Industrial carbon emission control method, device, equipment and medium

Abstract

Some embodiments of the invention disclose an industrial carbon emission control method, device, equipment and medium, the method collects and analyzes a large amount of data in an industrial production process, uses an artificial intelligence algorithm to establish a target neural network model, the model can accurately predict the carbon emission, and the carbon emission can be accurately controlled. An effective optimization strategy can be quickly generated by utilizing a dynamic parameter multi-objective differential evolution algorithm, the multi-objective differential evolution algorithm comprehensively considers factors of carbon emission reduction, cost control, energy consumption control and the like, carbon emission can be reduced to the greatest extent on the premise of meeting the requirements of enterprise economic benefits and product quality, and the economic benefit of enterprises is improved. According to the method, the limitation of traditional single target optimization is broken through, the production benefit and the product quality are not excessively sacrificed while an enterprise pursues energy conservation and emission reduction, coordination and unification of economic benefits and environmental benefits are achieved, and higher practicability and operability are achieved.

Description

Technical Field

[0001] This application relates to the technical field of carbon emissions, and particularly to an industrial carbon emission control method, device, equipment and medium. Background Art

[0002] The problem of carbon emissions in industrial production is becoming increasingly serious, posing a huge pressure on the environment. Traditional carbon emission control methods mainly rely on empirical models and manual parameter tuning. However, when facing complex carbon emission systems and changing environmental factors, these methods often fail to achieve ideal optimization effects.

[0003] Therefore, how to improve the accuracy and efficiency of industrial carbon emission control has become a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0004] This application provides an industrial carbon emission control method, device, electronic equipment and storage medium. This method collects and analyzes a large amount of data in the industrial production process, uses artificial intelligence algorithms to establish a target neural network model, which can accurately predict carbon emissions, and uses a dynamic parameter multi-objective differential evolution algorithm to quickly generate effective optimization strategies. This multi-objective differential evolution algorithm comprehensively considers various factors such as carbon emission reduction, cost control, and energy consumption control, and can minimize carbon emissions to the greatest extent on the premise of meeting the economic benefits and product quality requirements of enterprises. This breaks through the limitations of traditional single-objective optimization, enabling enterprises to pursue energy conservation and emission reduction without sacrificing production efficiency and product quality excessively, achieving the coordinated unity of economic benefits and environmental benefits, and having stronger practicability and operability.

[0005] In the first aspect, an embodiment of this application provides an industrial carbon emission control method, including:

[0006] Obtain target relevant factor data that affects carbon emissions;

[0007] Input the target relevant factor data into a target neural network model to output predicted carbon emissions; where the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model;

[0008] Define a multi-objective optimization function, which is determined according to the objective function of carbon emission reduction, the objective function of economic cost, and the objective function of energy consumption;

[0009] Based on the predicted carbon emissions and the multi-objective optimization function, use a dynamic parameter multi-objective differential evolution algorithm to determine the control parameters of industrial carbon emissions.

[0010] In some embodiments, the method further includes:

[0011] Obtain historical carbon emissions and data on historical relevant factors that affect the historical carbon emissions;

[0012] Preprocess the historical carbon emissions and the data on historical relevant factors;

[0013] Use the preprocessed historical carbon emissions and the preprocessed data on historical relevant factors to construct a grey model;

[0014] Use the grey model to predict future carbon emissions and data on future relevant factors that affect the future carbon emissions;

[0015] Determine a training set according to the preprocessed historical carbon emissions, the preprocessed data on historical relevant factors, the future carbon emissions, and the data on future relevant factors; wherein, the training set includes first training data and second training data, the first training data includes the preprocessed historical carbon emissions and the preprocessed data on historical relevant factors, and the second training data includes the future carbon emissions and the data on future relevant factors;

[0016] Use the first training data and the second training data to train a BP neural network model and an LSTM neural network model, and correspondingly obtain a target BP neural network model and a target LSTM neural network model;

[0017] Combine the target BP neural network model and the target LSTM neural network model to obtain a target neural network model.

[0018] In some embodiments, the preprocessed data on historical relevant factors and the data on future relevant factors respectively include production volume, energy consumption, temperature, and time;

[0019] The step of using the first training data and the second training data to train a BP neural network model and an LSTM neural network model, and correspondingly obtaining a target BP neural network model and a target LSTM neural network model includes:

[0020] Use the first training data and the second training data to train a BP neural network model to adjust the network weights corresponding to the production volume, energy consumption, and temperature in the BP neural network model, and obtain a target BP neural network model;

[0021] Use the first training data and the second training data to train an LSTM neural network model to enable the LSTM neural network model to learn the temporal law of carbon emissions according to the time, and obtain a target LSTM neural network model.

[0022] In some embodiments, the multi-objective function is the sum of the objective function of carbon emission reduction multiplied by a first coefficient, the objective function of economic cost multiplied by a second coefficient, and the objective function of energy consumption multiplied by a third coefficient.

