A method, device, medium and program product for monitoring carbon emissions of a thermal power unit

Abstract

The application discloses a thermal power unit carbon emission monitoring method, equipment, medium and program product, relates to the field of power system energy management, and comprises the following steps: acquiring a knowledge graph and a cubic function simulation curve; screening and analyzing an initial solution set and a high-quality solution set from the knowledge graph; performing multi-objective optimization on the solution set by using a genetic algorithm, and adjusting the crossover probability and the mutation probability of each population iteration of the genetic algorithm by using a deep Q network until the genetic algorithm reaches the maximum iteration number or the population convergence condition, and then outputting optimal output of the thermal power unit; determining a coal combustion coefficient of the thermal power unit according to the cubic function simulation curve, and calculating the carbon emission of the thermal power unit according to the coal combustion coefficient and the optimal output; regarding each thermal power unit as an entity node of the knowledge graph, configuring a carbon emission threshold for each entity node, and monitoring the carbon emission of the thermal power unit based on the carbon emission of the thermal power unit and the carbon emission threshold.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power system energy management, more particularly, it relates to a method, device, medium and program product for monitoring carbon emissions of a thermal power generating unit. BACKGROUND

[0002] Thermal power generating unit output optimization refers to improving the energy utilization efficiency of a thermal power generating unit and reducing its coal consumption to achieve the goal of energy saving and emission reduction by in-depth analysis and optimization of the operating performance of the thermal power generating unit.

[0003] Thermal power generating unit coal consumption modeling is an important method for describing the coal consumption characteristics of the unit under different power generation conditions. By establishing a mathematical model with power generation as the main input variable, using historical data or operating characteristics, the change rule of coal consumption is usually described in the form of a quadratic function. The main purpose of this model is to optimize the power generation efficiency of the unit, reduce the cost of coal consumption, and provide a scientific basis for dispatching optimization and carbon emission calculation. However, the traditional quadratic curve fitting method has many shortcomings in modeling, including the inability to accurately reflect the nonlinear change characteristics in the high load and low load intervals, the difficulty in describing complex high-order nonlinear changes, and the problem of insufficient fitting accuracy in a wide load range.

[0004] Secondly, there are currently genetic algorithms applied to the field of thermal power generating unit carbon emission calculation, and genetic algorithms as a heuristic search method, the crossover and mutation operations are the key steps to realize the evolution of the population. Crossover operation generates new offspring by exchanging part of the gene fragments of the parent individuals, aiming to integrate the excellent characteristics of the parents; mutation operation introduces diversity by randomly changing gene values to avoid the population falling into local optimum. The effective combination of these two operations makes genetic algorithms have strong search ability. However, in order to further improve the performance of the algorithm, the exploration ability and coverage of the population need to be evaluated and optimized through the population distribution diversity index and the hyper volume index. The population distribution diversity index is used to measure the breadth and uniformity of the population, while the hyper volume index reflects the volume occupied by the population in the target space, which can effectively evaluate the diversity and coverage of the solution set in the optimization process. However, the traditional genetic algorithm still has shortcomings in the generation of initial population and parameter setting. The randomly generated initial population often lacks coverage and diversity, leading to the optimization process easily falling into local optimum. In addition, the fixed crossover probability and mutation probability lack flexibility and cannot be dynamically adjusted according to the actual needs of the problem, thereby affecting the exploration efficiency of the genetic algorithm and the optimization ability of the global optimal solution. SUMMARY

[0005] The application aims to provide a thermal power unit carbon emission monitoring method, device, medium and program product, and solves the problem that the search optimization ability of the genetic algorithm provided by the related art is insufficient in the search process, and the calculated output of the thermal power unit is not the optimal output.

[0006] The above technical purposes of the application are achieved by the following technical solutions:

[0007] In a first aspect, the application provides a thermal power unit carbon emission monitoring method, which comprises:

[0008] Obtaining a knowledge graph constructed by operation parameters and historical operation data of the thermal power unit, and a cubic function simulation curve of coal consumption fitted from the historical operation data of the thermal power unit, wherein the historical operation data comprises load values and coal consumption corresponding to the load values;

[0009] Filtering and analyzing an initial solution set and a high-quality solution set that meet the operation constraints of the thermal power unit from the knowledge graph, and fusing the initial solution set and the high-quality solution set to obtain a solution set;

[0010] Performing multi-objective optimization on the solution set by using a genetic algorithm, and adjusting the crossover probability and the mutation probability of each population iteration of the genetic algorithm by using a deep Q network, until the genetic algorithm reaches a maximum iteration number or a population convergence condition, and outputting the optimal output of the thermal power unit;

[0011] Determining a coal consumption coefficient of the thermal power unit according to the cubic function simulation curve, and calculating the carbon emission of the thermal power unit according to the coal consumption coefficient and the optimal output;

[0012] Regarding each thermal power unit as an entity node of the knowledge graph, configuring a carbon emission threshold for each entity node, and monitoring the carbon emission of the thermal power unit based on the carbon emission of the thermal power unit and the carbon emission threshold.

