the density of the ethanolamine solution.

[0070] When constructing an atmospheric diffusion model, the carbon dioxide emission rate can be calculated based on the carbon dioxide emissions calculated by the carbon flow model and the flexible carbon capture model, and the model can be constructed based on the carbon dioxide emission rate. Specifically, to simplify diffusion modeling, the emissions from multiple thermal power units are aggregated and equivalently represented as emissions from a single common chimney. To ensure consistency between the diffusion model input and the emission output obtained from scheduling on both unit and time scales, the atmospheric diffusion model is expressed by the following formula:

[0071] ;

[0072] ;

[0073] in, For carbon dioxide emission rate. This represents the total number of thermal power units in the power system to be dispatched. For the first Duration of the period This refers to the near-surface carbon dioxide concentration. For time period average wind speed, and All are diffusion parameters. The horizontal distance is perpendicular to the wind direction. For effective source height, , For diffusion parameter coefficients, downwind distance Power of 1 downwind distance Power of 1.

[0074] When constructing a near-surface net disturbance concentration field model, it can be based on the power system's model of near-surface carbon dioxide concentration and atmospheric diffusion. Specifically, under the assumption of a fully mixed boundary layer, the accumulated ecosystem CO2 exchange during each scheduling period is converted into an equivalent concentration change by uniformly distributing this exchange amount within the boundary layer height, as shown in the following equation:

[0075] ;

[0076] The near-surface net disturbance concentration field model is defined as the sum of the concentration changes caused by the plume and the concentration changes caused by the ecosystem, as shown in the following equation:

[0077] ;

[0078] in, The near-surface carbon dioxide concentration is determined by the ecosystem within the power system to be dispatched. For position In the Carbon dioxide exchange flux during a given period. The boundary layer height is from ERA5. For the time period Internal, spatial location ground height The total change in CO2 concentration;

[0079] Subsequently, the upper tail disturbance intensity and average exposure level can be constructed based on the near-surface net perturbation concentration field model. Specifically, to quantify the spatiotemporal disturbance of the NNPCF (Near-surface Net Perturbation Concentration Field), the effective diffusion domain is used as the basis. On Two environmental indicators were constructed. These two indicators correspond to the intensity of the upper tail perturbation, respectively. and average exposure level The upper tail disturbance intensity is used to characterize the potential short-term accumulation characteristics within the plume's influence area, while the average exposure level is used to characterize the 95th percentile of the NNPCF, i.e., the upper tail disturbance intensity, as defined in the following formula:

[0080] To characterize the potential short-term accumulation characteristics within the plume's influence area, the intensity of the upper tail disturbance is expressed by the following formula:

[0081] ;

[0082] in, The intensity of the upper tail perturbation. To schedule the current time period Internal and effective diffusion domain The values ​​at the 95th percentile after sorting all grid points by net disturbance concentration values ​​from smallest to largest. For time period The effective diffusion domain within; this index reflects the intensity of the disturbance associated with the most severely affected part within this domain; compared to the maximum value, It has low sensitivity to isolated anomalous grid cells, can more robustly characterize hotspot conditions, and can objectively reflect the actual severity of high-concentration clusters.

[0083] To reflect the average level of carbon dioxide exposure borne by the ecosystem within the domain, the average exposure level is expressed by the following formula:

[0084] ;

[0085] in, For average exposure levels, The number of grid cells within the effective diffusion domain is used as the spatial average of NNPCF. This metric may be diluted when the domain expands to include low-perturbation boundary cells. Therefore, Need and They are used together to jointly characterize the overall degree of disturbance.

[0086] Finally, the carbon flow model, flexible carbon capture model, atmospheric diffusion model, near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level can be identified as the ecological coupling model.

[0087] For example, the Markov decision process can serve as a unified mathematical modeling framework for all reinforcement learning algorithms, describing the mathematical model of agent-environment interaction and sequential decision-making. In this embodiment of the invention, with the goal of balancing economic costs and environmental disturbances, the spatiotemporal characteristics of carbon dioxide disturbances can be characterized, and a Markov decision process for the power system to be dispatched can be constructed based on an ecological coupling model.