[0023] In some embodiments, the objective function of carbon emission reduction is the difference between the predicted carbon emissions and the optimized carbon emissions, divided by the predicted carbon emissions.

[0024] The objective function of economic cost is the economic cost data determined according to the optimized carbon emissions, divided by the maximum economic cost;

[0025] The objective function of energy consumption is the energy consumption determined according to the optimized carbon emissions, divided by the maximum energy consumption.

[0026] In some embodiments, the method further includes:

[0027] The target LSTM neural network model compares the difference between the target carbon emissions predicted by the target LSTM neural network model and the current carbon emissions with a preset difference; if the difference exceeds the preset difference, it is determined that the current carbon emissions are abnormal.

[0028] In some embodiments, the step of preprocessing the historical carbon emissions and the historical related factor data includes:

[0029] Removing invalid data and abnormal data in the historical carbon emissions and the historical related factor data, filling in missing data, and performing data standardization processing.

[0030] In a second aspect, an industrial carbon emission control device provided by an embodiment of the present application includes:

[0031] An acquisition unit, configured to acquire target related factor data affecting carbon emissions;

[0032] An input unit, configured to input the target related factor data into a target neural network model to output predicted carbon emissions; wherein the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model;

[0033] A definition unit, configured to define a multi-objective optimization function, where the multi-objective optimization function is determined according to an objective function of carbon emission reduction, an objective function of economic cost, and an objective function of energy consumption;

[0034] A determination unit, configured to determine control parameters for industrial carbon emissions based on the predicted carbon emissions and the multi-objective optimization function by using a dynamic parameter multi-objective differential evolution algorithm.

[0035] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the industrial carbon emission control method are implemented.

[0036] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the industrial carbon emission control method are implemented.

[0037] In the above embodiments, an industrial carbon emission control method, device, electronic device, and storage medium are provided. The method collects and analyzes a large amount of data in the industrial production process, uses an artificial intelligence algorithm to establish a target neural network model. This model can accurately predict carbon emissions, and uses a dynamic parameter multi-objective differential evolution algorithm to quickly generate effective optimization strategies. This multi-objective differential evolution algorithm comprehensively considers multiple factors such as carbon emission reduction, cost control, and energy consumption control. It can maximize carbon emission reduction on the premise of meeting the economic benefits and product quality requirements of the enterprise, breaking the limitations of traditional single-objective optimization. It enables the enterprise to pursue energy conservation and emission reduction without sacrificing production efficiency and product quality excessively, achieving the coordinated unity of economic benefits and environmental benefits, and having stronger practicability and operability. The method includes: obtaining target-related factor data that affects carbon emissions; inputting the target-related factor data into the target neural network model to output the predicted carbon emissions; where the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model; defining a multi-objective optimization function, and the multi-objective optimization function is determined according to the objective function of carbon emission reduction, the objective function of economic cost, and the objective function of energy consumption; based on the predicted carbon emissions and the multi-objective optimization function, using the dynamic parameter multi-objective differential evolution algorithm to determine the control parameters of industrial carbon emissions. Description of the Drawings

[0038] Figure 1 Exemplarily shows a flowchart of an industrial carbon emission control method provided according to some embodiments;

[0039] Figure 2 Exemplarily shows a flowchart of another industrial carbon emission control method provided according to some embodiments;

[0040] Figure 3 Exemplarily shows a structural schematic diagram of an industrial carbon emission control device provided according to some embodiments. Detailed Embodiments

[0041] To make the purpose and implementation manner of this application clearer, the following will clearly and completely describe the exemplary implementation manner of this application in conjunction with the drawings in the exemplary embodiments of this application. Obviously, the described exemplary embodiments are only a part of the embodiments of this application, rather than all of the embodiments.

[0042] It should be noted that the brief description of the terms in this application is only for the convenience of understanding the subsequent described implementation manner, rather than intending to limit the implementation manner of this application. Unless otherwise specified, these terms should be understood in their ordinary and common meanings.

[0043] In this application, the terms "first", "second", "third", etc. in the specification, claims and the above drawings are used to distinguish similar or homogeneous objects or entities, and do not necessarily mean to limit a specific order or sequence, unless otherwise noted. It should be understood that such terms can be interchanged under appropriate circumstances.

[0044] The terms "include" and "have" and any of their variants are intended to cover but not exclude inclusion. For example, a product or device that includes a series of components does not necessarily have to be limited to all the components clearly listed, but may include other components that are not clearly listed or are inherent to these products or devices.

