Power dispatching method and system based on carbon emission

Abstract

According to the power dispatching method and system based on carbon emission, power demand and carbon emission data are obtained in real time through sensors installed on a power user side and a thermal power generating unit; a dynamic clustering algorithm is adopted to analyze the data, and a graph neural network model is used for prediction; inputting a prediction result into a multi-agent reinforcement learning system by using an adaptive scheduling engine to generate decision data; and an optimized power generation combination plan is generated by combining a complex hybrid optimization algorithm, and the carbon quota is dynamically adjusted. According to the invention, the efficiency of power dispatching is improved, the flexibility of carbon emission management is enhanced, and the effective coupling of the power system and the carbon emission dispatching system is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of power dispatching and carbon emission management, and more specifically, to a power dispatching method and system based on carbon emission. Background Art

[0002] Power dispatching promotes sustainable energy development by rationally allocating power resources to ensure that carbon emissions are minimized while meeting power demand. However, existing power dispatching control methods (Chinese invention patent, publication number: CN117013518A, name: Power dispatching control method based on primary and secondary carbon emission quota dispatching) usually only focus on power generation costs, while ignoring the impact of carbon emissions on power dispatching. This situation leads to a lack of effective coupling between the power dispatching system and the carbon emission dispatching system, resulting in energy consumption and waste of resources. For example, the method based on primary and secondary carbon emission quota dispatching, although it attempts to incorporate carbon quota costs into power generation decisions, often fails to effectively integrate real-time carbon emission data in actual operation, resulting in delayed and inaccurate decision-making, and thus unable to achieve a more optimized power resource allocation. Summary of the invention

[0003] In view of the many problems existing in the above-mentioned prior art, the present invention provides a power dispatching method and system based on carbon emissions. The present invention uses real-time monitoring data to perform deep learning through graph neural networks to predict power demand and carbon emissions. On this basis, the prediction results are input into the multi-agent reinforcement learning system through an adaptive scheduling engine to generate intelligent decision-making data. Combined with complex hybrid optimization algorithms, the power generation combination plan is optimized to achieve dynamic adjustment and real-time trading of carbon quotas. The present invention not only improves the accuracy of power dispatching, but also significantly reduces carbon emissions, providing strong support for the realization of green power supply.

[0004] A first aspect of the present invention provides a carbon emission-based power dispatching method, comprising:

[0005] Through the sensors installed on the power user side and the thermal power units, the power demand data of the power users, the carbon emission data of the thermal power units and the environmental parameter data are obtained in real time. The obtained data are cleaned and standardized, and a dynamic clustering data set is generated in combination with a dynamic clustering algorithm.

[0006] Using a graph neural network model to train the dynamic clustering data set to generate predicted electricity demand data and predicted carbon emission data;

[0007] The predicted power demand data and the predicted carbon emission data are input into a multi-agent reinforcement learning system through an adaptive scheduling engine to generate agent decision data, and an optimized power generation combination plan is generated using a complex hybrid optimization algorithm;

[0008] The carbon quota demand data is calculated based on the optimized power generation combination plan, and the carbon quota is adjusted dynamically. At the same time, smart contracts are executed to realize carbon trading, form real-time carbon trading record data, and adjust the scheduling strategy through a feedback mechanism.

[0009] Preferably, the power demand data acquired in real time includes hourly power consumption data of power users, carbon emission data of thermal power units, and environmental parameter data, and the environmental parameter data includes temperature, humidity and wind speed.

[0010] Preferably, in the step of cleaning and standardizing the acquired data, the mean ± 3 times standard deviation method is used to remove outliers in the power demand data and the carbon emission data, and the data standardization scales the data to the [0, 1] interval.

[0011] Preferably, the dynamic clustering algorithm is a K-Means algorithm, which identifies the power consumption patterns of power users by iteratively optimizing cluster centers, and classifies users with similar power demands into the same category.

[0012] Preferably, the node features of the graph neural network model are the power output and carbon emission information of each unit, and the edge features are the power flow intensity between the units. The graph neural network model extracts the relationship between node features and edge features through multi-layer convolution operations.

[0013] Preferably, the multi-agent reinforcement learning system evaluates the current action according to the state of the environment and the historical behavior of each agent, and performs self-learning in combination with a reward mechanism.

