Gas concentration prediction method, computer readable storage medium, and electronic device

Abstract

The application provides a gas concentration prediction method, a computer readable storage medium and an electronic device. The method comprises: performing feature extraction on first gas concentration data by using a first model to obtain second gas concentration data; and processing the second gas concentration data by using an updated second model to obtain a prediction result corresponding to the second gas concentration data. In the scheme, the first gas concentration data can be obtained more accurately by performing noise reduction processing on the gas concentration data, the original sequence can be enhanced by performing feature extraction on the first gas concentration data by using a CNN model, the Elman model is improved in the scheme, the prediction result of the updated Elman model is more accurate, and then the updated Elman model is used to predict the gas concentration to obtain the prediction result corresponding to the second gas concentration data, thereby improving the accuracy of the result of the gas concentration prediction.

Description

Technical Field

[0001] This application relates to the field of coal mine detection technology, and more specifically, to a method, apparatus, computer-readable storage medium, and electronic device for predicting gas concentration. Background Technology

[0002] In coal mine integrated monitoring systems, the monitoring system plays a crucial role in daily safety management as an effective means of safety monitoring and control. Effective monitoring of disaster-causing factors and accurate judgment, prediction, and early warning of disaster trends are key to coal mine accident prevention and control. Due to the complexity of the forms of disasters involved in coal mine safety and their interrelationships, as well as the complexity of the correlation between disaster-causing factors in gas disasters themselves, research on predicting gas concentration changes is still relatively limited, and the systematic application of gas concentration prediction methods based on the processing of measured data from safety monitoring systems is even less common. Therefore, analyzing the correlation characteristics of daily gas monitoring data in coal mines to extract effective information and achieving more accurate and reliable gas concentration prediction for gas early warning analysis is of great significance for preventing gas disasters.

[0003] In traditional gas monitoring data prediction research, statistical models play a crucial role. However, these models, based on mathematical statistics, are linear and often struggle to accurately fit the changing curves of highly nonlinear gas monitoring data. With the development of artificial intelligence and big data, artificial neural networks and machine learning have gradually replaced traditional data processing methods. However, in practical applications, existing prediction models have been found to have drawbacks such as slow convergence and a tendency to get trapped in local optima, leading to inaccurate gas concentration predictions. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, computer-readable storage medium, and electronic device for predicting gas concentration, in order to solve the problem of inaccurate gas concentration prediction results in the prior art.

[0005] According to one aspect of the present invention, a method for predicting gas concentration is provided, comprising: acquiring gas concentration data in a coal mine and performing noise reduction processing on the gas concentration data to obtain first gas concentration data; extracting features from the first gas concentration data using a first model, wherein the first model is a CNN model; acquiring a second model and updating the second model to obtain an updated second model, wherein the second model is an Elman model; and processing the second gas concentration data using the updated second model to obtain a prediction result corresponding to the second gas concentration data, wherein the prediction result refers to predicting gas concentration data for a future time period using the second gas concentration data.

[0006] Optionally, obtaining a second model and updating the second model to obtain an updated second model includes: obtaining the second model, which includes an input layer, a hidden layer, and an output layer; updating the butterfly optimization algorithm to obtain an updated butterfly optimization algorithm; and using the updated butterfly optimization algorithm to update the weights corresponding to the output results of the input layer of the second model to obtain the updated second model.

[0007] Optionally, the butterfly optimization algorithm includes an initialization phase, an iteration phase, and a termination phase, wherein the initialization phase involves determining the population size, the number of iterations, and the initial position; the iteration phase involves calculating and sorting the fitness and fragrance intensity of all individual butterflies, and storing the optimal fitness and the corresponding butterfly's position information; and the termination phase involves performing a global or local search within the population to update the butterfly's position information.

[0008] Optionally, the butterfly optimization algorithm is updated to obtain an updated butterfly optimization algorithm, including: obtaining the target coefficient, wherein the target coefficient is κ(t)=κ max -η2(κ max -κ min )+η3(κ(t-1)-(κ max +κ min ) / 2), where κ represents the target coefficient, κ(t) represents the value of the κ coefficient in the t-th iteration, κ max κ represents the maximum value of the coefficient. min η1 represents the minimum value of the κ coefficient, η2 represents a random number within a predetermined range, η3 represents a random number within the predetermined range, and κ(t-1) represents the κ coefficient value in the (t-1)th iteration. Using the target coefficient and the DE algorithm, the population in the butterfly optimization algorithm is screened, and populations that do not meet the predetermined conditions are deleted to obtain the updated butterfly optimization algorithm.

[0009] Optionally, the gas concentration data is denoised to obtain first gas concentration data, including: decomposing the gas concentration data using a wavelet denoising algorithm, taking the gas concentration data greater than or equal to a concentration threshold as first type concentration data, and taking the gas concentration data less than the concentration threshold as second type concentration data; deleting the first type concentration data and retaining the second type concentration data; and reconstructing the second type concentration data using a wavelet denoising algorithm to obtain the first gas concentration data.

[0010] Optionally, the first model is used to extract features from the first gas concentration data to obtain the second gas concentration data, including: using the first model to extract features from the first gas concentration data, wherein the first model includes an input layer, a convolutional layer, a pooling layer and an output layer, the input layer is used to receive input data, the convolutional layer is used to perform convolution operations on the input data to obtain initial feature data, the pooling layer is used to perform pooling operations on the initial feature data to obtain the second gas concentration data, and the output layer is used to output the second gas concentration data, wherein the input data is the gas concentration data.

[0011] Optionally, the updated second model is used to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data, including: using the updated second model to predict the historical second gas concentration data to obtain the prediction result, wherein the updated second model is trained using multiple sets of training data, and each set of training data includes historical second gas concentration data obtained within a historical time period and historical detection results corresponding to the historical second gas concentration data.