[0088] In one possible manner, the construction of the Markov decision process for the power system to be dispatched, based on the ecological coupling model, includes:

[0089] Construct a state space based on the operating parameters of the power system to be dispatched;

[0090] The action space is determined based on the state space and preset exploration noise;

[0091] Based on the upper tail disturbance intensity and average exposure level in the ecological coupling model, an environmental objective function is constructed, and an economic objective function is constructed based on the operating costs of thermal power, carbon trading, carbon capture operation, renewable energy curtailment, and energy storage system in the power system to be dispatched. The reward function is then calculated based on the environmental and economic objective functions.

[0092] The state space, action space, and reward function are defined as a Markov decision process.

[0093] It should be understood that the state space can be used to summarize the key operating conditions required for scheduling, including wind turbine output, photovoltaic unit output, total load of the power system to be dispatched, energy storage capacity of the energy storage system, solvent volume of rich liquid storage tanks, carbon dioxide exchange flux, solvent volume of amine liquid storage tanks in the carbon capture system, total output of thermal power units, and boundary layer height. Specifically, it can be expressed by the following calculation formula:

[0094] ;

[0095] in, In the first The state space of a time period In the first Wind turbine output during different time periods In the first Photovoltaic unit output during a certain period of time In the first The total load of the power system to be dispatched during the specified time period. In the first The energy stored in the time-of-use energy storage system In the first Solvent volume in the rich liquid storage tank during the time period For position In the Carbon dioxide exchange flux during a given period. For time period The solvent volume of the amine storage tank in the carbon capture system. In the first Total output of thermal power units during the period The boundary layer height is derived from ERA5 data;

[0096] After observing the system's state space information, the agent selects an action from the action space based on the policy function and exploration noise. The exploration noise can be set by relevant operators, and the agent can be the decision-making core in deep reinforcement learning, i.e., an agent trained based on the TD3 algorithm. The action is then input into the model as the system's power value. The action space includes the output of thermal power units, flue gas split ratio, net output power of carbon capture units, operating energy consumption of carbon capture units, and output of energy storage systems, and can be expressed by the following formula:

[0097] ;

[0098] in, For the power system to be dispatched in the first The space of action during a period of time For thermal power units in the system to be dispatched In the Efforts during a specific time period The flue gas split ratio, For the system to be scheduled in the first Net output power of the carbon capture unit during the time period. For the system to be scheduled in the first Energy consumption of carbon capture unit during time period. For the system to be scheduled in the first Output of the time-limited energy storage system.

[0099] The reward function can be used to characterize the agent's performance. Only take action in a certain state Timely reporting is obtained, consistent with the objective function of the system. When calculating the reward function, it can be based on both the environmental and economic objective functions. Specifically, the environmental objective function can be calculated based on the intensity of the upper tail disturbance and the average exposure level, while the economic objective function can be constructed based on the operating costs of thermal power plants, carbon trading costs, carbon capture and storage operations, renewable energy curtailment costs, and energy storage system operations.

[0100] Specifically, to eliminate dimensional differences and ensure parameter interpretability, both the environmental and economic objective functions are normalized using a scaling method based on reference values, which can be expressed by the following formula:

[0101] ;

[0102] In the formula, As an indicator The reference scale is determined by using the traditional low-carbon economic dispatch solution as an unbiased benchmark.

[0103] The environmental objective function is obtained through the following method: The first value is obtained by processing the upper tail disturbance intensity of the scheduling system by a linear normalization operator based on the reference value; The second value is obtained by processing the average exposure level of the system to be scheduled using a linear normalization operator based on the reference value; An environmental objective function is constructed based on the first and second numerical values. This objective function can be expressed by the following formula:

[0104] ;

[0105] in, For the environmental objective function, For a linear normalization operator based on reference values, For the system to be scheduled in the first The intensity of the upper tail disturbance during the time period. For the system to be scheduled in the first Average exposure level over a period of time.

[0106] The economic objective function is defined based on total operating cost, and is obtained through the following method: Determine the operating costs of thermal power plants, carbon trading costs, carbon capture operation costs, renewable energy curtailment costs, and energy storage system operation costs in the system to be dispatched; The total operating cost is calculated based on the operating costs of thermal power plants, carbon trading costs, carbon capture and emission reduction costs, renewable energy curtailment costs, and energy storage system operating costs. This total operating cost is then processed using a linear normalization operator based on reference values ​​to obtain the economic objective function, which can be expressed as follows:

[0107] ;

[0108] in, Total operating cost, For the thermal power units in the waiting dispatch system in the first Operating costs of thermal power plants during different time periods For the system to be scheduled in the first Carbon trading costs for different time periods For the system to be scheduled in the first Carbon capture operation costs over time period For the system to be scheduled in the first The cost of renewable energy curtailment during a given period For the system to be scheduled in the first Operating costs of energy storage systems over time periods This is the economic objective function.

