A coal mine carbon emission assessment system

Abstract

The present application relates to the technical field of carbon emission monitoring, and provides a coal mine carbon emission evaluation system, which comprises a data acquisition module, a data preprocessing module, a carbon emission evaluation module, an optimized hyperparameter module and a decision support module; the system analyzes the dynamic characteristics of carbon emission through a multi-resolution mechanism, captures the carbon emission rules at different time scales, and combines an attention mechanism to improve the weight of key features, thereby generating accurate carbon emission evaluation data; the global exploration capability of the whale optimization algorithm is optimized by referencing a dynamic dimension weight mechanism, and the local development capability is optimized by introducing an acceleration factor and a dynamic step adjustment factor, so that the optimal configuration of the hyperparameters of the carbon emission evaluation system is realized, and the evaluation accuracy and stability are improved; the present application can comprehensively capture the dynamic characteristics of carbon emission, improve the accuracy and timeliness of the evaluation results, and provide important support for coal mine environmental protection and carbon emission reduction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon emission monitoring, and particularly relates to a coal mine carbon emission evaluation system. BACKGROUND

[0002] With the development of science and technology, the monitoring and evaluation of carbon emissions in the process of coal production have gradually become an important research direction in the field of environmental protection; in order to realize the sustainable development of the coal industry, more and more coal mines begin to use carbon emission evaluation systems to quantify carbon emission levels, optimize emission reduction strategies and evaluate environmental impact; however, in the prior art, the traditional coal mine carbon emission evaluation system has the following deficiencies: first, the traditional coal mine carbon emission evaluation system has weak dynamic characteristic capturing ability for carbon emission time series data, and cannot accurately reflect the short-term, medium-term and long-term influence of production activities, energy consumption and environmental factors on carbon emissions; especially for data with high time resolution, the existing system has low sensitivity, which leads to the evaluation results being unable to comprehensively reflect the fluctuation rule and key influencing factors of carbon emissions; in addition, the traditional system lacks accurate identification and weight distribution of key time points when analyzing the influencing factors of carbon emissions, and cannot highlight the features that have important influence on the evaluation results of carbon emissions, thereby affecting the accuracy and practicability of the evaluation; secondly, the existing coal mine carbon emission evaluation system is inefficient in model optimization and hyperparameter configuration in complex environments, and often cannot adapt to the processing requirements of multi-dimensional and high-frequency carbon emission data in actual coal mine scenes, which leads to the accuracy and stability of the evaluation results being difficult to meet the high requirements of carbon emission reduction and environmental management; therefore, there is an urgent need for a coal mine carbon emission evaluation system that can comprehensively capture the time series dynamic characteristics, accurately evaluate the carbon emission level of coal mines, and provide high-precision evaluation results to meet the needs of coal mine carbon emission management and emission reduction strategy formulation. SUMMARY

[0003] The application provides a coal mine carbon emission evaluation system, aiming to solve the problems of insufficient data processing capacity, inaccurate dynamic characteristics of carbon emission and low accuracy of evaluation results in the prior art; first, in the carbon emission evaluation process, the system adopts a multi-resolution processing mechanism to classify and analyze carbon emission data according to time resolution; by capturing the short-term, medium-term and long-term carbon emission change rules, the system comprehensively extracts the influence of different time scales on carbon emission in coal mine production activities, thereby generating high-quality carbon emission evaluation data covering the whole global; at the same time, the system uses the multi-resolution mechanism to highlight the influence of key time points on carbon emission evaluation, thereby further improving the accuracy and timeliness of the evaluation; secondly, the coal mine carbon emission evaluation system in the application optimizes the evaluation process globally through a dynamic dimension weight mechanism, significantly improving the comprehensiveness and global exploration ability of data processing; in the aspect of local optimization, the system introduces an acceleration factor and a dynamic step length adjustment factor, enhancing the adaptability to the complex environmental data characteristics of coal mines and significantly improving the processing efficiency of high-dimensional complex data; through this optimization means, the system can generate carbon emission evaluation data with higher precision, providing more scientific and reliable data support for coal mine carbon emission management and emission reduction strategies.