[0012] According to another aspect of the present invention, a gas concentration prediction device is also provided, comprising: a first acquisition unit, configured to acquire gas concentration data in a coal mine and perform noise reduction processing on the gas concentration data to obtain first gas concentration data, wherein the gas concentration data is obtained by detection using a target sensor; an extraction unit, configured to perform feature extraction on the first gas concentration data using a first model to obtain second gas concentration data, wherein the first model is a CNN model; a second acquisition unit, configured to acquire a second model and update the second model to obtain an updated second model, wherein the second model is an Elman model; and a prediction unit, configured to process the second gas concentration data using the updated second model to obtain a prediction result corresponding to the second gas concentration data, wherein the prediction result refers to predicting gas concentration data in a future time period using the second gas concentration data.

[0013] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein the program executes any one of the methods described.

[0014] According to another aspect of the present invention, an electronic device is also provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for performing any one of the methods described.

[0015] In this embodiment of the invention, firstly, methane concentration data from underground coal mines is acquired and denoised to obtain first methane concentration data. Then, a first model is used to extract features from the first methane concentration data to obtain second methane concentration data. Next, a second model is acquired and updated to obtain an updated second model. Finally, the updated second model is used to process the second methane concentration data to obtain the prediction result corresponding to the second methane concentration data. In this scheme, by denoising the methane concentration data, more accurate first methane concentration data can be obtained. Using a CNN model to extract features from the first methane concentration data enhances the original sequence. This scheme also improves the Elman model; the updated Elman model provides more accurate prediction results. Therefore, the updated Elman model is used to predict the methane concentration to obtain the prediction result corresponding to the second methane concentration data, thereby improving the accuracy of the methane concentration prediction result. Attached Figure Description

[0016] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0017] Figure 1 A schematic flowchart of a method for predicting gas concentration according to an embodiment of this application is shown;

[0018] Figure 2 A schematic diagram is shown showing the decomposition of gas concentration data using a wavelet denoising algorithm;

[0019] Figure 3 A schematic diagram of the structure of the first model is shown;

[0020] Figure 4 A schematic diagram of a gas concentration prediction device according to an embodiment of this application is shown;

[0021] Figure 5 A schematic diagram of another method for predicting gas concentration is shown. Detailed Implementation

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] It should be understood that when an element (such as a layer, film, region, or substrate) is described as being "on" another element, the element may be directly on the other element, or there may be an intermediate element present. Furthermore, in the specification and claims, when an element is described as being "connected" to another element, the element may be "directly connected" to the other element, or "connected" to the other element via a third element.

[0026] For ease of description, the following explains some of the nouns or terms used in the embodiments of this application:

[0027] Convolutional Neural Networks (CNNs) are among the most promising deep learning (DL) frameworks. They possess strong representation learning capabilities and can extract translation-invariant features from input data according to their network layer structure. The neural nodes in each layer of a CNN are arranged in 3D, so each layer has width, height, and depth. CNNs are not fully connected and share weights, meaning that the neural nodes in a convolutional layer are only connected to a small number of nodes in the previous layer and neurons in the same feature map share weights. This improves upon the redundancy and large number of parameters of fully connected networks.

[0028] In recent years, numerous studies have focused on applying time series analysis methods to mine gas prediction, using mine gas monitoring data as the research object. These studies primarily focus on assessing the risk and trend of coal and gas outbursts, predicting gas emission rates, and forecasting gas concentrations. Research applied to assessing the risk and trend of coal and gas outbursts includes fractal index-based trend assessment, pattern recognition-based outburst risk analysis, support vector machine-based outburst risk assessment, and probabilistic prediction-based outburst risk evaluation. Methods applied to gas concentration prediction mainly include those based on fractal theory, fuzzy data mining, grey prediction, neural networks, support vector machines, chaotic time series prediction, and probability-based prediction.

[0029] In traditional gas monitoring data prediction research, statistical models play a crucial role. However, these models, based on methods derived from mathematical statistics, are linear models. When dealing with highly nonlinear gas monitoring data series, they often struggle to accurately fit the changing curves. With the development of artificial intelligence and big data, artificial neural networks and machine learning have gradually replaced traditional data processing methods. Extensive research by scholars both domestically and internationally has demonstrated that ensemble prediction models outperform single prediction models, avoiding drawbacks such as slow convergence, getting trapped in local optima, and overfitting. Some traditional artificial neural networks suffer from slow convergence and a tendency to get trapped in local optima.

[0030] As mentioned in the background section, the results of gas concentration prediction in the prior art are inaccurate. In order to solve the above problems, in a typical embodiment of this application, a method, apparatus, computer-readable storage medium and electronic device for predicting gas concentration are provided.

[0031] According to an embodiment of this application, a method for predicting gas concentration is provided.

[0032] Figure 1 This is a flowchart of a gas concentration prediction method according to an embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0033] Step S101: Obtain gas concentration data in the coal mine and perform noise reduction processing on the gas concentration data to obtain the first gas concentration data;

[0034] Specifically, the gas concentration data can be obtained using a target sensor, which can be a gas concentration sensor.

[0035] In step S101 above, noise reduction processing can be performed on the gas concentration data, which can improve the signal-to-noise ratio of the gas concentration data and obtain the first gas concentration data. The first gas concentration data is more accurate, and the gas concentration can be predicted more accurately in the future.

[0036] Step S102: Use the first model to extract features from the first gas concentration data to obtain the second gas concentration data. The first model is a CNN model.

[0037] In step S102 above, the first model can be used to extract features from the first gas concentration data. This can enhance the original sequence features in the first gas concentration data and filter out unnecessary data to ensure the accuracy of subsequent prediction results.

[0038] Step S103: Obtain the second model and update the second model to obtain the updated second model, which is the Elman model;

[0039] In step S103 above, by updating the second model, more accurate prediction data can be obtained, which can make more accurate predictions and early warnings of gas disasters.