[0045] The problem of carbon emissions in industrial production processes is becoming increasingly serious, posing a huge pressure on the environment. Traditional carbon emission control methods mainly rely on empirical models and manual parameter tuning. However, when facing complex carbon emission systems and changing environmental factors, these methods often fail to achieve ideal optimization effects. Specifically, first, industrial production involves multiple interacting variables, and it is difficult for related technologies to comprehensively consider these complex relationships. Carbon emissions have non-linear characteristics, and it is difficult for traditional linear models to accurately describe them; second, environmental factors such as temperature and humidity change frequently, and it is difficult for traditional static models to adapt in real time. External disturbances and internal fluctuations increase the uncertainty of carbon emissions, and it is difficult for related technologies to effectively respond; third, related technologies rely on manual parameter tuning, which is highly subjective and difficult to ensure the optimal solution of carbon emissions. Empirical models are effective under specific conditions, but perform poorly when facing new situations; fourth, related technologies are difficult to adjust control strategies in real time, resulting in delayed optimization effects. Complex model calculations are time-consuming and cannot meet the real-time control requirements. Therefore, how to improve the accuracy and efficiency of industrial carbon emission control has become a technical problem that needs to be urgently solved by those skilled in the art.

[0046] To solve the above technical problems, the embodiments of the present application provide an industrial carbon emission control method, device, electronic device, and storage medium. The method collects and analyzes a large amount of data in the industrial production process, uses an artificial intelligence algorithm to establish a target neural network model, which can accurately predict carbon emissions, and uses a dynamic parameter multi-objective differential evolution algorithm to quickly generate effective optimization strategies. The multi-objective differential evolution algorithm comprehensively considers multiple factors such as carbon emission reduction, cost control, and energy consumption control, and can minimize carbon emissions to the greatest extent on the premise of meeting the enterprise's economic benefits and product quality requirements, breaking the limitations of traditional single-objective optimization. It enables enterprises to pursue energy conservation and emission reduction without sacrificing production efficiency and product quality excessively, achieving the coordinated unity of economic benefits and environmental benefits, and having stronger practicability and operability.

[0047] Figure 1 Exemplarily shown is a flowchart of an industrial carbon emission control method provided according to some embodiments. The method includes S100 - S400.

[0048] S100. Obtain data on target-related factors affecting carbon emissions.

[0049] In the embodiments of the present application, the data on target-related factors affecting carbon emissions involve factors such as society, economy, and environment. In some embodiments, the target-related factor data may include production volume, energy consumption, temperature, and time data. The production volume generally refers to the total output in the industrial production process, such as the output of the manufacturing industry (unit: ton / unit / set), the power generation of the power plant (unit: kilowatt-hour), the product output of the chemical enterprise (unit: cubic meter / liter), and specific industries can adjust the definition according to the application scenario. The energy consumption can be the energy consumption generated in the industrial production process. The temperature refers to the ambient temperature. The carbon emission is in CO2 equivalent, which can be specifically obtained through direct monitoring or conversion of energy consumption data, and the common units are: ton CO2 / hour, ton CO2 / month, etc. The above disclosed relevant factor data are parameters with relatively high importance affecting carbon emissions. Of course, the relevant factor data may also include other parameters affecting carbon emissions.

[0050] In the embodiments of the present application, during the industrial production process, various data related to carbon emissions (i.e., data on target-related factors affecting carbon emissions) are collected, which may specifically include energy consumption, production volume, and temperature, etc. In the subsequent process of the method of the embodiments of the present application, the complex relationships of various data related to carbon emissions can be comprehensively considered, and ultimately, industrial carbon emissions can be accurately controlled.

[0051] In the embodiments of the present application, in order to continuously and accurately control industrial carbon emissions, the current carbon emissions can be detected in real time and the data on target-related factors affecting carbon emissions can be obtained in real time, so as to dynamically optimize and adjust the industrial carbon emission control.

[0052] S200. Input the target relevant factor data into a target neural network model to output a predicted carbon emission; wherein the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model.

[0053] In the embodiment of the present application, the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model. The target BP neural network model is used for non-linear modeling of static data and influence factor analysis. The target LSTM neural network model is used for prediction of time series data and dynamic optimization control. In the embodiment of the present application, combining the target BP neural network model can construct a more powerful prediction and optimization model.

[0054] In the embodiment of the present application, the target neural network model can be determined according to the following steps. In some embodiments, Figure 2 Exemplarily shown is a flowchart of another industrial carbon emission control method provided according to some embodiments. The method further includes S500 - S1100.

[0055] S500. Obtain historical carbon emissions and historical relevant factor data that affect the historical carbon emissions.

[0056] In this embodiment, the historical carbon emissions refer to the actual carbon emissions generated historically. The historical relevant factor data refers to the actual relevant factor data that affect the historical carbon emissions, which may include production volume, energy consumption, temperature, and time data. It should be noted that the number of historical carbon emissions and historical relevant factor data can be several.

[0057] It should be noted that the target neural network model in the embodiment of the present application can also be continuously updated. As time goes by, the historical carbon emissions and historical relevant factor data continue to increase. The new added historical carbon emissions and historical relevant factor data and the previous historical carbon emissions and historical relevant factor data can be used to jointly construct a grey model, and subsequently jointly train a BP (Back Propagation Neural Network) neural network model and an LSTM (Long Short-Term Memory) neural network model, and finally obtain an updated target neural network model.

