Gas waste heat utilization method and system based on artificial intelligence

Abstract

The invention relates to the technical field of gas waste heat utilization, and discloses a gas waste heat utilization method and system based on artificial intelligence. The method comprises the steps that sensor parameters of the gas extraction system are collected in real time and processed, and a standardized data set is obtained; the standardized data set is input into a fusion convolution Transform neural network for analysis modeling, and a gas waste heat prediction model is obtained; performing decision analysis on gas waste heat utilization based on the gas waste heat prediction model and a preset expert knowledge base to obtain a control strategy; and performing parameter regulation and control on a heat exchanger, a waste heat boiler, a circulating pump and control valve equipment according to the control strategy to obtain a gradient utilization execution scheme. The gas waste heat recovery strategy is dynamically adjusted, and the safety of the system is guaranteed while it is ensured that gas waste heat is efficiently recovered.

Description

Technical Field

[0001] The present application relates to the technical field of gas waste heat utilization, and in particular to a gas waste heat utilization method and system based on artificial intelligence. Background Art

[0002] Currently, a large amount of waste heat from gas produced during coal mining operations is not fully utilized, resulting in energy waste and an unnecessary burden on the environment. Traditional methods for recovering waste heat from gas rely primarily on fixed heat exchange equipment and simple control strategies. These methods, mostly based on empirical data and simple adjustment rules, struggle to cope with dynamic changes in gas concentration and environmental conditions, resulting in low and unstable waste heat recovery efficiency. Although some intelligent control technologies have been applied to energy management, in the complex environment of coal mining, how to use real-time data to accurately adjust recovery strategies, improve waste heat utilization efficiency, and ensure system safety remains a technical challenge.

[0003] Conventional control systems often rely on manual settings and adjustments, lacking real-time monitoring and precise adjustment of parameters such as gas concentration, temperature, and flow rate, making them incapable of handling the complex variables and emergencies inherent in coal mine environments. This often results in inadequate recovery of waste heat from gas, posing safety risks. For example, when gas concentrations are high, existing systems may be unable to reduce equipment loads in a timely manner, increasing the risk of explosion. Conversely, when gas concentrations are low, the system may not be able to efficiently recover waste heat, resulting in energy waste. Summary of the Invention

[0004] The present application provides an artificial intelligence-based gas waste heat utilization method and system for dynamically adjusting the gas waste heat recovery strategy to ensure efficient recovery of gas waste heat while protecting the safety of the system.

[0005] In the first aspect, the present application provides a method for utilizing waste heat from gas based on artificial intelligence, which includes: real-time collection and processing of sensor parameters of the gas extraction system to obtain a standardized data set; inputting the standardized data set into a fused convolutional Transformer neural network for analysis and modeling to obtain a waste heat from gas prediction model; performing decision analysis on waste heat from gas utilization based on the waste heat from gas prediction model and a preset expert knowledge base to obtain a control strategy; and performing parameter control on heat exchangers, waste heat boilers, circulating pumps, and control valve equipment according to the control strategy to obtain a cascade utilization execution plan.

[0006] In a second aspect, the present application provides an artificial intelligence-based gas waste heat utilization system, the artificial intelligence-based gas waste heat utilization system comprising: The processing module is used to collect and process the sensor parameters of the gas extraction system in real time to obtain a standardized data set; A modeling module is used to input the standardized data set into a fused convolutional Transformer neural network for analysis and modeling to obtain a gas waste heat prediction model; An analysis module, configured to perform decision analysis on gas waste heat utilization based on the gas waste heat prediction model and a preset expert knowledge base to obtain a control strategy; The control module is used to control the parameters of the heat exchanger, waste heat boiler, circulation pump, and control valve equipment according to the control strategy to obtain a cascade utilization execution plan.

[0007] In a third aspect, an artificial intelligence-based gas waste heat utilization device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the artificial intelligence-based gas waste heat utilization device executes the above-mentioned artificial intelligence-based gas waste heat utilization method.