[0014] Preferably, the complex hybrid optimization algorithm combines genetic algorithm, particle swarm optimization and simulated annealing to achieve the optimal configuration of the unit output, wherein the calculation expression of the complex hybrid optimization algorithm is:

[0015] Fitness=α×Cost+β×Emission+γ×Reliability

[0016] Among them, Fitness is the fitness function; α, β and γ are weight coefficients; Cost is the power generation cost; Emission is the carbon emission; Reliability is the system reliability index.

[0017] Preferably, in the step of calculating the carbon quota demand data based on the optimized power generation combination plan, the overall carbon quota is calculated by the following calculation expression:

[0018]

[0019] Among them, Q needis the required carbon quota; P i is the output of unit i; CF i is the carbon emission factor of unit i; the overall carbon quota required under a given power generation portfolio is evaluated by taking the weighted sum of the output of all units and the corresponding carbon emission factors.

[0020] Preferably, the smart contract automatically triggers the purchase and sale of quotas based on real-time carbon trading record data, and the smart contract includes transaction conditions, transaction quantity and transaction price.

[0021] A second aspect of the present invention provides a system for implementing the carbon emission-based power dispatching method, comprising:

[0022] The sensor module is used to be installed on the power user end and the thermal power unit to obtain the power demand data of the power user, the carbon emission data of the thermal power unit and the environmental parameter data in real time;

[0023] The data processing module is used to clean and standardize the acquired data and generate a dynamic clustering data set in combination with a dynamic clustering algorithm;

[0024] A graph neural network module, used for training the dynamic clustering data set using a graph neural network model to generate predicted power demand data and predicted carbon emission data;

[0025] An adaptive scheduling engine, used for inputting the predicted power demand data and the predicted carbon emission data into a multi-agent reinforcement learning system to generate agent decision data;

[0026] Complex hybrid optimization algorithm module, used to generate optimized power generation combination plan using complex hybrid optimization algorithm;

[0027] A carbon quota calculation module, used to calculate carbon quota demand data based on the optimized power generation combination plan and dynamically adjust the carbon quota;

[0028] The smart contract execution module is used to execute smart contracts to realize carbon trading, form real-time carbon trading record data, and adjust the scheduling strategy through the feedback mechanism.

[0029] Compared with the prior art, the beneficial effects of the present invention include at least: the carbon emission-based power dispatching method and system provided by the present invention can accurately predict power demand and carbon emissions and achieve optimal configuration of power generation combinations by real-time monitoring of power demand and carbon emission data, and combining graph neural networks with multi-agent reinforcement learning systems. In addition, the system realizes carbon trading by dynamically adjusting carbon quotas and executing smart contracts, promoting green power supply and effectively reducing carbon emissions. This method significantly improves the accuracy and efficiency of power dispatching, promotes the transformation of the power industry to sustainable development, and has important economic and environmental benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flow chart of a carbon emission-based power dispatching method provided in accordance with an embodiment of the present invention;

[0031] Figure 2 is a data flow diagram provided according to an embodiment of the present invention;

[0032] Figure 3 is a carbon quota calculation flow chart provided in accordance with an embodiment of the present invention;

[0033] Figure 4 It is a smart contract execution flow chart provided according to an embodiment of the present invention;

[0034] Figure 5 It is a structural block diagram of a system provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0036] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0037] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0038] like Figure 1 As shown, a carbon emission-based power dispatching method includes:

[0039] Through the sensors installed on the power user side and the thermal power units, the power demand data of the power users, the carbon emission data of the thermal power units and the environmental parameter data are obtained in real time. The obtained data are cleaned and standardized, and a dynamic clustering data set is generated in combination with a dynamic clustering algorithm.

[0040] Preferably, Figure 2 As shown, the power demand data acquired in real time includes hourly power consumption data of power users, carbon emission data of thermal power units and environmental parameter data, and the environmental parameter data includes temperature, humidity and wind speed.

[0041] The hourly electricity consumption data of power users is monitored in real time by sensors installed at the power user end, which can accurately reflect the user's electricity demand in different time periods. This data not only includes the total amount of electricity consumption, but can also be broken down into the electricity consumption of each device or area for in-depth analysis. For example, a household may have higher electricity consumption at night and relatively lower electricity consumption during the day. By analyzing these hourly electricity consumption data, the dispatching system can identify peaks and valleys in electricity consumption, thereby optimizing the power distribution strategy and reducing the power generation burden during periods of high carbon emissions.