[0058] S600. Preprocess the historical carbon emissions and the historical relevant factor data.

[0059] The preprocessing in this embodiment refers to processing the data with problems in the historical carbon emissions and historical related factor data, so as to ensure that the subsequent historical carbon emissions after preprocessing and the historical related factor data after preprocessing can construct a grey model with better prediction effect, and the target BP neural network model and target LSTM neural network model with better prediction effect can be obtained when training the BP neural network model and LSTM neural network model, and finally the target neural network model with better prediction accuracy can be obtained.

[0060] In some embodiments, the steps of preprocessing the historical carbon emissions and the historical related factor data include: removing invalid data and abnormal data in the historical carbon emissions and the historical related factor data, filling in missing data, and performing data standardization processing.

[0061] The invalid data here may include values beyond physical meaning. For example, if the historical carbon emissions are negative, removing the invalid data can be deleting the historical carbon emissions and the historical related factor data affecting the historical carbon emissions, or reasonably filling in the historical carbon emissions data. Filling in missing data can be reasonably inferring the missing data based on other data. For example, if the energy consumption data is missing, the missing energy consumption data can be inferred and filled in according to the production volume, temperature and time data. The data standardization processing can be min-max normalization processing.

[0062] S700. Use the historical carbon emissions after preprocessing and the historical related factor data after preprocessing to construct a grey model (GM(1, N) model).

[0063] In the embodiment of the present application, the GM(1, N) model is a grey model containing N variables with a first-order equation, which can be used to describe a multi-variable system. The grey model can comprehensively consider multiple variables that affect each other in industrial production (including the historical carbon emissions after preprocessing and the historical related factor data after preprocessing). The multiple variables that affect each other have non-linear characteristics. The GM(1, N) model can accurately describe the non-linear characteristics and accurately predict the future carbon emissions and the future related factor data that affect the future carbon emissions.

[0064] S800. Use the grey model to predict the future carbon emissions and the future related factor data that affect the future carbon emissions.

[0065] In this embodiment, the grey model can be used to relatively accurately predict the future carbon emissions and the corresponding future related factor data. In this embodiment, since the amount of data of the historical carbon emissions and the historical related factor data is small in actual industrial production, in this embodiment, the grey model can use the data with a small amount of data to predict the future carbon emissions and the future related factor data with relatively high accuracy.

[0066] S900. Determine a training set based on the preprocessed historical carbon emissions, preprocessed historical related factor data, future carbon emissions, and future related factor data. The training set includes first training data and second training data. The first training data includes the preprocessed historical carbon emissions and the preprocessed historical related factor data, and the second training data includes the future carbon emissions and the future related factor data.

[0067] S1000. Use the first training data and the second training data to train a BP neural network model and an LSTM neural network model, and correspondingly obtain a target BP neural network model and a target LSTM neural network model.

[0068] In this embodiment, a BP neural network model and an LSTM neural network model can be trained using historical data (preprocessed historical carbon emissions and preprocessed historical related factor data) and future data (future carbon emissions and future related factor data). In this embodiment, in order to further accurately analyze the relationship between carbon emissions and related factor data on the basis of the gray model, the BP neural network model and the LSTM neural network model are used to further analyze the relationship between carbon emissions and related factor data.

[0069] In the embodiment of this application, the BP neural network is a classic feedforward neural network. By adjusting the network parameters through the backpropagation algorithm, it can fit complex non-linear relationships. The target BP neural network obtained by such training establishes a non-linear model between carbon emissions and industrial production process variables through non-linear modeling, supporting real-time optimization control. The target BP neural network can be used for non-linear modeling and influence factor analysis of static data. The LSTM neural network is a special recurrent neural network (RNN) specifically designed to process time series data and can capture long-term dependencies in the data. The target LSTM neural network model obtained by such training can be used for prediction and dynamic optimization control of time series data. Therefore, the target BP neural network model and the target LSTM neural network model can deeply explore the relationship between carbon emissions and related factor data.

[0070] In this embodiment, each preprocessed historical carbon emission and the preprocessed historical related factor data that affects the preprocessed historical carbon emission are used as a set of first training data. In this way, several sets of first training data will be generated through all the preprocessed historical carbon emissions and the preprocessed historical related factor data. Each future carbon emission and the future related factor data that affects the future carbon emission are used as a set of second training data. In this way, several sets of second training data will also be generated through all the future carbon emissions and the future related factor data. Using all the first training data and the second training data, a BP neural network model and an LSTM neural network model are trained. After training, the BP neural network model forms a target BP neural network model, and the LSTM neural network model forms a target LSTM neural network model.

[0071] S1100. Combine the target BP neural network model and the target LSTM neural network model to obtain a target neural network model.

[0072] In the embodiment of the present application, the target BP neural network model and the target LSTM neural network model can deeply explore the relationship between carbon emissions and related factor data. To further improve the prediction accuracy, the target BP neural network model and the LSTM neural network model are combined to obtain a target neural network model.