[0013] In an implementation scheme, the cubic function fitting curve of the coal consumption fitted from the historical operation data of the thermal power unit is obtained, and specifically:

[0014] Taking the coal consumption as a target variable and the load value as an independent variable, a cubic curve is generated;

[0015] The parameters of the cubic curve are fitted by using a least square method to obtain the cubic function simulation curve of the coal consumption.

[0016] In an implementation scheme, the initial solution set and the high-quality solution set that meet the operation constraints of the thermal power unit are filtered and analyzed from the knowledge graph, and specifically:

[0017] Filtering an initial solution set meeting the operation constraints of the thermal power generating unit from optimal operation modes of different thermal power generating units under different operation conditions in the knowledge graph; and

[0018] Extracting historical operation data of the thermal power generating unit from the knowledge graph as a historical optimal solution, and analyzing a high-quality solution set meeting the operation constraints of the thermal power generating unit according to the historical optimal solution and a preset rule base.

[0019] In an implementation scheme, a deep Q network is used to adjust the crossover probability and the mutation probability of each population iteration of the genetic algorithm until the genetic algorithm reaches a maximum iteration number or a population convergence condition, and an optimal output of the thermal power generating unit is output, specifically as follows:

[0020] Randomly initializing the weights of the deep Q network, and randomly selecting the crossover probability and the mutation probability; wherein, the input parameters of the genetic algorithm are used as the state function of the deep Q network, the adjustment strategy of the crossover probability and the mutation probability of the genetic algorithm is used as the action function of the deep Q network, and the population quality of the genetic algorithm is used as the reward function of the deep Q network;

[0021] In each population iteration of the genetic algorithm, the population state at the current time is calculated according to the randomly selected crossover probability and mutation probability, the crossover probability and mutation probability at the next time are output by the action function, the population state at the next time is calculated according to the crossover probability and mutation probability at the next time, and the population quality is calculated according to the reward function;

[0022] The population state at the current time, the crossover probability and the mutation probability, the population quality and the population state at the next time are randomly extracted to update the weights of the deep Q network, the crossover probability and the mutation probability of the next population iteration of the genetic algorithm are adjusted again based on the deep Q network with updated weights, and the optimal output of the thermal power generating unit is output until the genetic algorithm reaches the maximum iteration number or the population convergence condition.

[0023] In an implementation scheme, the optimization target of the genetic algorithm includes maximum joint output of wind and light, minimum net load fluctuation, minimum system operation cost and minimum carbon emission.

[0024] In an implementation scheme, the input parameters of the genetic algorithm include the number of non-dominated solutions, the population distribution diversity index, the hyper volume index and the population iteration number.

[0025] In an implementation scheme, the carbon emission of the thermal power generating unit is calculated according to the coal combustion factor and the optimal output, specifically as follows:

[0026] The coal consumption of the thermal power generating unit under the optimal output is determined according to the coal combustion factor;

[0027] The carbon emission of the thermal power generating unit is calculated according to the coal consumption and a preset carbon emission factor.

[0028] In a second aspect of the present application, an electronic device is provided, comprising a memory and a processor;

[0029] The memory is configured to store a computer program comprising program instructions;

[0030] The processor is configured to execute the program instructions to cause the electronic device to perform the steps of the method for monitoring carbon emission of a thermal power unit according to the first aspect of the present application.

[0031] In a third aspect of the present application, a computer program product comprising program instructions is provided, which when executed by an electronic device, causes the electronic device to perform the steps of the method for monitoring carbon emission of a thermal power unit according to the first aspect of the present application.

[0032] In a fourth aspect of the present application, a computer readable storage medium is provided, which comprises a computer program that, when executed by one or more processors, implements the method for monitoring carbon emission of a thermal power unit according to the first aspect of the present application.

[0033] Compared with the prior art, the present application has the following beneficial effects:

[0034] In the carbon emission monitoring method of the thermal power generating unit provided in the application, firstly, instead of randomly screening based on the historical operation data of the thermal power generating unit, the initial population of the genetic algorithm is obtained by combining the historical operation data and the deterministic data to construct a knowledge graph, screening the historical data based on the knowledge graph and comprehensively combining the results of the data reasoning to improve the feasibility and diversity of the initial population, thereby avoiding the low-quality solution problem caused by random initialization, significantly accelerating the convergence speed of the genetic algorithm, and further improving the efficiency and quality of the multi-objective optimization. Secondly, the crossover and mutation probability parameters of the genetic algorithm are not kept constant, but are dynamically adjusted by the deep Q network. In the early iteration, the mutation probability is increased to improve the exploration ability, and in the later iteration, the crossover probability is increased to focus on local development. At the same time, the deep Q network is used to learn the corresponding relationship between the population state and the optimal evolution strategy, which can make the algorithm more intelligent and avoid manual parameter adjustment. At the same time, the reward function is used to optimize the population quality and diversity, balance exploration and development, and effectively improve the quality of the Pareto front. Finally, the carbon emission coefficient of the thermal power generating unit is not an approximate value calculated based on the historical data, but is determined by the coal consumption function of the thermal power generating unit modeled by the cubic function curve. Using the cubic function curve to estimate the coal consumption of the thermal power generating unit can make the curve show more complex nonlinear characteristics, such as multiple inflection points, steep growth or slow change trend, etc. In addition, the cubic function curve can better reflect the dynamic performance of the unit during load regulation. The thermal power generating unit may exhibit significantly different coal consumption characteristics in different load intervals. In the low load interval, the coal consumption increases rapidly due to low equipment efficiency and high loss. In the medium load interval, the coal consumption and load show a regular change trend, which is close to a linear relationship. In the high load interval, when the unit is close to full load, the coal consumption may tend to be flat or show complex changes due to the unit approaching the limit operating state. The cubic curve can more accurately describe these nonlinear changes by using high-order terms of the polynomial, thereby improving the fitting accuracy and reducing the estimation error. Finally, the knowledge graph-based carbon emission monitoring method of the thermal power generating unit can quickly locate the thermal power generating unit with excessive emissions through the semantic association function of the knowledge graph to shorten the analysis time, update the carbon emission data in real time, and determine the state according to the carbon emission threshold to provide targeted adjustment plan, thereby reducing the carbon emission of the thermal power generating unit. BRIEF DESCRIPTION OF DRAWINGS