[0040] Step S104: The updated second model is used to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data. The prediction result refers to the prediction of gas concentration data in the future time period using the second gas concentration data.

[0041] In step S104 above, since the second model was optimized in advance, that is, the second model was updated, the optimization rate and convergence speed of the updated second model are faster. This can further ensure the accuracy of subsequent prediction results and avoid local optima and premature convergence.

[0042] The above method first acquires underground coal mine gas concentration data and performs noise reduction on the data to obtain first gas concentration data. Then, a first model is used to extract features from the first gas concentration data to obtain second gas concentration data. Next, a second model is acquired and updated to obtain an updated second model. Finally, the updated second model is used to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data. This scheme obtains more accurate first gas concentration data by performing noise reduction on the gas concentration data. Using a CNN model to extract features from the first gas concentration data enhances the original sequence. This scheme also improves the Elman model; the updated Elman model provides more accurate prediction results. Therefore, the updated Elman model is used to predict the gas concentration to obtain the prediction result corresponding to the second gas concentration data, thereby improving the accuracy of the gas concentration prediction results.

[0043] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0044] Elman neural network models possess strong learning capabilities, and shallow networks have a significant advantage in handling small-sample nonlinear problems. However, Elman neural network models are inherently prone to getting trapped in local optima and premature convergence. Furthermore, during the training process of Elman neural network models, the network layer weight coefficients and bias parameters are difficult to find optimal solutions during iterations, directly affecting the final prediction accuracy. Therefore, in order to optimize Elman neural network models, enabling them to escape local optima and reach the global optimum to accelerate convergence, one embodiment of this application obtains a second model and updates it to obtain an updated second model, specifically including the following steps:

[0045] Step S201: Obtain the second model described above, which includes an input layer, a hidden layer, and an output layer.

[0046] Step S202: Update the butterfly optimization algorithm to obtain the updated butterfly optimization algorithm;

[0047] In one specific embodiment of this application, the butterfly optimization algorithm includes an initialization phase, an iteration phase, and a termination phase. The initialization phase determines the population size, number of iterations, and initial position. The iteration phase calculates and sorts the fitness and fragrance intensity of all individual butterflies, storing the optimal fitness and the corresponding butterfly's position information. The termination phase performs a global or local search within the population to update the butterfly's position information. In this embodiment, the pre-obtained butterfly optimization algorithm allows for the definition of the weight coefficients and bias parameters of the Elman neural network model. This enables subsequent updates to the butterfly optimization algorithm to further optimize the weight coefficients and bias parameters of the Elman neural network model, resulting in a more efficient computational Elman neural network model.

[0048] Specifically, the butterfly optimization algorithm includes an initialization phase, an iteration phase, and a termination phase. The concepts of scent, location, and stimulus intensity in the butterfly algorithm constitute its specific optimization process. Through continuous iteration, it selects individuals with high fitness from the population, constantly seeking the optimal position, which is the optimal individual (i.e., the optimal value of the weights and thresholds). The three phases are described below:

[0049] Initialization phase: Define the objective function (fragrance is represented as a function of the physical intensity of the stimulus) and the space of solutions corresponding to the objective function, i.e., determine the population size, number of iterations, and initial position, specifically using the formula: f = CI a Where f represents the intensity of the fragrance, C represents the sensory modality, I represents the intensity of the stimulus, and a is the power exponent, taking values ​​between [0,1].

[0050] Iteration Phase: During the iteration process, all butterflies move to new locations, and their objective function is recalculated. This involves calculating and sorting the fitness and scent intensity of all individual butterflies, and storing the optimal fitness along with the corresponding butterfly's location information. The specific formula is: x i t+1 =x i t +(r 2 ×g * -x i t )×f i , where x i t+1 Let x represent the (t+1)th butterfly in the solution space. i t Let r represent the t-th butterfly in the solution space, where r is a random number and g is a random number. * f represents the optimal solution found among all solutions in the current iteration. i This represents the scent of the i-th butterfly.

[0051] Final stage: The final position update formula is:

[0052] This scheme uses the butterfly optimization algorithm to optimize the parameters of the Elman neural network model. Therefore, to further address the problems of slow convergence and susceptibility to local optima, the butterfly optimization algorithm can be further optimized. In another specific embodiment of this application, the butterfly optimization algorithm is updated to obtain an updated butterfly optimization algorithm, including: obtaining target coefficients, where the target coefficients are:

[0053] κ(t)=κ max -η2(κ max -κ min )+η3(κ(t-1)-(κ max +κ min ) / 2), where κ represents the target coefficient mentioned above, κ(t) represents the value of the κ coefficient in the t-th iteration, κ max κ represents the maximum value of the coefficient. minLet ηk represent the minimum value of the κ coefficient, η2 represent a random number within a predetermined range, η3 represent a random number within the same predetermined range, and κ(t-1) represent the κ coefficient value in the (t-1)th iteration. Using the aforementioned target coefficient and the DE algorithm, the population in the butterfly optimization algorithm is screened, and populations that do not meet the predetermined conditions are deleted, resulting in an updated butterfly optimization algorithm. In this embodiment, a nonlinear weight coefficient (target coefficient) and a crossover and mutation process using the differential evolution algorithm (DE algorithm) are introduced to optimize the butterfly optimization algorithm. This further addresses the problems of slow convergence speed and susceptibility to local optima.

[0054] In one alternative embodiment, the predetermined range can be the range [0,1], and the effective range of the κ coefficient can be [κ...]. min κ max The scope of ].

[0055] Specifically, to avoid premature convergence due to poor population diversity in the later stages of algorithm iteration, a target coefficient can be introduced into the butterfly optimization algorithm. This target coefficient can improve population diversity in the later stages of the butterfly optimization algorithm. The updated formula for the iteration stages of the butterfly optimization algorithm is as follows: The formula for the final stage of the updated butterfly optimization algorithm is: Because the DE algorithm has the characteristics of fast convergence speed and easy escape from local optima, the crossover process of the DE algorithm is introduced into the butterfly optimization algorithm to improve the solution performance of the butterfly optimization algorithm.