[0073] In an example, combining the two models can be completed by setting the weights corresponding to the target BP neural network model and the target LSTM neural network model respectively. For example, the weight of the target BP neural network model is 0.4, and the predicted carbon emission is 200. The weight of the target LSTM neural network model is 0.6, and the predicted carbon emission is 100. Then the predicted carbon emission of the target neural network model is 0.4 * 200 + 0.6 * 100, that is, 140.

[0074] In the embodiment of the present application, appropriate artificial intelligence algorithms (BP neural network and LSTM neural network) are selected to improve the accuracy of the target neural network model in predicting carbon emissions from different perspectives.

[0075] In some embodiments, the preprocessed historical related factor data and the future related factor data respectively include production volume, energy consumption, and temperature. The steps of using the first training data and the second training data to train the BP neural network model and the LSTM neural network model, and correspondingly obtaining the target BP neural network model and the target LSTM neural network model include:

[0076] Using the first training data and the second training data, train a BP neural network model to adjust the network weights corresponding to the production volume, energy consumption, and temperature in the BP neural network model, and obtain a target BP neural network model.

[0077] In this embodiment, the BP neural network is a classic feedforward neural network. By adjusting network parameters through the backpropagation algorithm, it can fit complex non-linear relationships. The target BP neural network model obtained through BP neural network training can predict future carbon emission trends, and the input variables can include influencing factors such as production volume, energy consumption, and temperature. By analyzing the network weights corresponding to the production volume, energy consumption, and temperature after training, analyze the influence degree of each input variable on carbon emissions, identify key influencing factors, and provide a basis for optimal control. Due to the frequent changes of environmental factors such as temperature and humidity, the models in the related technologies are difficult to adapt in real time, and external disturbances increase the uncertainty of carbon emissions. However, the target BP neural network model in the embodiments of the present application can analyze data such as temperature, and improve the prediction accuracy of carbon emissions in the presence of external disturbances.

[0078] Using the first training data and the second training data, train an LSTM neural network model to enable the LSTM neural network model to learn the temporal law of carbon emissions according to the time, and obtain a target LSTM neural network model.

[0079] In this embodiment, the LSTM model is a special recurrent neural network (RNN), specifically designed to process time series data and capable of capturing long-term dependencies in the data. The target LSTM model obtained through LSTM model training can predict future carbon emission trends and is applicable to industrial carbon emission data with obvious time dependencies. In this embodiment, through dynamic optimization control combined with the prediction results of the target LSTM model, industrial production parameters can be adjusted in real time, the carbon emission control strategy can be optimized, and real-time decision-making in a dynamic environment can be supported.

[0080] The target LSTM neural network model is used for time series prediction, and the input is time series data, which is the first training data and the second training data including time. The inputs of the target LSTM neural network model include historical carbon emissions (the preprocessed historical carbon emissions in the first training data and the future carbon emissions in the second training data), temperature, energy consumption, and time features (such as hours, day of the week, etc.). The model outputs the predicted value of the target predicted carbon emissions at the next time step (t) (univariate or multivariate).

[0081] Train the LSTM model using the first training data and the second training data so that it learns the temporal patterns (such as periodicity, trends) of carbon emissions, with the goal of minimizing the prediction error (such as mean squared error, MSE). The model learns to predict the normal carbon emission values at future times, and this normal carbon emission is the target carbon emission with an accuracy within a preset range.

[0082] In some embodiments, the method further includes: comparing the difference between the target carbon emission predicted by the target LSTM neural network model and the current carbon emission with a preset difference; if the difference exceeds the preset difference, it is determined that the current carbon emission is abnormal.

[0083] In this embodiment, the target LSTM neural network model is used to predict the carbon emission at the current time, that is, the target carbon emission. Calculate the difference between the target carbon emission and the current carbon emission, and compare the difference with the preset difference. If the difference exceeds the preset difference, it is marked as abnormal. In this embodiment, the preset difference is the difference between the normal carbon emission predicted based on history and the current carbon emission at the corresponding time of the predicted normal carbon emission, and the meaning of this normal carbon emission has been introduced above.

[0084] In this embodiment, the target LSTM model prediction can detect whether the target predicted carbon emission is in an abnormal situation, so that abnormal situations in the production process can be discovered in time and carbon emission fluctuations can be reduced.

[0085] S300. Define a multi-objective optimization function, which is determined according to the objective function of carbon emission reduction, the objective function of economic cost, and the objective function of energy consumption.

[0086] In the embodiments of the present application, the dynamic parameter multi-objective differential evolution algorithm needs to be used later, and a multi-objective optimization function is required in this algorithm, so the multi-objective optimization function is defined in advance. In the embodiments of the present application, for carbon emission reduction, economic cost, and energy consumption, they are all important parameters that need to be considered for industrial carbon emissions, so the defined multi-objective optimization function involves the objective function of carbon emission reduction, the objective function of economic cost, and the objective function of energy consumption.