[0035] The accompanying drawings, which are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings:

[0036] Figure 1 A flowchart of a method for monitoring carbon emissions of a thermal power generating unit is provided for an embodiment of the present application.

[0037] Figure 2 A constraint satisfaction rate comparison chart of using a knowledge graph to assist in initializing a population and a traditional random initial population is provided for an embodiment of the present application.

[0038] Figure 3 A convergence curve comparison chart of using a knowledge graph to assist in initializing a population and a traditional random initial population is provided for an embodiment of the present application.

[0039] Figure 4 A mutation probability and crossover probability dynamic adjustment curve chart after using NSGA-II combined with DQN to optimize the crossover and mutation strategy is provided for an embodiment of the present application.

[0040] Figure 5 A convergence cancellation comparison chart of using NSGA-II combined with DQN to optimize the crossover and mutation strategy and a traditional NSGA-II is provided for an embodiment of the present application.

[0041] Figure 6 A fitting effect comparison chart of using a cubic curve fitting coal consumption and a quadratic curve fitting coal consumption is provided for an embodiment of the present application.

[0042] Figure 7 A monitoring demonstration chart of a carbon emission situation of a thermal power generating unit is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present application clearer, further detailed description will be given to the present application in combination with embodiments and drawings, and the illustrative embodiments of the present application and the description thereof are only used to explain the present application, and should not be regarded as a limitation to the present application.

[0044] It should be noted that the term "include" or "may include" used in various embodiments of the present application indicates the existence of the claimed function, operation or element, and does not limit the addition of one or more functions, operations or elements. In addition, as used in various embodiments of the present application, the terms "include", "have" and their homonyms are only intended to indicate a specific feature, number, step, operation, element, component or combination of the foregoing, and should not be understood as first excluding the existence or addition of one or more other features, numbers, steps, operations, elements, components or combinations of the foregoing.

[0045] It should be understood that terms such as "first", "second" are used only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined as "first", "second" can be explicitly or implicitly included one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly specified.

[0046] Please refer to Figure 1 , Figure 1 A flowchart of a carbon emission monitoring method for a thermal power generating unit according to an embodiment of the present application is shown in FIG. 1, which comprises the following steps: Figure 1

[0047] S101, obtaining a knowledge graph constructed by operation parameters and historical operation data of the thermal power generating unit, and a cubic function simulation curve of coal consumption fitted by the historical operation data of the thermal power generating unit; wherein the historical operation data comprises load values and coal consumption corresponding to the load values.

[0048] In this embodiment, the knowledge graph is constructed: input the determined data (thermal power generating unit operation parameters such as steam turbine operation parameters, generator operation parameters, etc.) and historical operation data (such as thermal power generating unit historical operation data, maximum output of wind and light and load), define the entities related to the thermal power generating unit (unit type, operation state, fuel consumption, emission index, operation, etc.), relationships (such as thermal power generating unit-load, thermal power generating unit-fuel, thermal power generating unit output-wind and light unit maximum output, etc.) and constraint rules (load balance, etc.) based on the input data.

[0049] The process of fitting the cubic function fitting curve is as follows: taking the coal consumption as the target variable and the load value as the independent variable, a cubic curve is generated; the parameters of the cubic curve are fitted by the least square method to obtain the cubic function simulation curve of the coal consumption.

[0050] Specifically, first, the actual coal consumption of the thermal power generating unit under different loads is measured to obtain a set of load-coal consumption data points. Second, the measured data is screened to eliminate abnormal values or data with large errors to ensure data quality. Third, the measured data is substituted into the cubic curve model by using the polynomial regression method, and the least square method is used to solve the parameters of the cubic curve. Finally, the fitted cubic curve model is compared with the actual measured data to verify the fitting accuracy and error, and the parameters are adjusted if necessary. The finally determined cubic curve parameters are applied to the coal consumption estimation model of the thermal power generating unit for analyzing and optimizing the operation efficiency of the thermal power generating unit.