[0008] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, which, when executed on a computer, enables the computer to execute the above-mentioned artificial intelligence-based gas waste heat utilization method.

[0009] The technical solution provided in this application utilizes an artificial intelligence algorithm to dynamically analyze data such as gas concentration, flow rate, and temperature, effectively identifying and predicting the availability and optimal utilization of waste gas heat. Through precise model training, the system can adjust waste gas heat recovery strategies in real time and optimize control strategies based on factors such as gas concentration, flow rate, and temperature. This significantly improves waste gas heat utilization efficiency, avoiding the inefficiency and energy waste associated with traditional empirical control methods. The artificial intelligence algorithm also dynamically adjusts system operating parameters based on the correlation between real-time and historical data, ensuring that each device (such as heat exchangers, waste heat boilers, and circulating pumps) achieves optimal operating conditions under varying operating conditions. For example, when gas concentration fluctuates, the system automatically adjusts parameters such as heat exchanger flow rate and waste heat boiler load. When gas concentration is high, the heat exchanger flow rate and waste heat boiler load are reduced to prevent equipment overload and ensure safety. When gas concentration is low, the system increases the heat exchanger flow rate to improve waste heat recovery efficiency. In this scenario, artificial intelligence technology can process large amounts of sensor data in real time and make precise decisions, ensuring maximum gas waste heat recovery without overloading or inefficient equipment. By precisely controlling equipment operation, the risk of gas explosions caused by improper operation is avoided. For example, the system dynamically adjusts control strategies based on a gas concentration prediction model. When gas concentrations exceed safety thresholds, emergency measures are swiftly implemented, effectively mitigating safety risks. Furthermore, the system uses real-time data feedback to adjust and immediately identify and address any anomalies, ensuring the entire recovery process remains within a safe and controllable range. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 This is a schematic diagram of an embodiment of a method for utilizing waste heat from gas based on artificial intelligence in an embodiment of the present application; Figure 2 This is a schematic diagram of an embodiment of a gas waste heat utilization system based on artificial intelligence in an embodiment of the present application; Figure 3 It is a schematic block diagram of the structure of a gas waste heat utilization device based on artificial intelligence in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a method and system for utilizing waste heat from gas based on artificial intelligence. The terms first, second, third, fourth, etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms include or have and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the method for utilizing waste heat from gas based on artificial intelligence includes: Step S101: collecting and processing sensor parameters of the gas extraction system in real time to obtain a standardized data set; Step S102: Input the standardized data set into the fused convolutional Transformer neural network for analysis and modeling to obtain a gas waste heat prediction model; Step S103: performing decision analysis on gas waste heat utilization based on the gas waste heat prediction model and a preset expert knowledge base to obtain a control strategy; Step S104: Regulate the parameters of the heat exchanger, waste heat boiler, circulation pump, and control valve equipment according to the control strategy to obtain a cascade utilization execution plan.

[0014] It is understandable that the execution subject of this application can be an artificial intelligence-based gas waste heat utilization system, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0015] Specifically, a sensor network collects key parameters of the gas extraction system in real time, including gas concentration, flow rate, temperature, pressure, and oxygen content. Gas concentration is detected using infrared absorption spectroscopy, enabling high-precision measurements within the range of 0-100%, with a measurement error of no more than ±0.1%. The flow sensor utilizes ultrasonic technology, with an accuracy of ±1% and a measurement range of 0-100 m³ / min, ensuring accurate monitoring of gas flow. All of this data is transmitted to the data processing center via Industrial Ethernet using the Modbus-TCP protocol, ensuring stable real-time data transmission and processing. After preliminary processing, the collected data is subjected to a sliding window median filter algorithm for noise removal. This algorithm collects multiple samples around each data point based on a specific window size and takes the median, effectively removing sudden noise such as transient interference. A Kalman filter is then applied to smooth the data. A Kalman filter is a recursive estimation method based on a state-space model. It adjusts the current estimate based on the discrepancy between prediction and measurement, resulting in smoother data and reduced error. To handle abnormal data that may be caused by factors such as gas outbursts, the system uses the 3σ principle for outlier detection. Specifically, any outlier that deviates from the mean by more than three standard deviations is identified and marked as an error. These missing values ​​are then filled in using linear interpolation to ensure the integrity of the data series.