[0087] In some embodiments, the objective function is the sum of the objective function of carbon emission reduction multiplied by a first coefficient, the objective function of economic cost multiplied by a second coefficient, and the objective function of energy consumption multiplied by a third coefficient.

[0088] In this embodiment, the multi-objective optimization function can be expressed by the following formula:

[0089] Objective = w1·f1(x) + w2·f2(x) + w3·f3(x);

[0090] Among them, Objective is a multi-objective optimization function, w1 is the first coefficient, f1(x) is the objective function for reducing carbon emissions, w2 is the second coefficient, f2(x) is the objective function for economic cost, w3 is the third coefficient, and f3(x) is the objective function for energy consumption.

[0091] In this embodiment, the sum of the first coefficient, the second coefficient, and the third coefficient is 1. The three coefficients are set according to the actual needs of the user. For example, if the effect of carbon emission reduction is emphasized, the value of the first coefficient can be set relatively high, and the values of the other two coefficients can be set relatively low.

[0092] In some embodiments, the objective function for reducing carbon emissions is the difference between the predicted carbon emissions and the optimized carbon emissions, divided by the predicted carbon emissions. In this embodiment, the optimized carbon emissions are the carbon emissions generated under the scenario corresponding to the control parameters of industrial carbon emissions determined by the dynamic parameter multi-objective differential evolution algorithm (D-MOEA).

[0093] The objective function for economic cost is the economic cost data determined based on the optimized carbon emissions, divided by the maximum economic cost;

[0094] The objective function for energy consumption is the energy consumption determined based on the optimized carbon emissions, divided by the maximum energy consumption.

[0095] In the embodiment of the present application, the hybrid weighted algorithm in the multi-objective optimization function is adopted, which can comprehensively consider various factors such as carbon emission reduction, production cost control, and energy consumption, and generate an optimal production plan and operation parameter combination. This method can avoid the subjectivity of traditional methods relying on manual parameter adjustment, ensure the objectivity and optimal solution of the optimization strategy. Through model verification and feedback adjustment, the effectiveness and reliability of the optimization strategy in practical applications are ensured. The key index verification method is used to conduct feedback verification on the optimization and prediction processes to ensure that the carbon emission value meets the requirements of the multi-objective function and each constraint index.

[0096] In this embodiment, the objective function for reducing carbon emissions can be expressed by the following formula:

[0097]

[0098] Among them, f1(x) is the objective function for reducing carbon emissions, C0 is the predicted carbon emissions, and C(x) is the optimized carbon emissions.

[0099] The objective function for economic cost can be expressed by the following formula:

[0100]

[0101] Among them, f2(x) is the objective function of the economic cost, and E max is the maximum economic cost, and E(x) is the economic cost data determined by the optimized carbon emissions.

[0102] The objective function of energy consumption can be expressed by the following formula:

[0103]

[0104] Among them, f3(x) is the objective function of energy consumption, and P max is the maximum energy consumption, and E(x) is the energy consumption determined by the optimized carbon emissions, that is, the optimized energy consumption.

[0105] S400. Based on the predicted carbon emissions and the multi-objective optimization function, use the dynamic parameter multi-objective differential evolution algorithm (D-MOEA) to determine the control parameters of industrial carbon emissions.

[0106] In the embodiments of the present application, the solution set output by the dynamic parameter multi-objective differential evolution algorithm is the Pareto front (that is, the control parameters of industrial carbon emissions), indicating the trade-off between multiple objectives in the multi-objective optimization function. Specifically, by adjusting the population size, the number of iterations, the crossover probability, and the mutation probability to obtain better results, the optimal balance point between carbon emission minimization, economic cost control, and energy consumption control can be found.

[0107] Specifically, the execution steps of the dynamic parameter multi-objective differential evolution algorithm are as follows:

[0108] (1) Initialize the population

[0109] Randomly generate the initial population: X = x1, x2,..., x N , where each individual x i is a D-dimensional vector: x i = [x i1 , x i2 ,..., x iD , X = x1, x2,..., x N , D is the dimension of the problem, and N is the population size. The individual x i is the control parameter of industrial carbon emissions. The carbon emissions come from the energy consumption data, and the costs include raw materials and operating expenses. In one example, in the energy call scenario, the control parameters may include the output allocation of each unit, the charge and discharge strategy of the energy storage system, the new energy consumption ratio, production process parameters, etc. In another example, the control parameters may be different categories of production plans, processes, equipment, energy, supply chains, etc., and the specific measures under each category may include peak-shifting production, using low-carbon technologies, equipment upgrading, waste heat recovery, etc.

[0110] For each objective individual xi , generate a mutant vector V i :

[0111] V i = x r1 + F · (x r2 - x r3 )(6)

[0112] where: x r1 , x r2 , x r3 are three different individuals randomly selected from the population. F is the mutation factor that controls the scaling ratio of the difference vector.