[0109] Specifically, the operating cost of the thermal power plant is expressed by the following formula:

[0110] ;

[0111] in, For thermal power plant operating costs, For fuel costs, For start-stop costs, , and All are fuel cost coefficients. This refers to the collection of all thermal power units and carbon capture units in the power system to be dispatched. For thermal power units In the The on / off status during a certain period. For thermal power units Cost per startup For thermal power units Cost per downtime For thermal power units In the The on / off status during a given time period;

[0112] To characterize the nonlinear marginal pricing effect in the carbon market, this invention employs a tiered carbon trading mechanism, the carbon trading cost of which is expressed by the following formula:

[0113] ;

[0114] in, For carbon trading costs, In the first Emissions exceeding carbon allowances during the specified period. The length of the carbon emission ladder interval. For carbon trading prices, For the carbon trading price growth rate, In the first Carbon emission allowances for the specified period In the first Total carbon dioxide emissions during the period. This is the carbon quota allocation coefficient. This represents the total number of thermal power units in the power system to be dispatched.

[0115] The carbon capture operation cost is expressed by the following formula:

[0116] ;

[0117] in, For carbon capture operation costs, In the first Solvent unit cost coefficient within the time period In the first Costs of transporting and storing carbon dioxide over a given period of time. The solvent loss coefficient is... The unit price of MEA solvent. For the unit cost of transportation and storage, For the system to be scheduled in the first The amount of carbon dioxide emissions processed during the period;

[0118] The cost of renewable energy curtailment is expressed by the following formula:

[0119] ;

[0120] in, For the cost of curtailing renewable energy, The unit cost of power curtailment In the first Power wasted during the period;

[0121] The operating cost of the energy storage system is expressed by the following formula:

[0122] ;

[0123] in, For the operating costs of energy storage systems, The unit cost of operating an energy storage system. In the first The charging power of the energy storage system during the time period, In the first The discharge power of the energy storage system during the time period.

[0124] The reward function can then be expressed by the following formula:

[0125] ;

[0126] in, For the power system to be dispatched in the first Rewards obtained by the agent through interaction with the environment during a given time period. As a scalar, To balance the preference parameters between economic and environmental goals, Let the economic objective function be... This is the environmental objective function.

[0127] For example, such as Figure 3 As shown, after obtaining the Markov decision process, the target scheduling policy can be obtained by solving the Markov decision process using the Twin-Delayed Deep Deterministic Policy Gradient (TD3) method. The Twin-Delayed Deep Deterministic Policy Gradient is an offline, deterministic, continuous action deep reinforcement learning algorithm based on the Actor-Critic framework. The constructed Markov decision process contains continuous control variables, making the actor-critic method more applicable than the value function method based on discrete action sets. In the current problem, the coupled dynamics of the operational environment and the weighted rewards of the economic environment may lead to high-variance learning signals. To improve learning stability and robustness, this embodiment of the invention employs the TD3 algorithm (TwinDelayed Deep Deterministic Policy Gradient). By integrating a dual-critic network, a delayed policy update mechanism, and a target policy smoothing technique, the TD3 algorithm effectively alleviates overestimation bias and policy oscillation, thus providing a stable guarantee for continuous control policy learning in adaptive scheduling optimization.

[0128] Specifically, such as Figure 3 As shown, with a daily scheduling cycle and one hour as the time step for the daily scheduling task, the solution process is as follows: First, initialize the network parameters. The TD3 algorithm includes a policy network, a value network, and their corresponding target policy network and target value network; the network parameters are as follows: Policy network parameters Parameters of Value Network 1 Parameters of Value Network 2 Target policy network parameters Parameters of Target Value Network 1 Parameters of Target Value Network 2 Initialize the above network parameters and initialize the experience replay buffer.