[0004] The data acquisition module acquires relevant carbon emission data in coal mine production activities, including CO2 concentration, CH4 concentration, energy consumption data and coal production data collected by sensors, and integrated original carbon emission data;

[0005] The data preprocessing module processes the outliers in the original carbon emission data using the median absolute deviation method, fills in the missing data of the original carbon emission data by linear interpolation, and performs normalization processing to generate preprocessed carbon emission data;

[0006] The carbon emission evaluation module establishes a CNN-BiGRU model, optimizes the bidirectional gate recurrent unit through a multi-resolution mechanism, constructs an enhanced bidirectional gate recurrent unit, optimizes the CNN-BiGRU model through the enhanced bidirectional gate recurrent unit, constructs an enhanced CNN-BiGRU model, initializes the hyperparameters of the enhanced CNN-BiGRU model, processes the preprocessed carbon emission data using the enhanced CNN-BiGRU model, and generates carbon emission evaluation data;

[0007] The optimization hyperparameter module optimizes the global exploration of the whale optimization algorithm by referencing the dynamic dimension weight mechanism, improves the Levy flight mechanism by introducing an acceleration factor and a dynamic step adjustment factor, constructs an enhanced Levy flight mechanism, optimizes the local development of the whale optimization algorithm by the enhanced Levy flight mechanism, constructs an improved whale optimization algorithm, optimizes the hyperparameters of the enhanced CNN-BiGRU model by the improved whale optimization algorithm, finds the optimal hyperparameter configuration of the enhanced CNN-BiGRU model, optimizes the carbon emission evaluation data using the optimal hyperparameter configuration, and generates high-precision carbon emission evaluation data.

[0008] The decision support module uses high-precision carbon emission evaluation data for coal mine carbon emission management, carbon emission reduction strategy formulation, and environmental impact assessment.

[0009] Further, the carbon emission evaluation module generates carbon emission evaluation data, which includes the following steps:

[0010] Step P1: Data conversion: Organize the preprocessed carbon emission data according to the time sequence, convert it into multi-dimensional time sequence feature data, and define the carbon emission evaluation task output space.

[0011] The carbon emission evaluation task output space includes: total carbon emission, coal mine carbon emission reduction strategy, and carbon emission dynamic characteristics at hourly, daily, and monthly time resolution.

[0012] Step P2: Extract local features: Use one-dimensional convolution to operate on multi-dimensional time sequence feature data, capture local associations of multi-dimensional time sequence feature data in the time dimension, obtain convolution results, and map the convolution results through a ReLU activation function to generate activation feature data.

[0013] Step P3: Compress data: Pooling processing is performed on the activation feature data, key feature values in the local time period of the multi-dimensional time sequence feature data are extracted through the maximum pooling technology, the dimension of the activation feature data is reduced and the calculation amount is compressed, and the pooling local time feature data is generated. Key feature values include mean, variance, and peak value.

[0014] Step P4: Capture global dependencies: Input the pooling local time feature data into the enhanced bidirectional gated recurrent unit to generate BiGRU global time sequence feature data.

[0015] Step P5: Assign attention weights: Process the BiGRU global time sequence feature data through the attention mechanism to generate weighted comprehensive time sequence feature data.

[0016] Step P6: Map feature output: Process the weighted comprehensive time sequence feature data through the full connection layer, map it to the carbon emission evaluation task output space, and generate carbon emission evaluation data.

[0017] Further, step P4 specifically includes the following steps:

[0018] Step P41: split the pooled local time feature data according to the hour-level, day-level and month-level time resolution to form multiple groups of feature data; wherein the hour-level resolution is used to capture the device operation data and short-term carbon emission change pattern; the day-level resolution is used to capture the medium-term impact of production activities and energy consumption on carbon emissions; the month-level resolution captures the long-term trend impact of policy changes and device aging;

[0019] Step P42: input the multiple groups of feature data into the subnetwork of the enhanced bidirectional gated recurrent unit to extract hour-level global time sequence features, day-level global time sequence features and month-level global time sequence features; forward GRU and backward GRU are used inside the subnetwork to capture dynamic dependency at different time steps;

[0020] Step P43: fuse the hour-level global time sequence features, day-level global time sequence features and month-level global time sequence features into BiGRU global time sequence feature data through a weighted fusion mechanism.