[0016] All data undergoes Z-score normalization. This process transforms each data point into a standard normal distribution with a mean of 0 and a standard deviation of 1, eliminating differences between different dimensions. After data preprocessing, the normalized data is input into a fused convolutional Transformer neural network for analysis and modeling. The convolutional neural network (CNN) first extracts spatial features from the input data, specifically the spatial distribution of gas concentration. Each convolutional layer in the network operates using a 3×3 convolution kernel to extract local information from the data at different feature scales. For example, suppose that after the convolution operation, the spatial feature vector of gas concentration is a high-dimensional array. This is then reduced in dimensionality through max pooling, with a pooling kernel size of 2×2 and a stride of 2. This operation reduces the dimensionality of the data, removing unimportant details while retaining key information crucial for gas waste heat utilization.

[0017] The spatial features extracted by convolution are further processed by the Transformer model. The Transformer model is based on a self-attention mechanism, which learns and weights the relationships between input features by calculating the similarity between the query, key, and value matrices. The self-attention mechanism enhances the model's focus on important features by calculating a weighted average of each feature point. For example, in the analysis of gas concentration data, the self-attention mechanism dynamically adjusts its focus on gas concentration features based on historical information such as temperature and flow rate to improve model accuracy. The Transformer output undergoes further dimensionality reduction through a max pooling layer to reduce memory consumption and avoid overfitting.

[0018] The Long Short-Term Memory (LSTM) module plays a key role in further capturing time series features. LSTM is a powerful tool for processing time series data, effectively capturing long-term dependencies within the data. In analyzing gas flow and temperature data, LSTM focuses on past time steps to predict future trends. For example, when gas flow changes significantly, the LSTM uses a memory mechanism to remember these changes and, combined with current input data, predict future gas flow trends, thereby helping the model make more accurate predictions for waste heat utilization. Through the multi-layer combination and training of CNN, Transformer, and LSTM, the resulting waste heat prediction model can accurately predict waste heat availability and optimal utilization, providing strong support for subsequent decision-making.

[0019] Based on the trained gas waste heat prediction model and a pre-built expert knowledge base, the system conducts decision analysis on waste heat utilization. When the system predicts high gas concentrations, the expert knowledge base invokes high calorific value utilization rules, instructing the equipment to enhance waste heat recovery. For example, at high gas concentrations, the heat exchanger's heat transfer capacity needs to be increased to improve heat recovery efficiency. If the predicted concentration is in the medium or low range, the system adjusts the heat transfer efficiency, dynamically adjusting equipment parameters based on changes in gas concentration to ensure optimal resource utilization.

[0020] The control strategies generated by these decisions are translated into specific equipment control parameters. The control system, through a hierarchical control architecture encompassing equipment, control, and management layers, adjusts the operating parameters of equipment such as heat exchangers, waste heat boilers, circulation pumps, and control valves in real time. When gas concentrations fluctuate, the control system automatically adjusts the operating status of each device. For example, when gas concentrations are high, the combustion intensity of the waste heat boiler is reduced to avoid excessive heat waste, while when gas concentrations are low, the flow rate of the heat exchanger is increased to ensure sufficient heat recovery. Each device's operation is precisely calculated and executed within predefined safety thresholds, ensuring maximum waste heat recovery without compromising safety. Through continuous monitoring and optimization, the system implements a tiered utilization strategy: high-temperature waste heat is used for power generation, medium-temperature waste heat is used for industrial heating, and low-temperature waste heat is used for heating or hot water supply, thereby maximizing waste heat recovery.