[0042] Carbon emissions data of thermal power units are also monitored by sensors. These data include not only real-time carbon emissions, but also the carbon emission characteristics of the units under different loads. This data is crucial for optimizing the power generation mix, because during the scheduling process, the system needs to adjust the start and stop and output of the units according to changes in carbon emissions to achieve low-carbon and efficient power generation goals. For example, when electricity demand rises, the system may choose units with lower carbon emissions to generate electricity first to reduce the overall carbon footprint.

[0043] The acquisition of environmental parameter data provides an additional reference for the dispatching system. Environmental parameter data include factors such as temperature, humidity and wind speed, which will affect electricity demand and power generation efficiency. For example, in hot weather, the frequency of use of refrigeration equipment such as air conditioners increases significantly, leading to an increase in electricity demand; while changes in humidity and wind speed may affect the combustion efficiency of thermal power units. By combining these meteorological data with electricity demand and carbon emissions data, the dispatching system can more accurately predict electricity consumption trends and carbon emissions, and achieve flexible dispatching.

[0044] In summary, the principle of the real-time acquisition of power demand data is to provide comprehensive data support through comprehensive monitoring of power users' power consumption, carbon emissions of thermal power units and environmental meteorological factors, so as to achieve efficient and low-carbon power dispatching. This data-driven approach can effectively improve the accuracy and responsiveness of power dispatching, thereby optimizing resource allocation, reducing carbon emissions, and achieving dual benefits of the environment and the economy.

[0045] Preferably, in the step of cleaning and standardizing the acquired data, the mean ± 3 times standard deviation method is used to remove outliers in the power demand data and the carbon emission data, and the data standardization scales the data to the [0, 1] interval.

[0046] In the present invention, the steps of cleaning and standardizing the acquired data are crucial, aiming to ensure the accuracy and reliability of the data used in the power dispatching process. This step uses the mean ± 3 times standard deviation method to remove outliers in the power demand data and carbon emission data. This method is based on statistical principles. By calculating the mean and standard deviation of the data, a reasonable range is set. Data beyond this range will be regarded as outliers and eliminated. Specifically, the mean and standard deviation of the acquired power demand data and carbon emission data are first calculated, and then it is determined whether each data point falls within the range of the mean ± 3 times the standard deviation. Data beyond this range will be regarded as abnormal data, thereby avoiding misleading subsequent analysis and decision-making due to abnormal values ​​and improving the effectiveness of the model.

[0047] Data standardization is to scale all data to the [0,1] interval so that it has a uniform dimension, which is convenient for subsequent algorithm processing and analysis. The standardization process involves subtracting the minimum value of the data set from each data point and then dividing it by the range of the data set (the difference between the maximum and minimum values). For example, if the minimum value of the power demand data is 50 and the maximum value is 200, then through standardization, the standardized result of the data point x is:

[0048]

[0049] This standardization method helps to eliminate the dimensional effects between different data, balances the contribution of each feature during model training, and improves the convergence speed and accuracy of the algorithm.

[0050] Through these processes, the cleaned and standardized data obtained not only reduces the noise and abnormal effects in the data, but also ensures the effective application of the data in the algorithm model. The result is a more stable and reliable power demand forecast and carbon emission management, making the subsequent scheduling decisions based on graph neural network training and multi-agent reinforcement learning more accurate, and improving the overall performance of the carbon emission-based power dispatch system.

[0051] Preferably, the dynamic clustering algorithm is a K-Means algorithm, which identifies the power consumption patterns of power users by iteratively optimizing cluster centers, and classifies users with similar power demands into the same category.

[0052] In the present invention, the dynamic clustering algorithm is a K-Means algorithm, which aims to optimize the power dispatching strategy by analyzing the power consumption patterns of power users. The K-Means algorithm is an unsupervised learning algorithm, which is mainly used to divide the data set into K different clusters. The principle is to identify the power consumption patterns of power users by iteratively optimizing the cluster centers, and then classify users with similar power demands into the same category. At the beginning of this process, K initial cluster centers are randomly selected, and then the following steps are performed iteratively: first, the power consumption data of each power user is assigned to the nearest cluster center according to the Euclidean distance; second, the cluster center is updated to the mean of all data points in the cluster. This process is repeated until the cluster center no longer changes significantly, or the preset number of iterations is reached.