[0129] Secondly, in the ecological coupling model, the agent acts according to the current strategy. With exploring noise selection action This action drives the coupled system to evolve to the next state. The environment also calculates immediate rewards based on economic costs and environmental disturbance indicators. This reflects the long-term comprehensive benefits of the action rather than just the instantaneous gains, and will The samples are stored in an experience replay buffer, which holds samples from different time steps to break the temporal correlation, and are then used for subsequent small-batch sampling training.

[0130] Then, random sampling is performed from the experience playback buffer. Batch data The next action is generated through the target policy network. ,in Smooth the noise to the target; It is a cutoff function that limits the noise value to the interval middle, This indicates that the mean is 0 and the standard deviation is 0. The normal distribution; These are the upper and lower limits for truncation;

[0131] The target value is calculated using the target value network, as shown in the following formula:

[0132] ;

[0133] In the formula, For the first Target value for the time period; In the state Take action below The reward value obtained; Discount factor; These are the parameters for the first and second target value networks; For the first and second target value networks in state Take action below Target value;

[0134] Minimize the total loss function of the value network To update the network parameters, the model is as follows:

[0135] ;

[0136] ;

[0137] In the formula, , and They represent the first The status, actions, and target values ​​of each batch; For the first Parameters of a value network; For value network hyperparameters; The gradient of the total loss function;

[0138] Following this, the policy network parameters are updated every few steps according to the delayed update mechanism. The policy is improved by maximizing the target value network's evaluation of the action, as modeled below:

[0139] ;

[0140] ;

[0141] in, To determine the expected reward that can be obtained by taking the corresponding action distribution. The corresponding gradient; Indicates the state Calculate the gradient of the policy network below; Representing state With action The gradient of the value master network 1 is then calculated. These are the hyperparameters of the policy network.

[0142] The target network parameters are updated incrementally using a soft update strategy, as modeled below:

[0143] ;

[0144] In the formula, This is the soft update coefficient, used to ensure network stability through a soft update strategy.

[0145] Finally, through continuous interaction, sampling, and training, until the system scheduling target converges or the number of training rounds is reached, the optimal target scheduling strategy is obtained. The optimal target scheduling strategy may include the scheduling action sequence of thermal power unit output, flue gas split ratio, net output of carbon capture unit, output of regeneration tower and energy storage, as well as the operating cost, CO2 emissions, near-surface CO2 disturbance index and power balance of the power system to be scheduled.

[0146] Among the possible approaches, the constraints of the ecological coupling model include energy storage system constraints, energy conservation constraints, thermal power unit constraints, carbon capture system constraints, and power balance constraints.

[0147] The constraints of energy storage systems are expressed by the following formula:

[0148] ;

[0149] in, In the first The charging power of the energy storage system during the time period, In the first The charging status of the energy storage system during the time period. Maximum charging power, In the first The discharge power of the energy storage system during the time period In the first The discharge status of the energy storage system during the time period. For maximum discharge power, The total energy storage capacity of the energy storage system. This represents the lower limit of the energy storage capacity ratio. In the first The energy stored in the time-of-use energy storage system This represents the upper limit of the energy storage capacity ratio;

[0150] The energy conservation constraint is expressed by the following formula:

[0151] ;

[0152] in, For energy storage systems in the first Energy storage over a period of time Self-discharge rate For charging and discharging efficiency, For the first The duration of the period;

[0153] The constraints of thermal power units are expressed by the following formula:

[0154] ;

[0155] in, For thermal power units The minimum technical output, For thermal power units The maximum technical output, For thermal power units In the The on / off status during a certain period. This is the downhill climbing speed limit. This is the limit for uphill speed. For thermal power units In the Efforts during a specific time period For shutdown indicators, For thermal power units In the The on / off status during a certain period. To initiate the indicator, For thermal power units Cumulative power-on time before the state transition Minimum boot time, Minimum downtime;

[0156] Based on the constraints of conventional thermal power units, the CCPP model also considers the dynamic characteristics and capacity limitations of solvent storage tanks. The constraints of the carbon capture system are expressed by the following calculation formula:

[0157] ;

[0158] in, In the first The volume of solvent in the rich solution storage tank during a given period. In the first The solvent volume that shifts from a rich solvent to a poor solvent during a given period. In the first The volume of solvent in the rich solution storage tank during a given period. In the first The volume of solvent in the lean solution storage tank during a given period. In the first The volume of solvent in the lean solution storage tank during a given period. This represents the volume of solvent in the solvent-rich tank during the initial period. This represents the solvent volume in the solvent-rich tank during the final stage. For the maximum capacity of the storage tank, This represents the volume of solvent in the lean solvent tank during the initial period. This represents the volume of solvent in the lean solvent tank during the final stage.