[0021] Further, the process of optimizing the hyperparameters of the enhanced CNN-BiGRU model by the optimization hyperparameter module specifically includes the following steps:

[0022] Step B1: randomly generate white whale individuals of the population in the whale optimization algorithm, each white whale individual represents a set of hyperparameter configurations of the enhanced CNN-BiGRU model, define the search space, the hyperparameter configurations include the convolution kernel size, the number of convolution kernels, the BiGRU hidden unit number, the Attention mechanism type, the learning rate and the batch size, generate the initial white whale population according to the search space, set the fitness function, calculate the initial fitness value of the initial white whale population through the fitness function, and select the current global optimal solution according to the initial fitness value;

[0023] Step B2: globally explore the white whale individuals of the initial white whale population, simulate the swimming behavior of white whales and combine the dynamic dimension weight mechanism, randomly select a target whale individual, each white whale individual updates its own hyperparameter configuration through the random jump mechanism, combines the target whale individual and the current global optimal solution, and expands the search space to generate a global white whale population, the global white whale population is divided into a high-performance population and a low-performance population;

[0024] Step B3: locally optimize the high-performance population, adjust the step length of the white whale individuals in the high-performance population using the enhanced Levy flight mechanism, and optimize the hyperparameter configurations of the white whale individuals in the high-performance population;

[0025] Step B4: process the low-performance population through the whale falling mechanism, randomly reset the hyperparameter configurations of the white whale individuals in the low-performance population to prevent falling into local optimum.

[0026] Step B5: updating the global white whale population by local development and whale falling processing, generating an optimized global white whale population, calculating the optimized fitness value of the optimized global white whale population by the fitness function, and updating the current global optimal solution according to the optimized fitness value;

[0027] Step B6: setting the maximum number of iterations, iterating step B2-step B5 until the maximum number of iterations is reached, generating a final global optimal solution, and obtaining the optimal hyperparameter configuration.

[0028] The beneficial effects achieved by the above scheme are as follows:

[0029] The coal mine carbon emission evaluation system provided by the application significantly improves the accuracy and efficiency of coal mine carbon emission evaluation through data processing and optimization means. Firstly, the application realizes accurate capture of dynamic change rules of carbon emission at different time scales by analyzing and processing coal mine carbon emission data through a multi-resolution mechanism, solving the problem of incomplete capture of short-term fluctuations, medium-term trends and long-term influence characteristics of carbon emission by traditional coal mine carbon emission evaluation systems. Through analysis of hourly, daily and monthly time resolution, the application can dynamically present the influence of various factors such as coal mine production activities, equipment operating status and policy changes on carbon emission, providing more global and timely data support for coal mine carbon emission evaluation and enhancing the system's adaptability to complex carbon emission patterns.

[0030] Secondly, the application optimizes the global exploration capability by using a dynamic dimension weight mechanism, accurately adjusts local development by combining an acceleration factor and a dynamic step length adjustment factor, and improves the optimization efficiency of hyperparameters in the coal mine carbon emission evaluation process. Through the improved optimization strategy, the system can quickly find the optimal carbon emission data evaluation configuration, solving the problem of low hyperparameter configuration efficiency and difficulty in generating high-precision evaluation results in traditional evaluation systems, thereby further improving the accuracy and stability of coal mine carbon emission data analysis.

[0031] Finally, the decision support module of the application uses high-precision carbon emission evaluation data to provide a scientific basis for dynamic management of carbon emission, carbon emission reduction strategy formulation and environmental impact assessment of coal mines. Through efficient operation of the system, coal mines can realize life cycle management of carbon emission, formulate more targeted and effective carbon emission reduction measures, and improve the competitiveness and sustainable development ability of coal mines in the field of low-carbon environmental protection. In summary, the application makes an important contribution to the realization of fine carbon emission evaluation and environmental protection goals in the coal industry. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 A model schematic diagram of a coal mine carbon emission evaluation system is provided.

[0033] Figure 2 This is a flowchart illustrating the carbon emission assessment module in Example 2. Detailed Implementation

[0034] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0035] Example 1, according to Figure 1 The present invention provides a coal mine carbon emission assessment system, which includes a data acquisition module, a data preprocessing module, a carbon emission assessment module, an optimization hyperparameter module, and a decision support module.