[0053] In the present invention, by applying the K-Means algorithm, the system can effectively identify power users with different power consumption patterns, for example, household users and industrial users are classified into different clusters. For the power demand data of power users, the algorithm can identify the user groups with peak power consumption and the user groups with relatively stable power consumption in a specific time period. This cluster analysis enables the power dispatching system to formulate more targeted dispatching strategies according to the demand characteristics of different users, thereby improving the efficiency and sustainability of power supply.

[0054] For example, if household users in a certain area use significantly more electricity in the evening in summer than in other time periods, the K-Means algorithm can cluster these users together, and the system can dispatch low-carbon emission generators during peak hours to meet their needs. At the same time, the algorithm can also identify commercial users whose electricity consumption is relatively stable at night. The dispatch system can optimize the start and stop and output of the units based on these different electricity consumption patterns, thereby achieving more efficient resource allocation.

[0055] In summary, the application of the K-Means algorithm in the present invention can identify the power consumption patterns of power users through cluster analysis, improve the response speed and flexibility of the power dispatching system, and ultimately achieve the power dispatching goal based on carbon emissions and reduce overall carbon emissions.

[0056] Using a graph neural network model to train the dynamic clustering data set to generate predicted electricity demand data and predicted carbon emission data;

[0057] Preferably, the node features of the graph neural network model are the power output and carbon emission information of each unit, and the edge features are the power flow intensity between the units. The graph neural network model extracts the relationship between node features and edge features through multi-layer convolution operations.

[0058] In this invention, a graph neural network (GNN) model is used to train the dynamic clustering data set to generate predicted power demand data and predicted carbon emission data. This process fully utilizes the advantages of graph neural networks in processing graph structure data and can effectively capture the complex relationship between node features and edge features.

[0059] The node features of the graph neural network model include the power output and carbon emission information of each unit. These features reflect the performance indicators of the unit under different operating conditions and can provide basic data for prediction. The edge features represent the intensity of power flow between units and reflect the mutual influence between different units in the power network. By structuring this information into a graph, the model can capture the correlation between units. For example, when the power output of a unit increases, the carbon emissions of its adjacent units may be affected. The graph neural network can dynamically adjust its understanding of these relationships through the weights of the edge features.

[0060] During the training process, the graph neural network extracts the relationship between node features and edge features through multi-layer convolution operations. Each layer of convolution operation combines the features of the current node and the features of its adjacent nodes to update the node representation. This iterative update process allows the features of each node to gradually integrate its surrounding information. The resulting node representation not only contains the basic power and carbon emission information of the unit, but also comprehensively considers the power flow with other units. This approach significantly improves the accuracy of the prediction because it not only relies on the historical data of a single unit, but also takes into account the dynamic changes of the entire power system.

[0061] For example, by training on the historical power output and carbon emission data of a particular unit, the graph neural network can predict the changes in the power demand of the unit under specific load conditions, taking into account the output changes of other units connected to it. This prediction result will provide a strong basis for power dispatching decisions, enabling the dispatching system to reasonably allocate the operation of power generating units under carbon emission limits.

[0062] In general, the training process of using the graph neural network model can not only accurately predict electricity demand and carbon emission data, but also reveal the complex interactions in the power network, thereby optimizing carbon emission-based power dispatching strategies and promoting the realization of sustainable development goals.

[0063] The predicted power demand data and the predicted carbon emission data are input into a multi-agent reinforcement learning system through an adaptive scheduling engine to generate agent decision data, and an optimized power generation combination plan is generated using a complex hybrid optimization algorithm;

[0064] Multi-agent reinforcement learning systems play a key role in this process. Each agent represents an independent decision-making unit, such as a different generator set. Based on the state of the environment and the historical behavior of each agent, the system is able to evaluate the current action and make a decision. This evaluation process is combined with a reward mechanism, allowing the agent to self-learn and optimize its strategy by analyzing the results of past decisions. For example, if a unit successfully reduces carbon emissions during a high-demand period, the agent will receive a positive reward and will be more inclined to adopt the same strategy in similar situations in the future.

[0065] Specifically, the agent will select the corresponding power generation strategy based on real-time input data, such as predicted electricity demand and carbon emissions. In the decision-making process, the agent may weigh the startup, shutdown and output levels of different units to generate an optimized power generation combination plan. This plan not only takes into account meeting electricity demand, but also takes carbon emission restrictions into consideration in decision-making to achieve a balance between economic and environmental benefits.