[0051] ​In this embodiment, the simulation of the coal consumption of the thermal power generating unit using the cubic curve can make the simulation curve show more complex nonlinear characteristics, such as multiple inflection points, steep growth or slow change trend, etc. In addition, the cubic curve can better reflect the dynamic performance of the unit during load regulation. The thermal power generating unit can show obviously different coal consumption characteristics in different load intervals. In the low load interval, the coal consumption increases rapidly due to low equipment efficiency and high loss. In the medium load interval, the coal consumption and the load show a regular change trend, which is close to a linear relationship. In the high load interval, close to full load, due to the equipment close to the limit operating state, the coal consumption can tend to be flat or show complex changes. The cubic curve can more accurately describe these nonlinear changes through the high-order terms of the polynomial, avoid the oversimplification of the quadratic curve, thus improving the fitting accuracy and reducing the estimation error.

[0052] The cubic curve is used for modeling the coal consumption of the thermal power generating unit. The cubic fitting value, the quadratic fitting value and the actual coal consumption of the coal consumption under different load conditions are shown in FIG. 2. Figure 6 As shown in FIG. 2, the specific evaluation standard values are shown in Table 1. As can be seen from Table 1, the fitting error of the cubic curve in all load intervals is significantly lower than that of the quadratic curve, especially in the low load and high load intervals, which can better reflect the nonlinear change characteristics. The cubic curve has stronger fitting ability for the complex changes of the coal consumption (such as multiple inflection points), and the maximum error is significantly reduced. At the same time, it can be seen from Table 1 that the fitting effect of the cubic curve in different load intervals is better than that of the quadratic curve.

[0053] Table 1

[0054]

[0055] S102, filter and analyze the initial solution set and the high-quality solution set that meet the operation constraints of the thermal power generating unit from the knowledge graph, fuse the initial solution set and the high-quality solution set, and obtain a solution set.

[0056] In this embodiment, the determination of the initial solution set is as follows: based on the historical data associated with the knowledge graph, the optimal operation mode of different units under specific conditions is analyzed, a batch of solutions that meet the constraint rules are selected, and a candidate set of the initial population, i.e. the initial solution set, is constructed.

[0057] The specific steps for determining the set of high-quality solutions are as follows: extract historical operating data from the knowledge graph, including historical optimal solutions such as power generation costs, load demand, fuel consumption, and carbon emissions; build a rule base based on the above data (such as the power output range of thermal power units, minimum start and stop times, emission limits, etc.); use a logical reasoning engine to infer potential high-quality solutions that meet the constraints from the rule base as the set of high-quality solutions. For example, if the unit's power generation power is close to the historical optimal value and the fuel price drops, it is inferred that this power may be a high-quality solution.

[0058] Finally, the historical data screening results and the associated reasoning results are summarized; illegal solutions that do not meet the constraints are filtered out, and finally an initial population that takes both diversity and quality into account is generated, that is, the solution set.

[0059] By integrating historical data, rule constraints, and semantic associations, this embodiment enables the knowledge graph to generate high-quality solutions, improving the feasibility and diversity of the population, thereby avoiding the problem of low-quality solutions caused by random initialization and significantly accelerating the convergence of the algorithm. In addition, the reasoning ability of the knowledge graph can adapt to dynamic environmental changes in real time, generating solutions that meet the current constraints, enhancing the practical applicability and credibility of the optimization results, while reducing the generation of individuals that do not meet the constraints, reducing computational redundancy, and ultimately improving the efficiency and quality of multi-objective optimization.

[0060] S103, using a genetic algorithm to perform multi-objective optimization on the solution set, and using a deep Q network to adjust the crossover probability and mutation probability of each population iteration of the genetic algorithm until the genetic algorithm reaches the maximum number of iterations or the population convergence condition, and outputs the optimal output of the thermal power unit.

[0061] In this embodiment, a deep Q network is used to adjust the crossover probability and mutation probability of each population iteration of the genetic algorithm until the genetic algorithm reaches the maximum number of iterations or the population convergence condition, and then outputs the optimal output of the thermal power unit, specifically:

[0062] Randomly initialize the weights of the deep Q network, and randomly select the crossover probability and mutation probability; wherein, the input parameters of the genetic algorithm are taken as the state function of the deep Q network, the adjustment strategy of the crossover probability and mutation probability of the genetic algorithm is taken as the action function of the deep Q network, and the population quality of the genetic algorithm is taken as the reward function of the deep Q network; in each population iteration of the genetic algorithm, the population state at the current time is calculated according to the randomly selected crossover probability and mutation probability, the crossover probability and mutation probability at the current time are output by the action function, the population state at the next time is calculated according to the crossover probability and mutation probability at the next time, and the population quality is calculated according to the reward function; the population state at the current time, the crossover probability and mutation probability, the population quality and the population state at the next time are randomly selected to update the weights of the deep Q network, and the crossover probability and mutation probability of the next population iteration of the genetic algorithm are adjusted again based on the deep Q network with updated weights, until the genetic algorithm reaches the maximum iteration number or the population convergence condition, and the optimal output of the thermal power unit is output.

[0063] Specifically, the deep Q network dynamically adjusts the crossover and mutation parameters of the genetic algorithm, and the detailed description is as follows:

[0064] The genetic algorithm provided in the embodiment adopts the NSGA-II multi-objective optimization algorithm, and the optimization objectives include maximum joint output of wind and light, minimum net load fluctuation, minimum system operation cost and minimum carbon emission. In the optimization process, the deep Q network is introduced, the crossover probability and mutation probability are dynamically adjusted according to the current population evolution state, and finally the optimal output of the thermal power unit is obtained.