[0036] The data acquisition module collects relevant carbon emission data from coal mine production activities, including CO2 concentration, CH4 concentration, energy consumption data and coal mine output data collected by sensors, and integrates them to obtain raw carbon emission data.

[0037] The data preprocessing module uses the median absolute deviation method to process outliers in the original carbon emission data, fills in the missing data in the original carbon emission data through linear interpolation, and performs normalization to generate preprocessed carbon emission data.

[0038] The carbon emission assessment module establishes a CNN-BiGRU model, optimizes the bidirectional gated recurrent unit through a multi-resolution mechanism, constructs an enhanced bidirectional gated recurrent unit, optimizes the CNN-BiGRU model through the enhanced bidirectional gated recurrent unit, constructs an enhanced CNN-BiGRU model, initializes the hyperparameters of the enhanced CNN-BiGRU model, and uses the enhanced CNN-BiGRU model to process preprocessed carbon emission data to generate carbon emission assessment data.

[0039] The hyperparameter optimization module uses a dynamic dimension weighting mechanism to optimize the global exploration of the whale optimization algorithm, introduces an acceleration factor and a dynamic step size adjustment factor to improve the Levy flight mechanism, constructs an enhanced Levy flight mechanism, optimizes the local development of the whale optimization algorithm through the enhanced Levy flight mechanism, constructs an improved whale optimization algorithm, optimizes the hyperparameters of the enhanced CNN-BiGRU model through the improved whale optimization algorithm, finds the optimal hyperparameter configuration of the enhanced CNN-BiGRU model, uses the optimal hyperparameter configuration to optimize carbon emission assessment data, and generates high-precision carbon emission assessment data.

[0040] The decision support module utilizes high-precision carbon emission evaluation data to perform coal mine carbon emission management, carbon emission reduction strategy formulation, and environmental impact evaluation.

[0041] Embodiment Two, according to Figure 2 This embodiment is based on Embodiment One. In this embodiment, the process of generating carbon emission evaluation data by the carbon emission evaluation module specifically includes the following steps:

[0042] Step P1: Data conversion: organize the preprocessed carbon emission data according to time series, convert it into multi-dimensional time series feature data, and define the carbon emission evaluation task output space;

[0043] The carbon emission evaluation task output space includes: total carbon emission, coal mine carbon emission reduction strategy, and carbon emission dynamic characteristics at hourly, daily, and monthly time resolutions;

[0044] Step P2: Extract local features: use one-dimensional convolution to perform convolution operation on the multi-dimensional time series feature data, capture the local correlation of the multi-dimensional time series feature data in the time dimension, obtain the convolution result, and perform nonlinear mapping on the convolution result through the ReLU activation function to generate activation feature data;

[0045] Step P3: Compress data: perform pooling processing on the activation feature data, extract key feature values in the local time period in the multi-dimensional time series feature data through the maximum pooling technology, reduce the dimension of the activation feature data and compress the calculation amount, generate pooled local time feature data, and the key feature values include mean, variance, and peak value;

[0046] Step P4: Capture global dependencies: input the pooled local time feature data into the enhanced bidirectional gated recurrent unit to generate BiGRU global time series feature data, and the enhanced bidirectional gated recurrent unit includes: hourly global time series feature extraction, daily global time series feature extraction, and monthly global time series feature extraction;

[0047] Step P5: Assign attention weights: process the BiGRU global time series feature data through the attention mechanism to generate weighted comprehensive time series feature data;

[0048] Step P6: Map feature output: process the weighted comprehensive time series feature data through the full connection layer to map to the carbon emission evaluation task output space, and generate carbon emission evaluation data.