[0066] The application of complex hybrid optimization algorithms further enhances the effectiveness of decision-making. By combining multiple optimization methods, such as genetic algorithms and particle swarm optimization, the algorithm is able to find the best solution among multiple possible power generation combinations. When making decisions, the algorithm can adjust the operating status of the unit in real time to ensure that carbon emissions are minimized while meeting electricity demand. For example, in the case of a sudden increase in electricity demand, the optimization algorithm can quickly calculate the optimal unit output combination and give priority to low-carbon emission units, thereby quickly responding to market demand.

[0067] Preferably, the multi-agent reinforcement learning system evaluates the current action according to the state of the environment and the historical behavior of each agent, and performs self-learning in combination with a reward mechanism.

[0068] The assessment of environmental status includes current electricity demand, carbon emission limits, unit operating status, and external environmental factors (such as weather changes). The system converts this information into a state representation that can be understood by the agent, helping the agent to consider all relevant factors when making decisions. At the same time, each agent will store its own historical behavior and corresponding results, analyze the feedback of past actions, and evaluate the effect of current actions. This record of historical behavior enables the agent to identify which strategies are most effective in different situations, thereby gradually optimizing its decision-making.

[0069] The reward mechanism plays a key role in this process. When the agent takes an action, the system gives corresponding rewards or penalties based on the results (for example, whether the electricity demand is met or whether carbon emissions are reduced). The reward mechanism not only drives the agent to shift to a better strategy, but also encourages the agent to repeat effective decisions when faced with similar environments. For example, if an agent successfully mobilizes a low-carbon unit and reduces carbon emissions during a peak in electricity demand, the agent will receive a higher reward, thereby increasing its tendency to adopt the same strategy in similar situations in the future.

[0070] In specific implementation, when the system detects fluctuations in electricity demand or increases in carbon emissions, the intelligent agent will quickly analyze the current status and refer to historical data to evaluate whether it is necessary to adjust the start and stop or output of the unit. Through continuous self-learning and status evaluation, the intelligent agent can flexibly respond to changes in the electricity market and adjust the power generation combination plan in real time to ensure that carbon emissions are reduced while meeting demand.

[0071] Preferably, the complex hybrid optimization algorithm combines genetic algorithm, particle swarm optimization and simulated annealing to achieve the optimal configuration of the unit output, wherein the calculation expression of the complex hybrid optimization algorithm is:

[0072] Fitness=α×Cost+β×Emission+γ×Reliability

[0073] Among them, Fitness is the fitness function; α, β and γ are weight coefficients; Cost is the power generation cost; Emission is the carbon emission; Reliability is the system reliability index.

[0074] The core of the above algorithm lies in the design of its fitness function, which comprehensively considers power generation costs, carbon emissions and system reliability indicators to ensure the optimal balance between economy and environmental protection.

[0075] According to the specific electricity market environment and policy requirements, these weight coefficients can be dynamically adjusted to optimize the dispatch strategy. For example, when carbon emission policies become increasingly stringent, the weight of β can be increased to emphasize the importance of reducing carbon emissions.

[0076] In the calculation process, the power generation cost Cost reflects the operating cost of the unit at the current output level, usually including fuel costs, maintenance costs and fixed costs. Carbon emissions Emission is calculated based on the unit's carbon emission factor and output level, and is directly related to the achievement of environmental protection goals. The system reliability indicator Reliability measures the unit's ability to continuously supply power under specific operating conditions to ensure the stability of power supply.

[0077] Through genetic algorithms, the system can generate multiple candidate solutions (i.e. unit output configurations) in each generation and evaluate their pros and cons based on the fitness function. Configurations with better performance will be selected and crossover and mutation will be performed to form the next generation of configurations. Particle swarm optimization uses the concept of swarm intelligence to explore better configuration solutions by simulating multiple agents (particles) moving in the solution space. Simulated annealing introduces randomness in local searches to avoid falling into local optimal solutions, thereby enhancing the diversity of the overall search.

[0078] For example, if the current power demand is high and the reliability of a unit is reduced due to maintenance, the algorithm will reduce the fitness of the unit when evaluating the unit output, prompting the system to find output solutions among other available units, thereby achieving a more balanced power supply and environmental protection goals. Ultimately, through this complex hybrid optimization algorithm, the system can dynamically adjust the unit output configuration to meet power demand while minimizing carbon emissions and achieving a sustainable power dispatch strategy.