[0021] In the embodiments of this application, an artificial intelligence algorithm dynamically analyzes data such as gas concentration, flow rate, and temperature, effectively identifying and predicting the availability and optimal utilization of waste gas heat. Through precise model training, the system can adjust waste gas heat recovery strategies in real time and optimize control strategies based on factors such as gas concentration, flow rate, and temperature. This significantly improves waste gas heat utilization efficiency and avoids the inefficiency and energy waste associated with traditional empirical control methods. The artificial intelligence algorithm also dynamically adjusts system operating parameters based on the correlation between real-time and historical data, ensuring that each device (such as heat exchangers, waste heat boilers, and circulating pumps) achieves optimal operating conditions under varying operating conditions. For example, when gas concentration fluctuates, the system automatically adjusts parameters such as heat exchanger flow rate and waste heat boiler load. When gas concentration is high, the heat exchanger flow rate and waste heat boiler load are reduced to prevent equipment overload and ensure safety. When gas concentration is low, the system increases the heat exchanger flow rate to improve waste heat recovery efficiency. In this scenario, artificial intelligence technology can process large amounts of sensor data in real time and make precise decisions, ensuring maximum gas waste heat recovery without overloading or inefficient equipment. By precisely controlling equipment operation, the risk of gas explosions caused by improper operation is avoided. For example, the system dynamically adjusts control strategies based on a gas concentration prediction model. When gas concentrations exceed safety thresholds, emergency measures are swiftly implemented, effectively mitigating safety risks. Furthermore, the system uses real-time data feedback to adjust and immediately identify and address any anomalies, ensuring the entire recovery process remains within a safe and controllable range.

[0022] In a specific embodiment, the process of executing step S101 may specifically include the following steps: Parameters are collected through sensor devices at key nodes in the gas extraction system to obtain raw data on gas concentration, flow, temperature, pressure, and oxygen content; The original data were filtered using a sliding window median filter to remove the interference noise in the coal mine gas extraction environment and obtain preliminary filtered data. Performing data smoothing on the preliminary filtered data to obtain smoothed data; The 3σ principle is used to identify the outliers caused by sudden gas outburst on the smoothed data and fill them with linear interpolation to obtain the data series; The data series was normalized using the Z-Score standardization method to obtain the standardized parameters.

[0023] Specifically, sensors at key nodes in the gas extraction system collect real-time data. These sensors collect multiple parameters, including gas concentration, flow rate, temperature, pressure, and oxygen content. Gas concentration is typically measured using infrared absorption spectroscopy, which ensures accurate measurement data even under the drastic concentration fluctuations found in coal mine environments. Flow rate is measured using ultrasonic technology, while temperature and pressure are measured using platinum resistance sensors and piezoresistive sensors, respectively, providing highly accurate results. Sensor data is transmitted to the data processing center in real time via the Modbus-TCP protocol. The raw data is filtered using a sliding window median filter to remove noise caused by external interference. The sliding window median filter is a commonly used signal processing method that sorts the values ​​within a certain range around a data point and selects the median as the value for the current data point. This method effectively removes short-term noise caused by factors such as transient fluctuations and equipment jitter. The sliding window size is set to 11 data points, with a step size of 1 sample point. This method preserves signal characteristics while removing most transient noise. For example, suppose that at a certain moment, the collected original data of gas concentration is [5.3, 5.4, 5.1, 5.7, 5.2]. After sliding window median filtering, the new data point will be 5.3.

[0024] After initial filtering, the data needs to undergo data smoothing. The purpose of data smoothing is to reduce random fluctuations in the data and make data trends more distinct. The Kalman filter algorithm is used. Kalman filtering not only considers the current measurement value but also incorporates predictions about system dynamics, gradually adjusting the estimated value to reduce errors. For parameters that are significantly affected by external factors, such as gas flow and temperature, Kalman filtering can effectively eliminate errors and generate smoother values. For example, if the temperature data at a certain moment is affected by a sudden electrical fluctuation, causing the measured value to be higher, the Kalman filter will combine the measurement data from the previous moment to provide a more accurate correction value, thereby ensuring data stability.