[0049] Embodiment Three, based on Embodiment One, in this embodiment, the process of generating carbon emission evaluation data by the carbon emission evaluation module specifically includes the following steps:

[0050] Step T1: Data conversion: organize the preprocessed carbon emission data according to time series, convert it into multi-dimensional time series feature data, and define the carbon emission evaluation task output space;

[0051] The carbon emission evaluation task output space includes: total carbon emission, coal mine carbon emission reduction strategy and carbon emission dynamic characteristics at hour level, day level and month level time resolution;

[0052] Step T2: extracting local features: using one-dimensional convolution to perform convolution operation on multi-dimensional time series feature data, capturing the local correlation of multi-dimensional time series feature data in time dimension, obtaining convolution result, and generating activated feature data through ReLU activation function for non-linear mapping of the convolution result;

[0053] Step T3: compressing data: performing pooling processing on the activated feature data, extracting key feature values in the local time period in the multi-dimensional time series feature data through the maximum pooling technology, reducing the dimension of the activated feature data and compressing the calculation amount, generating pooled local time feature data, and the key feature values include mean, variance and peak value;

[0054] Step T4: capturing global dependence: inputting the pooled local time feature data into the bidirectional gated recurrent unit to generate BiGRU global time series feature data;

[0055] Step T5: assigning attention weight: processing the BiGRU global time series feature data through the attention mechanism to generate weighted comprehensive time series feature data;

[0056] Step T6: mapping feature output: processing the weighted comprehensive time series feature data through the full connection layer to map to the carbon emission evaluation task output space to generate carbon emission evaluation data.

[0057] In this embodiment, step P4 specifically includes the following steps:

[0058] Step P41: splitting the pooled local time feature data according to the hour level, day level and month level time resolution to form multiple groups of feature data; wherein, the hour level resolution is used to capture the device operation data and short-term carbon emission change mode; the day level resolution is used to capture the medium-term influence of production activities and energy consumption on carbon emission; and the month level resolution captures the long-term trend influence of policy changes and equipment aging;

[0059] Step P42: inputting the multiple groups of feature data into the subnetwork of the enhanced bidirectional gated recurrent unit to extract hour level global time series feature, day level global time series feature and month level global time series feature; the subnetwork internally uses forward GRU and backward GRU to capture the dynamic dependence relationship at different time steps;

[0060] Subnetwork feature extraction formula:

[0061] ;

[0062] wherein, represents a hour-level global timing feature, represents a bidirectional gated recurrent unit, represents a hour-level time resolution feature data in the plurality of sets of feature data;

[0063] ;

[0064] wherein, represents a day-level global timing feature, represents a day-level time resolution feature data in the plurality of sets of feature data;

[0065] ;

[0066] wherein, represents a month-level global timing feature, represents a month-level time resolution feature data in the plurality of sets of feature data;

[0067] Step P43: fusing the hour-level global timing feature, the day-level global timing feature and the month-level global timing feature into BiGRU global timing feature data through a weighted fusion mechanism.

[0068] Embodiment five, based on embodiment four, in this embodiment, the process of optimizing the hyperparameters of the enhanced CNN-BiGRU model by the optimization hyperparameter module, specifically includes the following steps:

[0069] Step B1: randomly generating a white whale individual of a population in the whale optimization algorithm, each white whale individual representing a set of hyperparameter configurations of the enhanced CNN-BiGRU model, defining a search space, the hyperparameter configurations including a convolution kernel size, a convolution kernel number, a BiGRU hidden unit number, an Attention mechanism type, a learning rate and a batch size, generating an initial white whale population according to the search space, setting a fitness function, calculating initial fitness values of the initial white whale population through the fitness function, and selecting a current global optimal solution according to the initial fitness values;

[0070] Step B2: globally exploring the white whale individuals of the initial white whale population, simulating a white whale swimming behavior and combining a dynamic dimension weight mechanism, randomly selecting a target whale individual, updating the hyperparameter configuration of each white whale individual through a random jumping mechanism, combining the target whale individual and the current global optimal solution, expanding the search space, and generating a global white whale population, the global white whale population being divided into a high-performance population and a low-performance population, and the used formula being as follows:

[0071] Dynamic updating formula of the position of the white whale individual:

[0072] Even dimension:

[0073] ;

[0074] in, Indicates the number of iterations. Indicates the current dimension index. Indicates dynamic dimension weights. This represents the current index of the beluga whale individual. Indicates the index of the target beluga whale individual. Indicates the first The first beluga whale individual in the A new position in each dimension, Indicates the first The first beluga whale individual in the Current position in each dimension and Represents a random number. Represents random numbers The sine value, This indicates that the target beluga whale individual was in the [number]th [year]. Position in each dimension Index representing a random individual beluga whale. This indicates that a random beluga whale individual is in the 1st... Position in each dimension;