[0079] The carbon quota demand data is calculated based on the optimized power generation combination plan, and the carbon quota is adjusted dynamically. At the same time, smart contracts are executed to realize carbon trading, form real-time carbon trading record data, and adjust the scheduling strategy through a feedback mechanism.

[0080] The formulation of optimized power generation mix plan aims to dynamically adjust the output of the units according to power demand and carbon emission targets, ensuring that carbon emissions are minimized while meeting demand. Figure 3 As shown, based on the optimization results, the system will calculate the required carbon quota demand data, which reflects the amount of carbon emission rights required by the unit to meet electricity demand under specific operating conditions.

[0081] Once the carbon quota demand data is calculated, the system will dynamically adjust the current carbon quota to ensure that it matches the actual demand. This adjustment is based on real-time monitoring, which means that the system will continuously analyze market dynamics, user needs and policy changes to maintain the flexibility and effectiveness of carbon quotas. The process of dynamically adjusting carbon quotas includes regular review of allocated quotas and expiring quotas to ensure a quick response to changes in market demand and environmental policies.

[0082] In addition, in order to realize carbon trading, the system will execute smart contracts, which is a form of contract that is automatically executed through blockchain technology. The application of smart contracts ensures the transparency and reliability of the carbon trading process, reduces the involvement of intermediaries, and reduces transaction costs. In specific implementation, the execution of carbon trading will automatically record the details of each transaction, including the two parties to the transaction, the transaction quantity and price, etc. This information forms real-time carbon trading record data, which can be queried and audited by relevant parties at any time.

[0083] Through the feedback mechanism, the system can effectively adjust the dispatch strategy according to the real-time carbon trading records and the dynamic adjustment of carbon quotas. For example, if it is found that the demand for carbon quotas is higher than the market supply, the system may give priority to mobilizing low-carbon emission units to reduce overall carbon emissions. This decision will be fed back into the optimization of the power generation portfolio plan, making subsequent dispatch decisions more accurate.

[0084] Preferably, in the step of calculating the carbon quota demand data based on the optimized power generation combination plan, the overall carbon quota is calculated by the following calculation expression:

[0085]

[0086] Among them, Q need is the required carbon quota; P i is the output of unit i; CF i is the carbon emission factor of unit i; the overall carbon quota required under a given power generation portfolio is evaluated by taking the weighted sum of the output of all units and the corresponding carbon emission factors.

[0087] The core of the above calculation expression is to perform a weighted summation of the output of all units and the corresponding carbon emission factors to evaluate the overall carbon quota required under a given power generation portfolio.

[0088] In principle, carbon quotas are set to limit and manage carbon emissions, and their demand is directly affected by power production activities. In the power dispatch process, the output and carbon emission factors of different units reflect their impact on the environment, so by combining them through a formula, an accurate carbon quota demand can be obtained. This calculation method effectively takes into account the characteristics of each unit, making the calculation of the overall carbon quota more scientific and reasonable.

[0089] In the specific implementation, the system first obtains the output and carbon emission factor of each unit based on real-time data. These data can be monitored in real time by sensors installed on the power user end and the thermal power unit. Then, using the above formula, the system multiplies the output of each unit and the corresponding carbon emission factor, and then sums them up to obtain the total carbon quota demand.

[0090] The effect of this process is remarkable. On the one hand, it ensures that the demand for carbon quotas is accurately assessed under different power generation combinations, which helps to formulate more effective scheduling strategies; on the other hand, the dynamic calculation characteristics enable the system to adjust carbon quotas in a timely manner according to changes in power demand and support real-time carbon trading. This flexibility and real-time performance are the key to dealing with complex power markets and environmental policies, and can effectively reduce carbon emissions and promote the realization of sustainable development goals.

[0091] For example, suppose a power grid consists of three units, with outputs of P1 = 100MW, P2 = 200MW and P3 = 150MW, and carbon emission factors of CF1 = 0.4tCO2 / MWh, CF2 = 0.3tCO2 / MWh and CF3 = 0.5tCO2 / MWh. By applying the above calculation formula, the system can calculate the required carbon quota as:

[0092] Q need =(100×0.4)+(200×0.3)+(150×0.5)=40+60+75=175tCO2

[0093] The results will guide subsequent carbon quota management and trading decisions and provide support for achieving low-carbon economic goals.