[0065] State function definition: state is a feature describing the current population quality and distribution, and the input parameters of the genetic algorithm include the number of non-dominated solutions, population distribution diversity index, hyper volume index and population iteration number, therefore, the state function includes:

[0066] : the number of non-dominated solutions.

[0067] : population distribution diversity index (such as average crowding distance).

[0068] : hyper volume index.

[0069] : current evolution generation, normalized to [0, 1].

[0070] State vector is represented as: .

[0071] Action function definition: action is the adjustment strategy of the crossover and mutation probability. The action space is continuous value: where, represents the crossover probability, represents the mutation probability.

[0072] These ranges ensure the rationality of the crossover and mutation operations.

[0073] The reward function defines the reward for evaluating the action, which can be designed as:

[0074] Population quality improvement reward: , which is the increment of the current hypervolume indicator.

[0075] Population diversity improvement reward:

[0076] Comprehensive reward: where, α, β represent the weighting coefficients (here, α = 0.7, β = 0.3).

[0077] The design of the deep Q network is described as follows:

[0078] The network structure includes the input layer: state vector , hidden layer: multi-layer fully connected network, each layer containing 128 neurons, and activation function ReLU, and output layer: action value Q(S, A) is the Q value for each possible action.

[0079] where, input layer: 4 nodes (corresponding to the 4-dimensional features of ). Hidden layer 1: 128 nodes. Hidden layer 2: 128 nodes. Output layer: Q value mapping of continuous action.

[0080] The parameter settings are as follows: learning rate η = 0.001, discount factor γ = 0.99, experience replay buffer size is 10000 experiences, small batch training sample size is 64, and the frequency of updating the target network is once every 100 iterations.

[0081] Strategy optimization process of deep Q network

[0082] Randomly initialize the weights of the DQN network, and randomly select the crossover and mutation probabilities.

[0083] In each generation, according to the current population state , the action is predicted by the DQN model, the crossover and mutation operations are performed to generate a new population, and the population state at the next time is calculated.

[0084] According to the population quality improvement, the reward is calculated and stored in the experience replay buffer.

[0085] Randomly extract batches of data from the buffer , update the weights of the Q network:

[0086] ; , where N represents the maximum number of iterations,

[0087] Regularly update target networks The weight of . Randomly select actions in the exploration phase, and gradually increase the probability of model selection as training progresses (ϵ-greedy strategy, initial value ϵ = 1.0, decaying by 0.995 per generation).

[0088] The process terminates after reaching the maximum number of iterations or the population convergence condition.

[0089] After constructing the knowledge graph and driving the generation of the solution set of the initial population, the constraint satisfaction rate of the initial population after each iteration is compared with that of the traditional randomly generated initial population after 100 generations under the same parameter environment. Figure 2 As shown by Figure 2 As can be seen, the constraint satisfaction rate for random initialization is low in the initial generations, only around 10%. It gradually improves with increasing iterations, but only reaches 80% by the 100th generation. This indicates that random initialization is inefficient in generating feasible solutions. The constraint satisfaction rate for knowledge graph-assisted initialization reaches 60% in the first generation and continues to grow rapidly, then levels off after the 40th generation, ultimately reaching 99.9%. This demonstrates that the reasoning power of the knowledge graph significantly enhances the constraint satisfaction rate of the population. Throughout the optimization process, the constraint satisfaction rate of the knowledge graph-assisted initialization method consistently exceeds that of the random initialization method, with a significant gap particularly early in the algorithm. This result demonstrates that using knowledge graph-assisted initialization can effectively improve the constraint satisfaction rate of the population, thereby reducing the generation of individuals that do not meet the constraints and improving the optimization efficiency and practical applicability of the algorithm.

[0090] At the same time, the initial population is compared with the traditional randomly generated initial population after 100 generations of iteration under the same parameter environment. The convergence (standard deviation) of the population after each iteration is as follows: Figure 3 As shown by Figure 3As can be seen, the target value of the random initialization method gradually decreases from the first to the 100th generation, but the initial target value is high and the rate of decrease is slow. The knowledge graph-assisted initialization method significantly outperforms random initialization in the first generation of target value and maintains a faster rate of decrease in subsequent iterations, ultimately reaching a lower target value. Starting around the 25th generation, the target value of the knowledge graph-assisted initialization method tends to stabilize, indicating that the algorithm converges quickly. This shows that the knowledge graph-assisted initialization method can provide a high-quality initial solution, accelerate the convergence of the algorithm, and improve optimization performance.

[0091] This embodiment uses a deep Q network to dynamically adjust the genetic algorithm crossover and mutation parameters. The crossover rate and mutation rate of each generation in the iterative process are as follows: Figure 4 As shown by Figure 4 As can be seen, in the early stages of the iteration (exploration phase), the algorithm aims to extensively explore the solution space to discover possible high-quality solutions. Therefore, the mutation rate is high (initially 0.3) to increase population diversity and thus explore more unknown areas. The crossover rate is initially relatively low (initially 0.7) to reduce overexploitation of local solutions and avoid premature convergence. In the mid-to-late stages of the iteration (exploitation phase), as the iterations progress, the population gradually approaches the Pareto front. At this point, more exploitation (local optimization) operations help improve the quality and accuracy of the solution. Therefore, the crossover rate is gradually increased (to 0.8) to more effectively combine the characteristics of existing high-quality solutions. At the same time, the mutation rate is gradually decreased (to 0.1) to minimize damage to the structure of high-quality solutions and reduce the generation of invalid solutions.