[0056] Step S203: The updated butterfly optimization algorithm is used to update the weights corresponding to the output results of the input layer of the second model to obtain the updated second model.

[0057] In steps S201 to S203 above, the improved butterfly optimization algorithm (IBOA) is used to optimize the weight coefficients and bias parameters of the Elman neural network model. This optimizes the Elman neural network model, that is, it updates the Elman neural network model. The updated Elman neural network model can escape local optima and obtain the global optimum, thus accelerating the convergence speed of the model.

[0058] When collecting methane concentration data, this data may contain noise. The sources of noise include factors such as data acquisition, data transmission, and data processing. During data acquisition, harsh underground conditions in coal mines, including toxic and harmful gases and high humidity, can reduce the sensitivity of sensor components. Furthermore, dust, pulverized coal, and water vapor can all contribute to the mixing of noise into the monitoring data. During data transmission, underground communication lines in coal mines may be subject to interference from non-detection equipment, resulting in noise in the transmitted data. Finally, during data processing, malfunctions or errors in monitoring system components can also cause noise to be mixed into the methane concentration data.

[0059] Since the collected gas concentration data may contain noise, and noise significantly affects the actual monitoring accuracy, leading to inaccurate detection, effective noise reduction processing can be performed on the collected gas concentration data to restore its authenticity. This process also restores the characteristics of the dispersion caused by various factors, thereby improving the detection accuracy and operational stability of this solution. In another embodiment of this application, the above-mentioned gas concentration data is subjected to noise reduction processing to obtain first gas concentration data, specifically including the following steps:

[0060] Step S301: The gas concentration data is decomposed using a wavelet denoising algorithm. The gas concentration data that is greater than or equal to the concentration threshold is taken as the first type of concentration data, and the gas concentration data that is less than the concentration threshold is taken as the second type of concentration data.

[0061] Specifically, the gas concentration data is decomposed using a wavelet denoising algorithm, which can decompose the gas concentration data into low-frequency component data (second type of concentration data) and high-frequency component data (first type of concentration data). The low-frequency component data can be further decomposed, and the decomposition process can be repeated to obtain low-frequency component data and high-frequency component data at any scale.

[0062] The process of decomposing gas concentration data using wavelet denoising algorithm is as follows: Figure 2 As shown, the decomposition can be performed multiple times; for example, four times:

[0063] First decomposition: The gas concentration data is decomposed to obtain low-frequency component data cA1 and high-frequency component data cD1;

[0064] Second decomposition: The low-frequency component data cA1 obtained from the first decomposition is decomposed again to obtain low-frequency component data cA2 and high-frequency component data cD2;

[0065] Third decomposition: The low-frequency component data cA2 obtained from the second decomposition is decomposed again to obtain low-frequency component data cA3 and high-frequency component data cD3;

[0066] Fourth decomposition: The low-frequency component data cA3 obtained from the third decomposition is decomposed again to obtain low-frequency component data cA4 and high-frequency component data cD4.

[0067] Step S302: Delete the first type of concentration data and retain the second type of concentration data;

[0068] Step S303: The wavelet denoising algorithm is used to reconstruct the second type of concentration data to obtain the first gas concentration data.

[0069] Multiple decompositions and reconstructions yield the denoised first gas concentration data. Each wavelet decomposition produces different low-frequency and high-frequency components. The low-frequency components are highly similar to the original gas concentration data, representing most of its features and thus possessing significant value. Conversely, the high-frequency components contain interference information. Retaining the low-frequency components while removing the high-frequency components acts as a filter. Each decomposition removes the high-frequency components, retaining only the low-frequency ones, and then repeats this process four times. Finally, the resulting low-frequency components are reconstructed to obtain the denoised first gas concentration data.

[0070] In steps S301 to S303 above, the gas concentration data is decomposed by using a wavelet denoising algorithm to remove noise and interference, thus improving the accuracy of the data. Furthermore, the second type of concentration data obtained after decomposition is reconstructed to restore the features of the gas concentration data. After decomposition and reconstruction, the first type of gas concentration data is more accurate, which improves the accuracy of the data.

[0071] Since the original Elman neural network model is a fully connected network model, meaning that nodes in adjacent network layers are connected to each other through weights, this increases the network training time and the risk of overfitting. To avoid this problem, this solution uses a suitable convolutional neural network model and an updated Elman neural network model to form a completely new prediction model. This reduces the network training time and the risk of overfitting.

[0072] Convolutional Neural Network (CNN) models extract features through convolutional kernels and pooling operations. Therefore, compared to some solutions that use traditional feature extractors, to overcome the limited information caused by feature extractors and multi-layer hidden neural networks, in another embodiment of this application, a first model is used to extract features from the first gas concentration data to obtain second gas concentration data. Specifically, this includes the following steps: using the first model to extract features from the first gas concentration data, wherein the first model includes an input layer, a convolutional layer, a pooling layer, and an output layer. The input layer receives input data, the convolutional layer performs convolution operations on the input data to obtain initial feature data, the pooling layer performs pooling operations on the initial feature data to obtain the second gas concentration data, and the output layer outputs the second gas concentration data. The input data is the gas concentration data. In this solution, using a CNN model to extract features from the denoised first gas concentration data further enhances the original sequence features, ensuring more efficient and accurate subsequent prediction of gas concentration.