[0094] Preferably, Figure 4 As shown, the smart contract automatically triggers the purchase and sale of quotas based on real-time carbon trading record data, and the smart contract includes transaction conditions, transaction quantity and transaction price.

[0095] A smart contract is a self-executing contract whose terms are written directly into the code and can be automatically triggered on the blockchain according to preset conditions. Specifically, the smart contract monitors carbon trading record data in real time to determine when to buy or sell quotas.

[0096] During the execution process, the smart contract first defines a series of transaction conditions, including but not limited to the upper limit of carbon emissions, the market price fluctuation range, the number of transactions, and the relevant contract price. When the real-time carbon trading record data is updated, the smart contract will automatically evaluate the current market status and compare it with these preset conditions. For example, if the real-time data shows that the market price of carbon quotas is lower than the preset purchase price, and the current demand for carbon quotas reaches a certain threshold, the smart contract will trigger the automatic purchase of quotas. Conversely, when the market price is higher than the preset selling price, the contract will execute the instruction to sell quotas.

[0097] This mechanism ensures the efficiency and accuracy of transactions, eliminates delays caused by human intervention, and greatly improves response speed. At the same time, the execution of smart contracts is based on transparent and tamper-proof blockchain technology, which provides trust protection for both parties to the transaction and reduces transaction risks.

[0098] For example, suppose a power company sets the trading conditions as follows: if the market price of carbon quotas is less than 5 yuan / ton and carbon emissions exceed 100 tons, 100 tons of carbon quotas will be automatically purchased. If the market price exceeds 7 yuan / ton and the carbon quotas held are less than 200 tons, 100 tons of carbon quotas will be automatically sold. With such a setting, smart contracts can automatically make decisions on buying or selling based on real-time data, thereby effectively managing carbon quotas, reducing corporate carbon emission costs, and enhancing market liquidity.

[0099] like Figure 5 As shown, a system for implementing the carbon emission-based power dispatching method comprises:

[0100] The sensor module is used to be installed on the power user side and the thermal power unit to obtain the power demand data of the power user, the carbon emission data of the thermal power unit and the environmental parameter data in real time; these data are the basis for subsequent analysis. Through real-time monitoring, the system can respond to user needs and environmental changes in a timely manner, thereby optimizing power dispatch.

[0101] The data processing module is used to clean and standardize the acquired data, and generate a dynamic clustering data set in combination with the dynamic clustering algorithm; the data processing module cleans and standardizes the acquired data, uses the mean ± 3 times standard deviation method to remove outliers, and scales the data to the [0,1] interval. This step ensures the accuracy and consistency of the data, laying a good foundation for the subsequent dynamic clustering algorithm to generate a dynamic clustering data set. Dynamic clustering groups users with similar power demands into the same category, making it easier to identify power consumption patterns.

[0102] The graph neural network module is used to train the dynamic clustering data set using a graph neural network model to generate predicted power demand data and predicted carbon emission data; this method can capture the complex relationship between user electricity consumption behavior and carbon emission characteristics, making the prediction more accurate and supporting intelligent scheduling decisions.

[0103] An adaptive scheduling engine is used to input the predicted power demand data and the predicted carbon emission data into a multi-agent reinforcement learning system to generate agent decision data; the system continuously optimizes the agent's scheduling decisions by combining environmental status and historical behavior, thereby improving scheduling efficiency and reducing carbon emissions.

[0104] The complex hybrid optimization algorithm module is used to generate an optimized power generation combination plan using a complex hybrid optimization algorithm; the complex hybrid optimization algorithm module uses genetic algorithms, particle swarm optimization, simulated annealing and other technologies to generate an optimized power generation combination plan. This module is based on the fitness function and comprehensively considers factors such as power generation cost, carbon emissions and system reliability to ensure the optimal configuration of the power generation combination.

[0105] The carbon quota calculation module is used to calculate the carbon quota demand data based on the optimized power generation combination plan and dynamically adjust the carbon quota; this process ensures that the system achieves economical and efficient power dispatch while complying with carbon emission regulations.

[0106] The smart contract execution module is used to execute smart contracts to realize carbon trading, form real-time carbon trading record data, and adjust the scheduling strategy through the feedback mechanism. This mechanism ensures the transparency and efficiency of transactions and promotes sustainable development.