[0092] At the same time, the convergence degree (standard deviation) of each generation of the traditional NSGA-Ⅱ and NSGA-Ⅱ+DQN algorithms in the iterative process of this patent is as follows: Figure 5 As shown by Figure 5 As can be seen, the standard deviation of traditional NSGA-II decreases slowly in the early stages, and convergence slows in the middle and late stages. The standard deviation of NSGA-II + DQN decreases more rapidly, demonstrating stronger convergence, especially in the early and middle stages. By the 100th generation, the standard deviation of traditional NSGA-II is still around 40, while the standard deviation of NSGA-II + DQN has dropped to around 20, demonstrating better solution distribution concentration and stability.

[0093] S104, determining the coal combustion coefficient of the thermal power unit according to the cubic function simulation curve, and calculating the carbon emissions of the thermal power unit according to the coal combustion coefficient and the optimal output.

[0094] Specifically, in an implementation scheme, the carbon emission of the thermal power unit is calculated according to the coal consumption factor and the optimal output, specifically: the coal consumption of the thermal power unit under the optimal output is determined according to the coal consumption factor; the carbon emission of the thermal power unit is calculated according to the coal consumption and a preset carbon emission factor.

[0095] Specifically, the coal consumption factor of the thermal power unit is determined by calculating the fuel consumption based on the fitted cubic function simulation curve. The optimal output of the thermal power unit is converted into the coal consumption of the thermal power unit through the determined coal consumption factor of the thermal power unit, and combined with the carbon emission factor to finally convert into the optimized carbon emission, specifically, carbon emission = coal consumption x carbon emission factor.

[0096] S105, each thermal power unit is taken as an entity node of the knowledge graph, and a carbon emission threshold is configured for each entity node, and the carbon emission of the thermal power unit is monitored based on the carbon emission of the thermal power unit and the carbon emission threshold.

[0097] In the present embodiment, the monitoring of the carbon emission of the thermal power unit based on the knowledge graph is described in detail:

[0098] (1) Define the entity class and attribute of the knowledge graph:

[0099] Thermal Power Unit (Thermal Power Unit): unitID (unique identifier), unitName (unit name), capacity (installed capacity), emissionFactor (carbon emission factor);

[0100] Carbon emission threshold (EmissionThreshold): thresholdID, thresholdValue.

[0101] Carbon emission monitoring data (EmissionData): dataID, emissionValue (real-time emission value), timestamp (timestamp).

[0102] Dispatching state (UnitStatus): statusID, statusColor (green / red), adjustmentRequired (whether adjustment is required).

[0103] Adjustment suggestion (AdjustmentPlan): planID, suggestion (specific adjustment suggestion, such as reducing load by 10%).

[0104] (2) Define the relationship of the knowledge graph:

[0105] ThermalPowerUnit hasThreshold → EmissionThreshold: Each thermal power unit (ThermalPowerUnit) has a corresponding carbon emission threshold (EmissionThreshold), which represents the maximum carbon emissions allowed for that unit. When the thermal power unit is running, its actual carbon emissions value needs to be compared with this threshold to determine whether it exceeds the standard.

[0106] ThermalPowerUnit hasEmissionData → EmissionData: Each thermal power unit generates carbon emission monitoring data (EmissionData) related to its operation, including real-time carbon emissions values and timestamps. This data is used to monitor the unit's running state in real time and analyze whether its carbon emissions meet environmental standards.

[0107] ThermalPowerUnit hasStatus → UnitStatus: Each thermal power unit has a status (UnitStatus) that is identified by color (green for normal, red for exceeding the standard). This status is used to describe whether its carbon emissions during operation meet the regulations. Through the knowledge graph, the running state of each unit can be visually displayed to help operation and maintenance personnel quickly identify units that exceed the standard.

[0108] ThermalPowerUnit needsAdjustment → AdjustmentPlan: If the carbon emission status of a thermal power unit is exceeding the standard (red), a corresponding adjustment plan (AdjustmentPlan) needs to be generated to reduce its carbon emissions. The system generates adjustment suggestions based on the unit's status and real-time emissions values, such as reducing output or prioritizing low-emission units.

[0109] (3) Modeling tools and formats: Use Protégé tool to establish OWL ontology. Data storage format is RDF or Turtle.

[0110] (4) Data collection

[0111] The data sources are coal consumption (t / h), power generation (MWh), carbon emission coefficient (kg CO2 / t coal), and time dimension data (sampled every hour or minute). Real-time data is obtained through the monitoring system of thermal power units (such as SCADA), and the pandas and requests libraries in Python are used to connect data interfaces and complete data pulling.