[0075] Odd-numbered dimensions:

[0076] ;

[0077] in, Represents random numbers The cosine value;

[0078] Step B3: Perform local optimization on the high-performance population. Utilize the enhanced Levy flight mechanism to adjust the step size of individual beluga whales within the high-performance population, thereby optimizing the hyperparameter configuration of these individuals. The formula used is as follows:

[0079] ;

[0080] in, Indicates the acceleration factor. Represents the acceleration proportionality constant. Indicates the current individual beluga whale fitness value, Indicates in Beluga whale individuals during iteration fitness value, Indicates the scaling factor. This represents the minimum fitness value in the current population;

[0081] ;

[0082] in, Indicates the first A single beluga whale The new position after the next iteration Indicates the first A single beluga whale Position in the next iteration Represents a random individual beluga whale. In the Position in the next iteration Represents a random number. This represents the dynamic step size adjustment factor. This represents the Levy flight step size function. This indicates the beluga whale individual with the best fitness value in the current population at the [number]th [year]. Position in the next iteration;

[0083] Step B4: Use the whale fall mechanism to handle low-performance populations by randomly resetting the hyperparameter configurations of individual beluga whales in the low-performance population to prevent them from getting trapped in local optima.

[0084] Step B5: Update the global beluga whale population through local development and whale fall treatment, generate an optimized global beluga whale population, calculate the optimized fitness value of the optimized global beluga whale population through the fitness function, and update the current global optimal solution based on the optimized fitness value;

[0085] Step B6: Set the maximum number of iterations, iterate through steps B2 to B5 until the maximum number of iterations is reached, generate the final global optimal solution, and obtain the optimal hyperparameter configuration.

[0086] Example 6, based on Example 4, focuses on optimizing the hyperparameters of the enhanced CNN-BiGRU model. The specific steps include:

[0087] Step R1: Randomly generate beluga whale individuals in the population of the whale optimization algorithm. Each beluga whale individual represents a set of hyperparameter configurations of the enhanced CNN-BiGRU model. Define the search space. The hyperparameter configurations include kernel size, number of kernels, number of BiGRU hidden units, type of attention mechanism, learning rate, and batch size. Generate an initial beluga whale population based on the search space. Set the fitness function. Calculate the initial fitness value of the initial beluga whale population using the fitness function. Select the current global optimum based on the initial fitness value.

[0088] Step R2: Conduct a global exploration of the individual beluga whales in the initial beluga whale population, simulate beluga whale swimming behavior, randomly select target whale individuals, and each beluga whale individual updates its own hyperparameter configuration by combining the target whale individual and the current global optimal solution through a random jump mechanism, and expands the search space to generate a global beluga whale population. The global beluga whale population is divided into a high-performance population and a low-performance population.

[0089] Step R3: Perform local optimization on the high-performance population by using the Levy flight mechanism to adjust the step size of individual beluga whales in the high-performance population and optimize the hyperparameter configuration of individual beluga whales in the high-performance population.

[0090] Step R4: Use the whale fall mechanism to handle low-performance populations by randomly resetting the hyperparameter configurations of individual beluga whales in the low-performance population to prevent them from getting trapped in local optima.

[0091] Step R5: Update the global beluga whale population through local development and whale fall treatment, generate an optimized global beluga whale population, calculate the optimized fitness value of the optimized global beluga whale population through the fitness function, and update the current global optimal solution based on the optimized fitness value;

[0092] Step R6: Set the maximum number of iterations, iterate through steps R2 to R5 until the maximum number of iterations is reached, generate the final global optimal solution, and obtain the optimal hyperparameter configuration.