[0025] The smoothed data is then tested for outliers. Since gas extraction systems may experience sudden gas outbursts during actual operation, causing sampled data at a given moment to deviate significantly from the normal range, it is necessary to identify outliers in the data. In this paper, the 3σ principle is used to identify these outliers. The 3σ principle is based on the standard deviation principle in statistics, stating that if a data point deviates from the mean by more than three times the standard deviation, it is considered an outlier. Specifically, for each measured parameter, its mean and standard deviation are first calculated. For example, if the mean of the gas concentration at a given moment is 5.0 and the standard deviation is 0.2, then a measured value of 5.7 at that moment has a deviation of 0.7, exceeding the limit of three times the standard deviation (0.6) and therefore considered an outlier. Linear interpolation is used to process outliers, filling in the gaps by using the linear relationship between the two preceding and following normal data points to ensure data continuity. After smoothing and outlier processing, the data is then subjected to Z-score normalization, converting parameters of different dimensions into a unified scale, enabling comparison and calculation within the same model. All data are transformed into a standard normal distribution with a mean of 0 and a standard deviation of 1. After standardization, measurements in different units such as gas concentration, flow rate, and temperature will have the same scale.

[0026] In a specific embodiment, the process of executing step S102 may specifically include the following steps: The standardized dataset was input into a convolutional neural network with three convolutional layers for processing. Each convolution layer used 32, 64, and 128 convolution kernels of 3×3 size, respectively, to obtain the spatial characteristics of the gas concentration distribution. The spatial features are reduced in dimension by using a batch normalization layer and a 2×2 maximum pooling operation with a stride of 2 after each convolution layer to obtain the reduced-dimensional spatial feature representation; The spatial feature representation after dimensionality reduction is input into a multi-head self-attention mechanism with 8 heads, the hidden layer dimension is set to 256, and the feedforward network dimension is set to 1024 to obtain the attention-weighted feature; A 2×2 maximum pooling operation with a step size of 2 is used on the attention weighted features to reduce memory consumption and obtain a simplified attention feature; The simplified attention features are input into a two-layer bidirectional LSTM structure, with the first layer containing 128 hidden units and the second layer containing 64 hidden units, to obtain the temporal variation features of gas flow and temperature. By training a learnable attention network to calculate the weight coefficients of spatial features, attention features and temporal features, a loss function is constructed for optimization training to obtain a gas waste heat prediction model.

[0027] Specifically, when a standardized dataset is fed into a convolutional neural network consisting of three convolutional layers, multidimensional parameters such as gas concentration, flow rate, and temperature are organized into a two-dimensional tensor format, with each parameter representing a channel. The first convolutional layer extracts features from the input data using 32 3×3 convolution kernels with a stride of 1 and padding of 1. The output feature map has the same size as the input, but with 32 channels. The second convolutional layer further extracts features from the output of the first layer using 64 3×3 convolution kernels, also maintaining a stride of 1 and padding of 1, increasing the number of output channels to 64. The third convolutional layer further extracts deeper features using 128 3×3 convolution kernels with the same stride and padding, bringing the number of output channels to 128. Using multiple convolution kernels, the spatial distribution of gas parameters can be learned, for example, by identifying spatial correlation patterns between gas concentration and temperature, thereby capturing the spatial characteristics of gas parameters.

[0028] For dimensionality reduction of spatial features, a batch normalization layer and a maximum pooling operation are immediately followed by each convolution layer. The batch normalization layer normalizes the data by calculating the mean and variance of the input data, and then performs a linear transformation, effectively reducing the problem of internal covariate shift. The maximum pooling operation uses a 2×2 window configuration with a stride of 2, outputting the maximum value in each 2×2 area as the representative value of the area, reducing the feature map to half its original size in both the horizontal and vertical directions. After three layers of convolution and pooling, the spatial size of the feature map is significantly reduced, but the key spatial feature information of the gas concentration distribution is retained.