[0112] (5) Data preprocessing

[0113] After cleaning the data by removing missing values or invalid data, perform denoising (use mean or median filling). Then standardize the data: unify units (e.g., convert coal consumption to standard units t / h), and calculate actual carbon emissions:

[0114] Carbon emissions (kg CO2 / h) = Coal consumption (t / h) x Carbon emission factor (kg CO2 / t)

[0115] (6) Data mapping to knowledge graph

[0116] Map each thermal power unit's data to the corresponding entity in the knowledge graph. Update the UnitStatus entity after comparing the real-time carbon emissions with the threshold. Use the data import function of GraphDB or Neo4j. Use Apache Jena or RDF4J for data mapping.

[0117] (7) Rule and reasoning design

[0118] State monitoring rules, use rule reasoning engine (such as Drools or Neo4j APOC) to define the following rules:

[0119] State update rule: if carbon emissions > carbon emission threshold, then update UnitStatus.statusColor = red. If carbon emissions ≤ carbon emission threshold, then update UnitStatus.statusColor = green.

[0120] Adjustment suggestion rule: if UnitStatus.statusColor = red, generate adjustment suggestions, such as reducing the output power of the thermal power unit to reduce carbon emissions. The adjustment priority is based on the emissions of the thermal power unit, with high-emission units being adjusted first.

[0121] Use SPARQL query rules and description logic-based reasoning engine (such as HermiT or Pellet) to perform emissions state reasoning and generate adjustment suggestions.

[0122] (8) Visualization system design

[0123] The running state of the thermal power unit is displayed in the form of a node, and the color represents the emission state (green / red). The real-time data dashboard is used to display the real-time emissions value and threshold of each unit.

[0124] (9) Automatic adjustment model design

[0125] Adjustment scheme generation: dynamically reduce the output of high-emission units based on the carbon emission state.

[0126] Power balance constraint: the total output needs to meet the load demand.

[0127] Adjustment feedback mechanism: after each adjustment, real-time update of carbon emission data, and automatic writing of adjusted data to the knowledge graph.

[0128] Based on the above description, the power plant carbon emission monitoring method provided by the embodiment of the application organizes the operation data, carbon emission threshold and state information of the power plant in the form of a knowledge graph using semantic association and reasoning technology. The essence is to simplify and automate the complex power plant carbon emission monitoring and adjustment process using the modeling, semantic reasoning and visualization capabilities of the knowledge graph. The effect is reflected in the aspects of improved monitoring efficiency, accurate decision support, strong scalability and low-carbon target assistance, and is an effective means of intelligent energy management.

[0129] The actual demonstration effect of using the power plant carbon emission monitoring method based on the knowledge graph is shown in FIG. Figure 7 Figure 7 Each node is a power plant entity, and the green, yellow and red nodes represent the state of the relevant entity nodes being below, close to and exceeding the carbon emission threshold after reasoning and calculation in the knowledge graph. Through the semantic association function of the knowledge graph, the over-standard unit can be quickly located, replacing the traditional table format monitoring method, greatly shortening the data analysis time, and being able to update the carbon emission data in real time and make state judgments according to the threshold, providing targeted adjustment plans, thereby reducing carbon emissions.

[0130] The embodiment of the application also provides an electronic device. The electronic device includes a processor, a memory, a communication interface and at least one communication bus for connecting the processor, the memory and the communication interface. The memory includes but is not limited to random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (PROM) or portable read-only memory (CD-ROM), and is used for related instructions and data.

[0131] ​The communication interface is configured to receive and send data. The processor can be one or more CPUs, and in the case of one CPU, the CPU can be a single-core CPU or a multi-core CPU. The processor in the electronic device is configured to read one or more programs stored in the memory and perform the following operations: obtaining a knowledge graph constructed by operation parameters and historical operation data of a thermal power unit and a cubic function simulation curve of coal consumption fitted by the historical operation data of the thermal power unit, wherein the historical operation data includes load values and coal consumption corresponding to the load values; screening and analyzing an initial solution set and a high-quality solution set that meet operation constraints of the thermal power unit from the knowledge graph, fusing the initial solution set and the high-quality solution set to obtain a solution set; performing multi-objective optimization on the solution set by using a genetic algorithm, and adjusting a crossover probability and a mutation probability of each population iteration of the genetic algorithm by using a deep Q network until the genetic algorithm reaches a maximum iteration number or a population convergence condition, and outputting optimal output of the thermal power unit; determining a coal consumption factor of the thermal power unit according to the cubic function simulation curve, and calculating carbon emissions of the thermal power unit according to the coal consumption factor and the optimal output; regarding each thermal power unit as an entity node of the knowledge graph, and configuring a carbon emission threshold for each entity node, and monitoring carbon emissions of the thermal power unit based on the carbon emissions of the thermal power unit and the carbon emission threshold.

[0132] It should be noted that the specific implementation of each operation can be described above Figure 1 It should be noted that the specific implementation of each operation can be described above

[0133] The embodiment of the present application further provides a computer readable storage medium, which is a memory device in a computer device and is used for storing programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the computer device, and of course can include an extended storage medium supported by the computer device. The computer readable storage medium provides a storage space, which stores an operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory or a non-volatile memory, for example, at least one disk memory. The one or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the power plant carbon emission monitoring method in the above embodiment. It should be understood by those skilled in the art that the embodiment of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0134] The embodiment of the present application further provides a computer program product containing program instructions. The computer program product can be a software or program product containing program instructions, which can be run on a computing device or stored in any available medium. When the computer program product is run on at least one electronic device, the at least one electronic device is caused to perform a power plant carbon emission monitoring method.