[0159] The power balance constraint is expressed by the following formula:

[0160] ;

[0161] in, In the first The output of the wind turbines during the period In the first The output of the photovoltaic units during the period In the first Power curtailment during the time period In the first The total load of the power system to be dispatched during the specified time period.

[0162] like Figure 4 , Figure 5 and Figure 6The figure shows the results of the operation of the low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control proposed in this invention.

[0163] The algorithm training of this invention is implemented based on the PyTorch framework, using Python 3.8 as the programming language, and the hardware platform configuration is an Intel Core i7-13650HX processor and an NVIDIA GeForce RTX 4060 graphics card. Regarding the algorithm's hyperparameter settings, the maximum number of training epochs for the TD3 algorithm is set to 2000, the soft update coefficient is 0.005, the batch size is 256, and the experience replay buffer capacity is 10000. The learning rates for the actor network and the critic network are set to 0.003 and 0.0003, respectively, to ensure stable convergence during the policy evaluation and policy improvement processes.

[0164] To verify the effectiveness and performance of the proposed TD3 algorithm, it was compared with the DDPG (Deep Deterministic Policy Gradient) algorithm. All algorithms were trained on the same system optimization scheduling model and compared under the same training environment and hyperparameter configuration. The change curves of the reward function during training are shown below. Figure 4 As shown.

[0165] In terms of convergence speed, the DDPG algorithm exhibits the fastest initial convergence rate, stabilizing around 1000 training rounds. However, its final convergence value is significantly lower than the other two algorithms, indicating relatively poor performance. The TD3 algorithm converges slightly slower than DDPG, reaching a stable phase after approximately 2000 training rounds. However, its convergence value is significantly improved, demonstrating superior policy optimization capabilities. The TD3 algorithm shows the most rapid increase in its reward function during the early stages of training, stabilizing around 1000 rounds. Its convergence speed is significantly better than the DDPG algorithm, and its final reward value is also higher, indicating that it can learn a more effective scheduling policy. Regarding convergence stability, although the DDPG algorithm converges quickly, it exhibits large fluctuations after reaching a steady state, resulting in poor stability. The TD3 algorithm shows a significant improvement in stability compared to DDPG.

[0166] The TD3 algorithm is an optimization of the DDPG algorithm. By introducing a dual-Q network, it adds a delayed policy update mechanism and a target policy smoothing mechanism. Therefore, it is much better than the DDPG algorithm in terms of convergence speed and stability. Moreover, the TD3 algorithm has smaller variance in the later training stage.

[0167] Based on the above optimal parameters, the power balance results during the algorithm's solution process are as follows: Figure 5 As shown. Figure 5The figure presents the power balance results under a calm, stable wind scenario. As can be seen from the figure, wind power, photovoltaic (PV), thermal power units, carbon capture units, and energy storage systems (energy storage discharge and energy storage charging) can work together to meet the system load demand. The load curves are largely consistent with the stacking results of various power sources and energy storage outputs, indicating that the proposed dispatch strategy can maintain system power balance under different diffusion conditions. Wind power output accounts for a higher proportion at night and evening, especially around 0:00–7:00 and 17:00–23:00, providing the main renewable energy power support for the system. PV output is mainly concentrated between 9:00–16:00, reaching a high level at noon, exhibiting typical daytime concentrated output characteristics. Conventional thermal power units and carbon capture units, as the main adjustable power sources in the system, increase their output during periods of insufficient wind and solar power output or high load to ensure power supply reliability; while during the noon period when PV output is high, their output levels decrease accordingly, thus freeing up space for renewable energy consumption. Energy storage systems charge during periods of surplus renewable energy and discharge during periods of high load or insufficient renewable energy output, thus playing a role in peak shaving, valley filling, and transferring renewable energy power across time periods. In both scenarios, wind power, solar power, and load input remain consistent, but the output allocation of thermal power, carbon capture units, and energy storage systems differs, indicating that the introduction of environmental targets alters the system's allocation of flexible resources. Overall, Figure 5 This demonstrates that the proposed environmentally sensitive low-carbon scheduling method can coordinate the operation of new energy sources, thermal power, carbon capture units, and energy storage systems while meeting power balance constraints, providing a feasible operational basis for subsequent near-ground CO2 disturbance control.