[0107] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0108] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

[0109] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for power dispatching based on carbon emissions, characterized in that: include: Through the sensors installed on the power user side and the thermal power unit, the power demand data of the power user, the carbon emission data of the thermal power unit and the environmental parameter data are obtained in real time, the obtained data are cleaned and standardized, and a dynamic clustering data set is generated in combination with a dynamic clustering algorithm; Using a graph neural network model to train the dynamic clustering data set to generate predicted electricity demand data and predicted carbon emission data; The predicted power demand data and the predicted carbon emission data are input into a multi-agent reinforcement learning system through an adaptive scheduling engine to generate agent decision data, and an optimized power generation combination plan is generated using a complex hybrid optimization algorithm; The carbon quota demand data is calculated based on the optimized power generation combination plan, and the carbon quota is dynamically adjusted. At the same time, smart contracts are executed to realize carbon trading, form real-time carbon trading record data, and adjust the scheduling strategy through a feedback mechanism.

2. The carbon emission-based power dispatching method according to claim 1, characterized in that: The power demand data acquired in real time includes hourly power consumption data of power users, carbon emission data of thermal power units, and environmental parameter data, wherein the environmental parameter data includes temperature, humidity, and wind speed.

3. The carbon emission-based power dispatching method according to claim 1, characterized in that: In the step of cleaning and standardizing the acquired data, the mean ± 3 times standard deviation method is used to remove outliers in the power demand data and the carbon emission data, and the data standardization process scales the data to the [0, 1] interval.

4. The carbon emission-based power dispatching method according to claim 1, characterized in that: The dynamic clustering algorithm is a K-Means algorithm, which identifies the power consumption patterns of power users by iteratively optimizing cluster centers and classifies users with similar power demands into the same category.

5. The carbon emission-based power dispatching method according to claim 1, characterized in that: The node features of the graph neural network model are the power output and carbon emission information of each unit, and the edge features are the power flow intensity between the units. The graph neural network model extracts the relationship between node features and edge features through multi-layer convolution operations.

6. The carbon emission-based power dispatching method according to claim 1, characterized in that: The multi-agent reinforcement learning system evaluates the current action according to the environment state and the historical behavior of each agent, and performs self-learning in combination with the reward mechanism.

7. The carbon emission-based power dispatching method according to claim 1, characterized in that: The complex hybrid optimization algorithm combines genetic algorithm, particle swarm optimization and simulated annealing to achieve the optimal configuration of the unit output. The calculation expression of the complex hybrid optimization algorithm is: Fitness=α×Cost+β×Emission+γ×Reliability Among them, Fitness is the fitness function; α, β and γ are weight coefficients; Cost is the power generation cost; Emission is the carbon emission; Reliability is the system reliability index.

8. The carbon emission-based power dispatching method according to claim 1, characterized in that: In the step of calculating the carbon quota demand data based on the optimized power generation combination plan, the overall carbon quota is calculated by the following calculation expression: Among them, Q need is the required carbon quota; P i is the output of unit i; CF i is the carbon emission factor of unit i; the overall carbon quota required under a given power generation portfolio is evaluated by taking the weighted sum of the output of all units and the corresponding carbon emission factors.

9. The carbon emission-based power dispatching method according to claim 1, characterized in that: The smart contract automatically triggers the purchase and sale of quotas based on real-time carbon trading record data, and the smart contract includes transaction conditions, transaction quantity and transaction price.

10. A system for implementing the carbon emission-based power dispatching method according to any one of claims 1 to 9, characterized in that: include: The sensor module is used to be installed on the power user end and the thermal power unit to obtain the power demand data of the power user, the carbon emission data of the thermal power unit and the environmental parameter data in real time; The data processing module is used to clean and standardize the acquired data and generate a dynamic clustering data set in combination with a dynamic clustering algorithm; A graph neural network module, used for training the dynamic clustering data set using a graph neural network model to generate predicted power demand data and predicted carbon emission data; An adaptive scheduling engine, used for inputting the predicted power demand data and the predicted carbon emission data into a multi-agent reinforcement learning system to generate agent decision data; Complex hybrid optimization algorithm module, used to generate optimized power generation combination plan using complex hybrid optimization algorithm; A carbon quota calculation module, used to calculate carbon quota demand data based on the optimized power generation combination plan and dynamically adjust the carbon quota; The smart contract execution module is used to execute smart contracts to realize carbon trading, form real-time carbon trading record data, and adjust the scheduling strategy through the feedback mechanism.

Images