[0135] The above detailed description further explains the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above detailed description is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for monitoring carbon emissions from thermal power plants, characterized in that: The method comprises: Obtaining a knowledge graph constructed from operating parameters and historical operating data of a thermal power unit, and a cubic function simulation curve of coal consumption fitted from the historical operating data of the thermal power unit; wherein the historical operating data includes load values ​​and coal consumption corresponding to the load values; Screening and analyzing the initial solution set and high-quality solution set that meet the operating constraints of the thermal power unit from the knowledge graph, and fusing the initial solution set and the high-quality solution set to obtain the solution set; A genetic algorithm is used to perform multi-objective optimization on the solution set, and a deep Q network is used to adjust the crossover probability and mutation probability of each population iteration of the genetic algorithm until the genetic algorithm reaches the maximum number of iterations or the population convergence condition, and then the optimal output of the thermal power unit is output; wherein, the weights of the deep Q network are randomly initialized, and the crossover probability and mutation probability are randomly selected; wherein, the input parameters of the genetic algorithm are used as the state function of the deep Q network, the adjustment strategy of the crossover probability and mutation probability of the genetic algorithm is used as the action function of the deep Q network, and the population quality of the genetic algorithm is used as the reward function of the deep Q network; in each population iteration of the genetic algorithm, according to the randomly selected The crossover probability and mutation probability of the selected population are used to calculate the population state at the current moment, and the action function outputs the crossover probability and mutation probability at the current moment. The population state at the next moment is calculated based on the crossover probability and mutation probability at the current moment, and the population quality is calculated based on the reward function. The population state at the current moment, the crossover probability and mutation probability at the current moment, the population quality, and the population state at the next moment are randomly extracted to update the weight of the deep Q network. The crossover probability and mutation probability of the next population iteration of the genetic algorithm are readjusted based on the deep Q network after the weight update. After the genetic algorithm reaches the maximum number of iterations or the population convergence condition, the optimal output of the thermal power unit is output. The coal combustion coefficient of the thermal power unit is determined based on the cubic function simulation curve, and the carbon emissions of the thermal power unit are calculated based on the coal combustion coefficient and the optimal output; Each thermal power unit is regarded as an entity node in the knowledge graph, and a carbon emission threshold is configured for each entity node. The carbon emissions of the thermal power unit are monitored based on the carbon emissions and carbon emission threshold of the thermal power unit.

2. A method for monitoring carbon emissions from a thermal power plant according to claim 1, characterized in that: Obtain the cubic function fitting curve of coal consumption fitted by the historical operating data of the thermal power unit, specifically: With coal consumption as the target variable and load value as the independent variable, a cubic curve is generated; The cubic function simulation curve of coal consumption was obtained by fitting the parameters of the cubic curve using the least square method.

3. The method for monitoring carbon emissions from a thermal power plant according to claim 1, wherein: The initial solution set and high-quality solution set that meet the operating constraints of the thermal power unit are screened and analyzed from the knowledge graph, specifically: Filter out the initial solution set that meets the operating constraints of the thermal power units from the optimal operating modes of different thermal power units under different operating conditions in the knowledge graph; and The historical operating data of thermal power units is extracted from the knowledge graph, including the historical optimal solutions of power generation cost, load demand, fuel consumption and carbon emissions. Based on the historical optimal solutions and the preset rule base, a set of high-quality solutions that meet the operating constraints of the thermal power units is analyzed.

4. A method for monitoring carbon emissions from a thermal power plant according to claim 1, characterized in that: The optimization objectives of the genetic algorithm include maximizing the combined wind and solar power output, minimizing the net load fluctuation, minimizing the system operation cost and minimizing the carbon emission.

5. A method for monitoring carbon emissions from a thermal power plant according to claim 1, characterized in that: The input parameters of the genetic algorithm include the number of non-dominated solutions, a population distribution diversity index, a super volume index and the number of population iterations.

6. A method for monitoring carbon emissions from a thermal power plant according to claim 1, characterized in that: The carbon emissions of thermal power units are calculated based on the coal combustion coefficient and optimal output, specifically: Determine the coal burning amount of the thermal power unit at the optimal output according to the coal burning coefficient; The carbon emissions of thermal power units are calculated based on the amount of coal burned and the preset carbon emission factor.

7. An electronic device, characterized in that: including memory and processor; a memory for storing a computer program, wherein the computer program includes program instructions; The processor is used to execute the program instructions so that the electronic device performs the steps of the method for monitoring carbon emissions of a thermal power unit as described in any one of claims 1 to 6.

8. A computer program product comprising program instructions, characterized in that When the program instructions are executed by the electronic device, the electronic device executes the steps of the method for monitoring carbon emissions of a thermal power unit as claimed in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a computer program, and when the computer program is executed by one or more processors, it implements the method for monitoring carbon emissions of a thermal power unit according to any one of claims 1 to 6.

Images