[0073] Specifically, a convolutional neural network (CNN) model includes an input layer, convolutional layers, pooling layers, and an output layer. The following is an introduction to convolutional layers and pooling layers:

[0074] Convolutional Layers: Convolutional layers are the core of convolutional neural networks. Their function is to extract features from the input data. A convolutional layer contains multiple convolutional kernels, each sharing the same weight coefficients and bias parameters (bias). These kernels repeatedly operate on the entire receptive field (neuron, similar to the receptive field of a visual cortex cell) of the convolutional layer, performing convolution operations on the input data. The result constitutes the feature map of the input data. Convolutional layers have two important operations: local correlation, which simplifies computation, and sliding window, which is used to calculate feature data within the receptive field.

[0075] Pooling layers: The purpose of pooling layers is to reduce the dimensionality of convolutional features. In convolutional neural networks, pooling layers are often inserted consecutively after convolutional layers to reduce the size of the data representation, thereby reducing computational cost and controlling overfitting. Pooling layers include max pooling and average pooling operations. In practice, max pooling is more commonly used than average pooling. This invention uses the max pooling method, employing the max() operation to spatially adjust the size of the convolutional features. It has translation invariance, and typically uses a 2x2 filter with a stride of 2.

[0076] In one optional embodiment, the specific structure of the first model is as follows: Figure 3As shown, the system includes an input layer, a convolutional layer, a buffer layer, a fully connected layer, a pooling layer, and an output layer. The input data of the feature extractor is the training set (input data) of each gas monitoring data subsequence obtained after data processing. After being processed by the convolutional and pooling layers in the convolutional neural network, the feature data of each subsequence (second gas concentration data) is output. Then, the feature data of each dataset is merged as the input data for training the upper-layer predictor based on the improved Elman neural network model. Finally, the updated Elman neural network model is used to obtain the output result.

[0077] With the second model already updated, the updated second model can be used to predict gas concentration, further ensuring the accuracy of the prediction results. In another embodiment of this application, the updated second model is used to process the aforementioned second gas concentration data to obtain the prediction results corresponding to the aforementioned second gas concentration data. Specifically, this includes the following steps: using the updated second model to predict the historical second gas concentration data to obtain the aforementioned prediction results. The updated second model is trained using multiple sets of training data. Each set of training data includes historical second gas concentration data acquired within a historical time period and the corresponding historical detection results. In this embodiment, because the prediction results of the updated second model are more accurate, the gas concentration prediction results obtained using the updated second model in this embodiment have higher accuracy.

[0078] In one specific embodiment, if the gas concentration corresponding to the predicted result is greater than or equal to the concentration threshold, it is determined that the gas concentration is too high and a gas disaster will occur. An alarm message can be generated to remind staff to leave the danger zone in time. If the gas concentration corresponding to the predicted result is less than the concentration threshold, it is determined that the gas concentration is low and a gas disaster will not occur. In this case, no alarm message can be generated, or the gas data can be sent to the three-dimensional visualization platform.

[0079] Optionally, the concentration threshold can be 5% or 16%, but it is not limited to these two cases. Those skilled in the art can also set an appropriate threshold according to the actual conditions in a coal mine.

[0080] This application uses real-time monitoring data collected by a coal mine safety monitoring system as the research object, and applies efficient information processing and calculation analysis methods to predict and analyze mine gas concentration. It provides decision support for safety early warning and reasonable response to gas disasters in coal mines, which is of great practical significance for improving ventilation safety management and gas disaster prevention in coal mines.

[0081] This application also provides a gas concentration prediction device. It should be noted that the gas concentration prediction device of this application embodiment can be used to execute the gas concentration prediction method provided in this application embodiment. The gas concentration prediction device provided in this application embodiment is described below.

[0082] Figure 4 This is a schematic diagram of a gas concentration prediction device according to an embodiment of this application. Figure 4 As shown, the device includes:

[0083] The first acquisition unit 10 is used to acquire gas concentration data in the coal mine and perform noise reduction processing on the gas concentration data to obtain the first gas concentration data.

[0084] The first acquisition unit described above can perform noise reduction processing on the gas concentration data, which can improve the signal-to-noise ratio of the gas concentration data and obtain the first gas concentration data. The first gas concentration data is more accurate, and the gas concentration can be predicted more accurately in the future.

[0085] Extraction unit 20 is used to extract features from the first gas concentration data using a first model to obtain second gas concentration data. The first model is a CNN model.

[0086] The extraction unit described above can use the first model to extract features from the first gas concentration data. This can enhance the original sequence features in the first gas concentration data and filter out unnecessary data to ensure the accuracy of subsequent prediction results.

[0087] The second acquisition unit 30 is used to acquire the second model and update the second model to obtain the updated second model, wherein the second model is an Elman model.

[0088] The second acquisition unit mentioned above can obtain more accurate prediction data by updating the second model, which can make more accurate predictions and early warnings of gas disasters.

[0089] The prediction unit 40 is used to process the second gas concentration data using the updated second model to obtain the prediction result corresponding to the second gas concentration data. The prediction result refers to the prediction of gas concentration data in the future time period using the second gas concentration data.

[0090] The aforementioned prediction unit optimizes the second model in advance, i.e., updates the second model. The updated second model has a faster optimization rate and convergence speed, which can further ensure the accuracy of subsequent prediction results and avoid local optima and premature convergence.

[0091] In the aforementioned apparatus, the first acquisition unit acquires methane concentration data from underground coal mines and performs noise reduction processing on the methane concentration data to obtain first methane concentration data. The extraction unit uses a first model to extract features from the first methane concentration data to obtain second methane concentration data. The second acquisition unit acquires a second model and updates the second model to obtain an updated second model. The prediction unit uses the updated second model to process the second methane concentration data to obtain the prediction result corresponding to the second methane concentration data. In this scheme, by performing noise reduction processing on the methane concentration data, more accurate first methane concentration data can be obtained. Using a CNN model to extract features from the first methane concentration data can enhance the original sequence. This scheme also improves the Elman model, and the updated Elman model has more accurate prediction results. Therefore, the updated Elman model is used to predict the methane concentration to obtain the prediction result corresponding to the second methane concentration data, thereby improving the accuracy of the methane concentration prediction result.