[0113] (2) Generate a trial vector u through a crossover operation i :

[0114]

[0115] where: rand(0, 1) is a random number with a uniform distribution. CR is the crossover probability that controls the proportion of the mutant vector in the trial vector. j rand . is a randomly selected dimension index to ensure that at least one dimension comes from the mutant vector. Select the individuals of the next generation according to the objective function values:

[0116]

[0117] f(·) is the objective function. In the multi-objective optimization function, the goal is to optimize multiple objective functions f(x) = [f(x1), f(x2),..., f(x M )] simultaneously.

[0118] D-MOEA extends DE (Differential Evolution) in the following way:

[0119] (3) Define the Pareto dominance relationship:

[0120] The solution x1 dominates the solution x2, denoted as x1 < x2, if and only if:

[0121]

[0122] and there exists at least one j such that: f j (x1) < f j (x2)

[0123] Sort the individuals in the population according to the Pareto dominance relationship to generate the non-dominated front (Pareto Front). Calculate the crowding degree of each individual to maintain the diversity of the population:

[0124]

[0125] (4) Dynamic parameter adjustment:

[0126] F t = F min + (F max - F min ) * rand(0,1) (11)

[0127] CR t = CR min + (CR max - CR min ) * rand(0,1) (12)

[0128] Finally, determine the control parameters for industrial carbon emissions.

[0129] In some embodiments, the above algorithm process can also be summarized as follows: including 1. Initializing the population X. 2. Calculating the objective function value f(x i ). 3. Performing non-dominated sorting and crowding degree calculation on the population. 4. Dynamically adjusting the parameters F and CR. 5. Performing mutation, crossover, and selection operations to generate a new population. 6. Repeating step 5 until the termination condition is met.

[0130] The D-MOEA algorithm can combine economic costs for multi-objective optimization, find the balance point between costs and emissions reduction, effectively solve the carbon emission optimization problem in industrial production, and provide a more accurate, efficient, and reliable solution.

[0131] By adopting the D-MOEA algorithm and introducing a parameter adaptive adjustment strategy, the global optimization ability and convergence ability of the algorithm are improved. This method can adapt to changes in environmental factors in real time, such as temperature, humidity, etc., and effectively cope with external interference and internal fluctuations. By real-time monitoring of carbon emission data and production operation conditions, the optimization strategy is dynamically adjusted according to the feedback results. This closed-loop feedback mechanism can timely discover and solve problems that may occur in the actual application of the optimization strategy, ensuring the persistence and stability of the optimization effect.

[0132] The D-MOEA algorithm can complete the optimization calculation in a short time to meet the requirements of real-time control. This method not only improves the calculation efficiency but also ensures the real-time adjustment and implementation of the optimization strategy. Through real-time scheduling and optimization technologies, based on the dynamic adjustment thread pool method of real-time monitoring, it can effectively manage computing resources, improve the response speed and processing ability of the system, and ensure the real-time and effectiveness of the optimization strategy.

[0133] In the above embodiments, an industrial carbon emission control method, device, electronic device, and storage medium are provided. The method collects and analyzes a large amount of data in the industrial production process, uses an artificial intelligence algorithm to establish a target neural network model, which can accurately predict carbon emissions, and uses a dynamic parameter multi-objective differential evolution algorithm to quickly generate effective optimization strategies. The multi-objective differential evolution algorithm comprehensively considers various factors such as carbon emission reduction, cost control, and energy consumption control. It can minimize carbon emissions to the greatest extent on the premise of meeting the enterprise's economic benefits and product quality requirements, breaking the limitations of traditional single-objective optimization. It enables enterprises to pursue energy conservation and emission reduction without overly sacrificing production efficiency and product quality, achieving the coordinated unity of economic benefits and environmental benefits, and having stronger practicability and operability.

[0134] The embodiment of the present application also provides an industrial carbon emission control device; Figure 3 An exemplary structural schematic diagram of an industrial carbon emission control device provided according to some embodiments is shown. The device includes an acquisition unit 301, an input unit 302, a definition unit 303, and a determination unit 304.

[0135] The acquisition unit is used to acquire target-related factor data that affects carbon emissions;

[0136] The input unit is used to input the target-related factor data into the target neural network model to output the predicted carbon emissions; where the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model;

[0137] The definition unit is used to define a multi-objective optimization function, and the multi-objective optimization function is determined according to the objective function for carbon emission reduction, the objective function for economic cost, and the objective function for energy consumption;

[0138] The determination unit is used to determine the control parameters of industrial carbon emissions based on the predicted carbon emissions and the multi-objective optimization function by using a dynamic parameter multi-objective differential evolution algorithm.

[0139] The embodiment of the present application also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the industrial carbon emission control method are implemented.

[0140] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the industrial carbon emission control method are implemented.