[0093] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A coal mine carbon emission assessment system, the system comprising a data preprocessing module, the data preprocessing module generating preprocessed carbon emission data; characterized in that: The system also includes a carbon emission assessment module and an optimization hyperparameter module; The carbon emission assessment module constructs an enhanced CNN-BiGRU model, initializes the hyperparameters of the enhanced CNN-BiGRU model, uses the enhanced CNN-BiGRU model to process preprocessed carbon emission data, and generates carbon emission assessment data. The hyperparameter optimization module constructs an improved whale optimization algorithm, which optimizes the hyperparameters of the enhanced CNN-BiGRU model. The hyperparameter optimization module optimizes the hyperparameters of the enhanced CNN-BiGRU model, specifically including the following steps: Step B1: Randomly generate beluga whale individuals in the population of the whale optimization algorithm. Each beluga whale individual represents a set of hyperparameter configurations of the enhanced CNN-BiGRU model. Define the search space, generate the initial beluga whale population, and select the current global optimum. Step B2: Conduct a global exploration of the initial beluga whale population, simulating beluga whale swimming behavior and incorporating a dynamic dimension weighting mechanism. Randomly select target whale individuals. Each beluga whale updates its hyperparameter configuration and expands the search space through a random jump mechanism, combining the target whale individual and the current global optimum. This generates a global beluga whale population, which is divided into a high-performance population and a low-performance population, using the following formula: The dynamic update formula for the location of individual beluga whales: Even-numbered dimensions: ； in, Indicates the number of iterations. Indicates the current dimension index. Indicates dynamic dimension weights. This represents the current index of the beluga whale individual. Indicates the index of the target beluga whale individual. Indicates the first The first beluga whale individual in the A new position in each dimension, Indicates the first The first beluga whale individual in the Current position in each dimension and Represents a random number. Represents random numbers The sine value, This indicates that the target beluga whale individual was in the [number]th [year]. Position in each dimension Index representing a random individual beluga whale. This indicates that a random beluga whale individual is in the 1st... Position in each dimension; Odd-numbered dimensions: ； in, Represents random numbers The cosine value; Step B3: Perform local optimization on the high-performance population. Utilize the enhanced Levy flight mechanism to adjust the step size of individual beluga whales within the high-performance population, thereby optimizing the hyperparameter configuration of these individuals. The formula used is as follows: ； in, Indicates the acceleration factor. Represents the acceleration proportionality constant. Indicates the current individual beluga whale fitness value, Indicates in Beluga whale individuals during iteration fitness value, Indicates the scaling factor. This represents the minimum fitness value in the current population; ； in, Indicates the first A single beluga whale The new position after the next iteration Indicates the first A single beluga whale Position in the next iteration Represents a random individual beluga whale. In the Position in the next iteration Represents a random number. This represents the dynamic step size adjustment factor. This represents the Levy flight step size function. This indicates the beluga whale individual with the best fitness value in the current population at the [number]th [year]. Position in the next iteration; Step B4: Use the whale fall mechanism to process low-performance populations by randomly resetting the hyperparameter configurations of individual beluga whales in the low-performance population; Step B5: Update the global beluga whale population through local development and whale fall processing, generate an optimized global beluga whale population, and update the current global optimal solution; Step B6: Iterate through steps B2-B5 to optimize hyperparameter configuration.

2. The coal mine carbon emission assessment system according to claim 1, characterized in that: A CNN-BiGRU model is established, and the bidirectional gated recurrent unit is optimized through a multi-resolution mechanism to construct an enhanced CNN-BiGRU model.

3. The coal mine carbon emission assessment system according to claim 1, characterized in that: The carbon emission assessment module generates carbon emission assessment data, specifically including the following steps: Step P1: Convert the preprocessed carbon emission data into multidimensional time-series feature data; Step P2: Perform convolution operations on the multidimensional temporal feature data to generate activation feature data; Step P3: Perform pooling on the activation feature data to generate pooled local temporal feature data; Step P4: Input the pooled local temporal feature data into the enhanced bidirectional gated recurrent unit to generate BiGRU global temporal feature data; Step P5: Process the BiGRU global temporal feature data through an attention mechanism to generate weighted composite temporal feature data; Step P6: Process the weighted integrated time series feature data through a fully connected layer to generate carbon emission assessment data.

4. The coal mine carbon emission assessment system according to claim 3, characterized in that: Step P4 specifically includes the following steps: Step P41: Split the pooled local time feature data according to hourly, daily and monthly time resolutions to form multiple sets of feature data; Step P42: Input multiple sets of feature data into the sub-network of the enhanced bidirectional gated recurrent unit to extract hourly global time series features, daily global time series features, and monthly global time series features; Step P43: The hourly, daily, and monthly global time series features are fused into BiGRU global time series feature data through a weighted fusion mechanism.

Images