[0092] Elman neural network models possess strong learning capabilities, and shallow networks have a significant advantage in handling small-sample nonlinear problems. However, Elman neural network models are inherently prone to getting trapped in local optima and premature convergence. Furthermore, during the training process of Elman neural network models, the network layer weight coefficients and bias parameters are difficult to find optimal solutions during iterations, directly affecting the final prediction accuracy. Therefore, in order to optimize Elman neural network models, enabling them to escape local optima and reach the global optimum, thereby accelerating convergence, in one embodiment of this application, the second acquisition unit includes an acquisition module, a first update module, and a second update module. The functions of each module are as follows:

[0093] The acquisition module is used to acquire the second model mentioned above, which includes an input layer, a hidden layer, and an output layer.

[0094] The first update module is used to update the butterfly optimization algorithm to obtain the updated butterfly optimization algorithm;

[0095] In one specific embodiment of this application, the butterfly optimization algorithm includes an initialization phase, an iteration phase, and a termination phase. The initialization phase determines the population size, number of iterations, and initial position. The iteration phase calculates and sorts the fitness and fragrance intensity of all individual butterflies, storing the optimal fitness and the corresponding butterfly's position information. The termination phase performs a global or local search within the population to update the butterfly's position information. In this embodiment, the pre-obtained butterfly optimization algorithm allows for the definition of the weight coefficients and bias parameters of the Elman neural network model. This enables subsequent updates to the butterfly optimization algorithm to further optimize the weight coefficients and bias parameters of the Elman neural network model, resulting in a more efficient computational Elman neural network model.

[0096] This scheme uses the butterfly optimization algorithm to optimize the parameters of the Elman neural network model. Therefore, to further address the problems of slow convergence speed and easy getting trapped in local optima, the butterfly optimization algorithm can be further optimized. In another specific embodiment of this application, the first update module includes an acquisition submodule and an update submodule. The acquisition submodule is used to acquire the target coefficients, which are:

[0097] κ(t)=κ max -η2(κ max -κ min )+η3(κ(t-1)-(κ max +κ min ) / 2), where κ represents the target coefficient mentioned above, κ(t) represents the value of the κ coefficient in the t-th iteration, κ max κ represents the maximum value of the coefficient. min ηk represents the minimum value of the κ coefficient, η2 represents a random number within a predetermined range, η3 represents a random number within the same predetermined range, and κ(t-1) represents the κ coefficient value in the (t-1)th iteration. The update submodule is used to filter the population in the butterfly optimization algorithm using the target coefficient and the differential evolution algorithm (DE algorithm), deleting populations that do not meet the predetermined conditions, and obtaining the updated butterfly optimization algorithm. In this embodiment, a nonlinear weight coefficient (target coefficient) and a crossover and mutation process of the differential evolution algorithm (DE algorithm) are introduced to optimize the butterfly optimization algorithm, which can further solve the problems of slow convergence speed and easy getting trapped in local optima.

[0098] The second update module is used to update the weights corresponding to the output results of the input layer of the second model using the updated butterfly optimization algorithm, so as to obtain the updated second model.

[0099] The aforementioned acquisition module, first update module, and second update module are optimized using the Improved Butterfly Optimization Algorithm (IBOA). This optimization can improve the weight coefficients and bias parameters of the Elman neural network model, thereby updating the Elman neural network model. The updated Elman neural network model can then escape local optima and reach the global optimum, accelerating the model's convergence speed.

[0100] Since the collected gas concentration data may contain noise, and noise significantly affects the actual monitoring accuracy, leading to inaccurate detection, effective noise reduction processing can be performed on the collected gas concentration data to restore its authenticity. This process also restores the characteristics of the dispersion caused by various factors, thereby improving the detection accuracy and operational stability of this solution. In another embodiment of this application, the first acquisition unit includes a first processing module, a second processing module, and a third processing module. The functions of each module are as follows:

[0101] The first processing module is used to decompose the gas concentration data using a wavelet denoising algorithm, and to take the gas concentration data that is greater than or equal to the concentration threshold as the first type of concentration data, and the gas concentration data that is less than the concentration threshold as the second type of concentration data.

[0102] The second processing module is used to delete the first type of concentration data and retain the second type of concentration data.

[0103] The third processing module is used to reconstruct the second type of concentration data using a wavelet denoising algorithm to obtain the first type of gas concentration data.

[0104] The first, second, and third processing modules described above decompose the gas concentration data using a wavelet denoising algorithm to remove interference and improve data accuracy. Furthermore, they reconstruct the second type of concentration data obtained after decomposition to restore the features of the gas concentration data. After decomposition and reconstruction, the obtained first type of gas concentration data is more accurate, thus improving data precision.

[0105] Convolutional Neural Network (CNN) models extract features through convolutional kernels and pooling operations. Therefore, compared to some solutions that use traditional feature extractors, to overcome the limited information caused by feature extractors and multi-layer hidden neural networks, in another embodiment of this application, the extraction unit includes an extraction module. The extraction module is used to extract features from the first gas concentration data using the aforementioned first model. The first model includes an input layer, a convolutional layer, a pooling layer, and an output layer. The input layer receives input data, the convolutional layer performs convolution operations on the input data to obtain initial feature data, the pooling layer performs pooling operations on the initial feature data to obtain the second gas concentration data, and the output layer outputs the second gas concentration data. The input data is the gas concentration data. In this solution, using a CNN model to extract features from the denoised first gas concentration data further enhances the original sequence features, ensuring more efficient and accurate subsequent prediction of gas concentration.