[0141] Through the description of the above embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus a necessary general hardware platform, or can also be implemented by hardware.

[0142] Those skilled in the art can understand that the drawings are only schematic diagrams of a preferred embodiment scenario, and the units or processes in the drawings are not necessarily essential for implementing this application. Those skilled in the art can understand that the units in the devices in the embodiment scenario can be distributed in the devices in the embodiment scenario according to the description of the embodiment scenario, or can be correspondingly changed and located in one or more devices different from this embodiment scenario. The units in the above embodiment scenario can be combined into one unit, or can be further split into multiple sub-units.

[0143] The above serial numbers of this application are only for description and do not represent the advantages or disadvantages of the embodiment scenarios. The above disclosure is only several specific embodiment scenarios of this application. However, this application is not limited thereto, and any changes that can be thought of by those skilled in the art should fall within the protection scope of this application.

Claims

1. An industrial carbon emission control method, characterized in that: include: Obtain target-related data on factors that affect carbon emissions; Inputting the target-related factor data into a target neural network model to output predicted carbon emissions; wherein the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model; defining a multi-objective optimization function, wherein the multi-objective optimization function is determined according to an objective function of carbon emission reduction, an objective function of economic cost, and an objective function of energy consumption; Based on the predicted carbon emissions and the multi-objective optimization function, a dynamic parameter multi-objective differential evolution algorithm is used to determine control parameters for industrial carbon emissions.

2. The method according to claim 1, characterized in that Also includes: Obtaining historical carbon emissions and historical relevant factor data affecting the historical carbon emissions; Preprocessing the historical carbon emissions and the historical related factor data; The grey model is constructed by using the pre-processed historical carbon emissions and pre-processed historical related factor data; Using the grey model to predict future carbon emissions and future related factor data affecting the future carbon emissions; Determine a training set according to the preprocessed historical carbon emissions, the preprocessed historical relevant factor data, the future carbon emissions and the future relevant factor data; wherein the training set includes first training data and second training data, the first training data includes the preprocessed historical carbon emissions and the preprocessed historical relevant factor data, and the second training data includes the future carbon emissions and the future relevant factor data; Using the first training data and the second training data, training a BP neural network model and an LSTM neural network model, and obtaining a target BP neural network model and a target LSTM neural network model accordingly; The target BP neural network model and the target LSTM neural network model are combined to obtain a target neural network model.

3. The method according to claim 2, characterized in that The pre-processed historical related factor data and future related factor data respectively include production volume, energy consumption, temperature and time; The step of using the first training data and the second training data to train the BP neural network model and the LSTM neural network model and correspondingly obtaining the target BP neural network model and the target LSTM neural network model comprises: Using the first training data and the second training data, training a BP neural network model to adjust network weights in the BP neural network model corresponding to the production volume, energy consumption, and temperature, respectively, to obtain a target BP neural network model; The first training data and the second training data are used to train the LSTM neural network model so that the LSTM neural network model learns the temporal law of carbon emissions according to the time to obtain a target LSTM neural network model.

4. The method according to claim 1, characterized in that The multi-objective function is the sum of the objective function of carbon emission reduction multiplied by a first coefficient, the objective function of economic cost multiplied by a second coefficient, and the objective function of energy consumption multiplied by a third coefficient.

5. The method according to claim 4, characterized in that The objective function for reducing carbon emissions is the difference between the predicted carbon emissions and the optimized carbon emissions, divided by the predicted carbon emissions; The objective function of the economic cost is the economic cost data determined according to the optimized carbon emissions divided by the maximum economic cost; The objective function of the energy consumption is the energy consumption determined according to the optimized carbon emissions divided by the maximum energy consumption.

6. The method according to claim 1, characterized in that Also includes: The target LSTM neural network model compares the difference between the target carbon emissions predicted by the target LSTM neural network model and the current carbon emissions with a preset difference; If the difference exceeds the preset difference, it is determined that the current carbon emission is abnormal.

7. The method according to claim 1, characterized in that The step of preprocessing the historical carbon emissions and the historical related factor data comprises: Invalid data and abnormal data in the historical carbon emissions and the historical related factor data are removed, missing data are filled, and data standardization is performed.

8. An industrial carbon emission control device, characterized in that: include: An acquisition unit, used to acquire target-related factor data affecting carbon emissions; An input unit, used to input the target-related factor data into a target neural network model to output predicted carbon emissions; wherein the target neural network model is obtained by combining a target BP neural network model and a target LSTM neural network model; A definition unit, used to define a multi-objective optimization function, wherein the multi-objective optimization function is determined according to an objective function of carbon emission reduction, an objective function of economic cost, and an objective function of energy consumption; A determination unit is used to determine the control parameters of industrial carbon emissions based on the predicted carbon emissions and the multi-objective optimization function using a dynamic parameter multi-objective differential evolution algorithm.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the industrial carbon emission control method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the industrial carbon emission control method according to any one of claims 1 to 7 are implemented.

Images