[0168] Based on the feasibility of power balance Figure 6 The results for two indicators under the wind condition change scenario are further presented: the results for changes in meteorological conditions and environmental disturbance indicators. Among them, Figure 6 The intensity of the upper tail perturbation is given respectively. and average exposure level The changes. From Figure 6 It can be seen that around 12:00–18:00, strong horizontal transport capacity and a deep boundary layer are conducive to CO2 diffusion and dilution. Therefore, during this period, both environmental indicators remain at low levels, and the difference between Case 2 and Case 3 is small, indicating that under favorable diffusion conditions, near-surface CO2 disturbances are inherently weak, and the marginal improvement space for environmental awareness and scheduling is limited. Conversely, during 20:00–23:00, wind speed decreases significantly, and the boundary layer height decreases rapidly, significantly weakening atmospheric dilution capacity, leading to easier CO2 accumulation near the ground. Affected by this, under Case 2… and All showed a rapid increase, peaking around 22:00, indicating strong local disturbances and average exposure risk during this period. Compared to Case 2, Case 3 significantly reduced both types of environmental indicators during the unfavorable diffusion period of 20:00–23:00, with particularly noticeable reductions around 21:00 and 22:00. This demonstrates that the proposed method can proactively adjust carbon capture operation strategies based on changes in wind speed and boundary layer height, strengthening CO2 capture and emission control during periods of higher environmental risk. Therefore, Figure 6 This not only demonstrates that near-surface CO2 disturbances have significant meteorological sensitivity, but also verifies that environmental awareness-based scheduling strategies can focus on controlling unfavorable diffusion periods, thereby achieving a more effective near-surface disturbance suppression effect than simple low-carbon economic scheduling.

[0169] As demonstrated by the above embodiments, this invention proposes a low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control. This method constructs a low-carbon dispatching framework for the coordinated operation of wind power, photovoltaic, thermal power units, carbon capture units, and energy storage systems. It further introduces atmospheric diffusion processes and ecosystem CO2 exchange mechanisms to establish a near-surface net disturbance concentration field, achieving deep coupling between power system operation, carbon flow regulation, and local environmental disturbance assessment. Simultaneously, this invention further couples a Gaussian plume diffusion model with the ecosystem CO2 exchange process to construct a near-surface net disturbance concentration field, using the upper tail intensity of the disturbance and the average exposure level as environmental assessment indicators, thus expanding the dispatching objective from traditional total carbon emission control to near-surface environmental disturbance control. In terms of algorithm implementation, the TD3 deep reinforcement learning algorithm is used to adaptively optimize continuous dispatching variables such as thermal power output, carbon capture operation, flue gas split ratio, and energy storage charging and discharging, improving the strategy stability and dispatching robustness under complex and uncertain environments. The results show that this invention can maintain system power balance under different meteorological diffusion conditions and effectively suppress the risk of near-surface CO2 accumulation during unfavorable diffusion periods such as weak winds and low boundary layer heights, while maintaining good operational economics. This method provides a feasible technical solution for the flexible utilization of carbon capture resources, proactive control of environmental risks, and low-carbon economic dispatch in high-proportion renewable energy power systems.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control, characterized in that, include: An ecological coupling model for the power system to be dispatched is constructed. The ecological coupling model includes a carbon flow model, a flexible carbon capture model, an atmospheric diffusion model, a near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level. The carbon flow model is used to calculate the carbon dioxide emissions generated by thermal power units in the power system. The flexible carbon capture model is used to calculate the carbon dioxide emissions generated by carbon capture units in the power system. The atmospheric diffusion model is used to calculate the near-surface carbon dioxide concentration caused by plume diffusion in the power system. The near-surface net disturbance concentration field model is used to calculate the sum of the near-surface carbon dioxide concentration changes caused by the plume and the near-surface carbon dioxide concentration changes caused by the ecosystem. With the goal of balancing economic costs and environmental disturbances, a Markov decision process for the power system to be dispatched is constructed based on an ecological coupling model. The Markov decision process includes the state space, action space, and reward function of the power system to be dispatched. The Markov decision process is solved by using the double-delay deep deterministic policy gradient method to obtain the target scheduling policy.