[0106] With the second model already updated, the updated second model can be used to predict gas concentration, further ensuring more accurate prediction results. In another embodiment of this application, the prediction unit includes a prediction module, which is used to predict the historical second gas concentration data using the updated second model to obtain the prediction result. The updated second model is trained using multiple sets of training data, each set including historical second gas concentration data acquired within a historical time period and corresponding historical detection results. In this embodiment, because the prediction results of the updated second model are more accurate, the gas concentration prediction result obtained using the updated second model in this embodiment has higher accuracy.

[0107] The aforementioned gas concentration prediction device includes a processor and a memory. The first acquisition unit, extraction unit, second acquisition unit, and prediction unit are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0108] The processor contains a kernel, which retrieves the corresponding program unit from memory. One or more kernels can be configured, and adjusting kernel parameters can improve the accuracy of gas concentration prediction results.

[0109] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0110] This invention provides a computer-readable storage medium storing a program that, when executed by a processor, implements the aforementioned method for predicting gas concentration.

[0111] This invention provides a processor for running a program, wherein the program executes the gas concentration prediction method.

[0112] This application also provides an electronic device including one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for performing any of the above-described methods.

[0113] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs at least the following steps:

[0114] Step S101: Obtain gas concentration data in the coal mine and perform noise reduction processing on the gas concentration data to obtain the first gas concentration data;

[0115] Step S102: Use the first model to extract features from the first gas concentration data to obtain the second gas concentration data. The first model is a CNN model.

[0116] Step S103: Obtain the second model and update the second model to obtain the updated second model, which is the Elman model;

[0117] Step S104: The updated second model is used to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data. The prediction result refers to the prediction of gas concentration data in the future time period using the second gas concentration data.

[0118] The devices mentioned in this article can be servers, PCs, tablets, mobile phones, etc.

[0119] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having at least the following method steps:

[0120] Step S101: Obtain gas concentration data in the coal mine and perform noise reduction processing on the gas concentration data to obtain the first gas concentration data;

[0121] Step S102: Use the first model to extract features from the first gas concentration data to obtain the second gas concentration data. The first model is a CNN model.

[0122] Step S103: Obtain the second model and update the second model to obtain the updated second model, which is the Elman model;

[0123] Step S104: The updated second model is used to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data. The prediction result refers to the prediction of gas concentration data in the future time period using the second gas concentration data.

[0124] To enable those skilled in the art to better understand the technical solution of this application, the technical solution and technical effects of this application will be described below in conjunction with specific embodiments.

[0125] Example

[0126] This embodiment relates to a method for predicting gas concentration, such as... Figure 5 As shown, the method includes: first, acquiring data by inputting gas concentration data from underground coal mines; creating an Elman neural network model; updating the butterfly optimization algorithm to obtain the globally optimal solution set; updating the Elman neural network model using the updated butterfly optimization algorithm; adjusting the parameters of the Elman neural network model; using the updated Elman neural network model to predict the gas concentration; obtaining the prediction result; and ending the prediction process.

[0127] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0129] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0130] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0131] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0132] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0133] 1) The gas concentration prediction method of this application first acquires gas concentration data from underground coal mines and performs noise reduction processing on the gas concentration data to obtain first gas concentration data. Then, a first model is used to extract features from the first gas concentration data to obtain second gas concentration data. Next, a second model is acquired and updated to obtain an updated second model. Finally, the updated second model is used to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data. In this scheme, by performing noise reduction processing on the gas concentration data, more accurate first gas concentration data can be obtained. Using a CNN model to extract features from the first gas concentration data can enhance the original sequence. This scheme also improves the Elman model; the updated Elman model's prediction results are more accurate. Therefore, the updated Elman model is used to predict the gas concentration to obtain the prediction result corresponding to the second gas concentration data, thereby improving the accuracy of the gas concentration prediction result.

[0134] 2) The gas concentration prediction device of this application comprises: a first acquisition unit acquiring gas concentration data from underground coal mines and performing noise reduction processing on the gas concentration data to obtain first gas concentration data; an extraction unit using a first model to extract features from the first gas concentration data to obtain second gas concentration data; a second acquisition unit acquiring a second model and updating the second model to obtain an updated second model; and a prediction unit using the updated second model to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data. In this scheme, by performing noise reduction processing on the gas concentration data, more accurate first gas concentration data can be obtained. Using a CNN model to extract features from the first gas concentration data can enhance the original sequence. This scheme also improves the Elman model; the updated Elman model's prediction result is more accurate. Therefore, the updated Elman model is used to predict the gas concentration to obtain the prediction result corresponding to the second gas concentration data, thereby improving the accuracy of the gas concentration prediction result.

[0135] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method of predicting gas concentration, characterized by, include: Obtain methane concentration data from underground coal mines, and perform noise reduction processing on the methane concentration data to obtain first methane concentration data, including: The gas concentration data is decomposed using a wavelet denoising algorithm. Gas concentration data greater than or equal to a concentration threshold are classified as first-type concentration data, and gas concentration data less than the threshold are classified as second-type concentration data. This process involves decomposing the gas concentration data into second-type and first-type concentration data, then further decomposing the second-type concentration data, and repeating this process cyclically to obtain second-type and first-type concentration data at arbitrary scales. The first-type concentration data is then deleted, and the second-type concentration data is retained. Finally, the second-type concentration data is reconstructed using the wavelet denoising algorithm to obtain the first gas concentration data. The first gas concentration data after denoising is obtained by multiple decompositions and reconstructions. The second type of concentration data obtained after each wavelet decomposition is different from the first type of concentration data. The second type of concentration data has a high similarity to the original gas concentration data and represents most of the features of the original gas concentration data. The first type of concentration data contains some interference information. The second type of concentration data is retained while the first type of concentration data is removed for filtering. Each decomposition removes the first type of concentration data and retains the second type of concentration data. Then, the decomposition is performed again, removing the first type of concentration data and retaining the second type of concentration data. After multiple operations, the second type of concentration data is reconstructed to obtain the first gas concentration data after denoising. The first model is a CNN model. Obtain the second model and update it to obtain the updated second model, which is an Elman model, including: Obtain the second model, which includes an input layer, a hidden layer, and an output layer; Obtain the target coefficient, which is: ,in, This represents the target coefficient. Indicates the first The next iteration coefficient value, express The maximum value of the coefficient. express The minimum value of the coefficient. Represents a random number within a predetermined range. Represents a random number within the predetermined range. Indicates the first The next iteration Coefficient values; using the target coefficients and the DE algorithm, the population in the butterfly optimization algorithm is screened, and populations that do not meet the predetermined conditions are deleted to obtain an updated butterfly optimization algorithm. The target coefficients are then introduced into the butterfly optimization algorithm, and the formula for the iterative stage of the updated butterfly optimization algorithm is as follows: The formula for the final stage of the updated butterfly optimization algorithm is: r is a random number. This represents the optimal solution found among all solutions in the current iteration. The scent of the i-th butterfly; The updated butterfly optimization algorithm is used to update the weights corresponding to the output results of the input layer of the second model, so as to obtain the updated second model. The updated second model is used to process the second gas concentration data to obtain the prediction result corresponding to the second gas concentration data. The prediction result refers to the prediction of gas concentration data in the future time period using the second gas concentration data.

2. The method of claim 1, wherein, The butterfly optimization algorithm includes an initialization phase, an iteration phase, and a termination phase, wherein... The initialization phase includes determining the population size, number of iterations, and initial position. The iterative phase involves calculating and sorting the fitness and fragrance intensity of all individual butterflies, and storing the optimal fitness and the corresponding butterfly's location information. The final stage involves performing a global or local search within the population to update the butterfly's location information.

3. The method of claim 1, wherein, The first model is used to extract features from the first gas concentration data to obtain the second gas concentration data, including: The first model is used to extract features from the first gas concentration data. The first model includes an input layer, a convolutional layer, a pooling layer, and an output layer. The input layer is used to receive input data. The convolutional layer is used to perform convolution operations on the input data to obtain initial feature data. The pooling layer is used to perform pooling operations on the initial feature data to obtain the second gas concentration data. The output layer is used to output the second gas concentration data. The input data is the gas concentration data.

4. The method of claim 1, wherein, The updated second model is used to process the second gas concentration data to obtain the prediction results corresponding to the second gas concentration data, including: The updated second model is used to predict the historical second gas concentration data to obtain the prediction result. The updated second model is trained using multiple sets of training data. Each set of training data includes historical second gas concentration data acquired within a historical time period and the historical detection results corresponding to the historical second gas concentration data.

5. A gas concentration prediction device characterized by comprising: include: The first acquisition unit is used to acquire gas concentration data in the coal mine and perform noise reduction processing on the gas concentration data to obtain first gas concentration data. The gas concentration data is obtained by a target sensor. The unit includes a first processing module, a second processing module, and a third processing module. The first processing module is used to decompose the gas concentration data using a wavelet denoising algorithm. Gas concentration data greater than or equal to a concentration threshold is classified as first-type concentration data, and gas concentration data less than the concentration threshold is classified as second-type concentration data. The gas concentration data is decomposed into second-type and first-type concentration data, and the second-type concentration data is further decomposed. This decomposition process is repeated cyclically to obtain second-type and first-type concentration data at any scale. The second processing module is used to delete the first-type concentration data and retain the second-type concentration data. The third processing module is used to reconstruct the second-type concentration data using a wavelet denoising algorithm to obtain the first gas concentration data. The device is further used to obtain the denoised first gas concentration data by multiple decompositions and reconstructions. The second type concentration data obtained after each wavelet decomposition is different from the first type concentration data. The second type concentration data has a high similarity to the original gas concentration data and represents most of the features of the original gas concentration data. The first type concentration data is some interference information. The second type concentration data is retained while the first type concentration data is removed for filtering. Each decomposition removes the first type concentration data and retains the second type concentration data. Then, the device is decomposed again to remove the first type concentration data and retain the second type concentration data. After multiple operations, the obtained second type concentration data is reconstructed to obtain the denoised first gas concentration data. An extraction unit is used to extract features from the first gas concentration data using a first model to obtain the second gas concentration data, wherein the first model is a CNN model. The second acquisition unit is used to acquire the second model and update the second model to obtain the updated second model. The second model is an Elman model. It includes an acquisition module, a first update module, and a second update module. The acquisition module is used to acquire the second model, which includes an input layer, a hidden layer, and an output layer. The first update module is used to acquire the target coefficients, which are: ,in, This represents the target coefficient. Indicates the first The next iteration coefficient value, express The maximum value of the coefficient. express The minimum value of the coefficient. Represents a random number within a predetermined range. Represents a random number within the predetermined range. Indicates the first The next iteration Coefficient values; using the target coefficients and the DE algorithm, the population in the butterfly optimization algorithm is screened, and populations that do not meet the predetermined conditions are deleted to obtain an updated butterfly optimization algorithm. The target coefficients are then introduced into the butterfly optimization algorithm, and the formula for the iterative stage of the updated butterfly optimization algorithm is as follows: The formula for the final stage of the updated butterfly optimization algorithm is: r is a random number. This represents the optimal solution found among all solutions in the current iteration. The i-th butterfly's scent is represented by the second update module, which uses the updated butterfly optimization algorithm to update the weights corresponding to the output of the input layer of the second model, thereby obtaining the updated second model. The prediction unit is used to process the second gas concentration data using the updated second model to obtain the prediction result corresponding to the second gas concentration data. The prediction result refers to the prediction of gas concentration data in the future time period using the second gas concentration data.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program performs the method according to any one of claims 1 to 4.

7. An electronic device, comprising: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising methods for performing any one of claims 1 to 4.

Images