2. The low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control according to claim 1, characterized in that, The ecological coupling model is constructed using the following method: Calculate the output of thermal power units in the power system to be dispatched, and construct a carbon flow model based on the output of thermal power units; Calculate the total generating power of carbon capture units in the power system to be dispatched, and construct a flexible carbon capture model based on the total generating power; The carbon dioxide emission rate is calculated based on the carbon flow model and the flexible carbon capture model, and an atmospheric diffusion model is constructed based on the carbon dioxide emission rate. Calculate the near-surface carbon dioxide concentration caused by the power system, and construct a near-surface net disturbance concentration field model based on carbon dioxide concentration and atmospheric diffusion model; The upper tail disturbance intensity and average exposure level were constructed based on the near-surface net disturbance concentration field model. The carbon flow model, flexible carbon capture model, atmospheric diffusion model, near-surface net disturbance concentration field model, upper tail disturbance intensity, and average exposure level were identified as the ecological coupling model.

3. The low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control according to claim 1, characterized in that, The Markov decision process for the power system to be dispatched, based on the ecological coupling model, includes: Construct a state space based on the operating parameters of the power system to be dispatched; The action space is determined based on the state space and preset exploration noise; Based on the upper tail disturbance intensity and average exposure level in the ecological coupling model, an environmental objective function is constructed, and an economic objective function is constructed based on the operating costs of thermal power, carbon trading, carbon capture operation, renewable energy curtailment, and energy storage system in the power system to be dispatched. The reward function is then calculated based on the environmental and economic objective functions. The state space, action space, and reward function are defined as a Markov decision process.

4. The low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control according to claim 3, characterized in that, The state space includes the output of wind turbines, the output of photovoltaic units, the total load of the power system to be dispatched, the energy storage of the energy storage system, the solvent volume of the rich liquid storage tank, the carbon dioxide exchange flux, the solvent volume of the amine liquid storage tank in the carbon capture system, the total output of thermal power units, and the boundary layer height. The operational space includes the output of thermal power units, flue gas split ratio, net output power of carbon capture units, operating energy consumption of carbon capture units, and output of energy storage systems. The reward function is expressed by the following formula: ； in, For the power system to be dispatched in the first Rewards obtained by the agent through interaction with the environment during a given time period. As a scalar, To balance the preference parameters between economic and environmental goals, Let the economic objective function be... This is the environmental objective function.

5. The low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control according to claim 3, characterized in that, The environmental objective function is obtained through the following method: The first value is obtained by processing the upper tail disturbance intensity of the scheduling system by a linear normalization operator based on the reference value; The second value is obtained by processing the average exposure level of the system to be scheduled using a linear normalization operator based on the reference value; Construct an environmental objective function based on the first and second values; The economic objective function is obtained through the following method: Determine the operating costs of thermal power plants, carbon trading costs, carbon capture operation costs, renewable energy curtailment costs, and energy storage system operating costs in the system to be dispatched; The total operating cost is calculated based on the operating costs of thermal power plants, carbon trading costs, carbon capture operation costs, renewable energy curtailment costs, and energy storage system operation costs. The total operating cost is then processed using a linear normalization operator based on reference values ​​to obtain the economic objective function.

6. The low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control according to claim 1, characterized in that, The intensity of the upper tail disturbance is expressed by the following formula: ； in, The intensity of the upper tail perturbation. To schedule the current time period Internal and effective diffusion domain The values ​​at the 95th percentile after sorting all grid points by net disturbance concentration values ​​from smallest to largest. For time period The effective diffusion domain within; The average exposure level is expressed by the following formula: ； in, For average exposure levels, This represents the number of grid cells within the effective diffusion domain.

7. The low-carbon dispatching method for power systems based on near-surface carbon dioxide disturbance control according to claim 3, characterized in that, The constraints of the ecological coupling model include energy storage system constraints, energy conservation constraints, thermal power unit constraints, carbon capture system constraints, and power balance constraints.

Images