[0029] When the reduced spatial feature representation is input into the multi-head self-attention mechanism, linear projection is performed to generate three vectors: query (Q), key (K), and value (V). In an eight-head configuration, the feature vector is split into eight parts and processed in parallel, with each head focusing on a different feature subspace, corresponding to different parameter focus points in the gas waste heat utilization system. A hidden layer dimension of 256 means each head has a dimension of 32, sufficient to capture the complex correlations between gas parameters. The feedforward network dimension of 1024 provides sufficient nonlinear transformation capabilities, enhancing feature representation. During the self-attention calculation process, certain gas parameters (such as high-concentration areas) are given higher weights, forming attention-weighted features. A 2×2 max pooling operation with a stride of 2 is applied to the attention-weighted features to further compress the feature representation, extract the most significant features, and reduce computational overhead. When these refined attention features are input into a two-layer bidirectional LSTM structure, the LSTM's gating mechanism enables them to effectively capture long-term dependencies between gas parameters. The first layer contains 128 hidden units, and the second layer contains 64 hidden units. This bidirectional design allows for learning sequence information from both the past and future directions. This structure is particularly suitable for processing parameters such as gas flow and temperature that have obvious timing characteristics, and can identify timing patterns such as sudden changes in gas concentration, flow fluctuations, and temperature gradients.

[0030] A learnable attention network is trained to fuse spatial, attentional, and temporal features. This is achieved by constructing a fully connected neural network whose input is a concatenated vector of the three features and whose output is three weight coefficients. The loss function is designed as a weighted combination of mean squared error, cross entropy, and smoothed L1 loss, specifically targeting heat prediction, utilization classification, and system parameter prediction.

[0031] In a specific embodiment, the process of executing step S103 may specifically include the following steps: The fusion prediction results of spatial features, attention features, and time series features are obtained from the gas residual heat prediction model. When the predicted gas concentration is higher than the upper limit of the safe explosion, it is determined to be a high concentration interval. When the predicted gas concentration is between the lower and upper limits of the safe explosion, it is determined to be a medium concentration interval. When the predicted gas concentration is lower than the lower limit of the safe explosion, it is determined to be a low concentration interval. The gas residual heat grade classification data is obtained. According to the gas waste heat level classification data, the expert knowledge base organized based on the ontological structure is queried. When the concentration is in the high range, the high calorific value utilization rule is called; when the concentration is in the medium range, the medium calorific value utilization rule is called; when the concentration is in the low range, the low calorific value utilization rule is called to obtain the preliminary control reference value; A decision scoring function is constructed for the preliminary control reference value. By calculating the ratio of energy recovery efficiency to the theoretical maximum value, the margin between equipment operating parameters and safety thresholds, the difference between the economic value of recovered heat and energy consumption costs, and the reduced carbon emissions, a weighted comprehensive evaluation index is obtained. The weighted comprehensive evaluation index is input into the hierarchical multi-agent reinforcement learning framework as a reward signal. The main agent decomposes the tasks and assigns them to sub-agents. The sub-agents are responsible for optimizing the heat exchange system, boiler system, and circulation system, and a collaboratively optimized state-action strategy is obtained. Based on the collaborative optimization state-action strategy, multiple sets of candidate control parameters are generated. When the ambient temperature in the coal mine is higher than the upper comfortable limit, the ventilation and cooling scheme is given priority. When the ambient temperature is lower than the lower comfortable limit, the heating scheme is given priority. When the ambient temperature is between the lower and upper comfortable limits, the power generation and heating schemes are balanced, thus obtaining a multi-objective trade-off control parameter set. The multi-objective trade-off control parameter set is subjected to evolutionary multi-objective optimization based on the super volume index. The reference point adaptive adjustment mechanism and the decision space diversity maintenance strategy are introduced. The control values ​​of the heat exchanger, waste heat boiler, circulating pump and control valve are determined according to the gas extraction working conditions to obtain the control strategy.

[0032] Specifically, data is collected from key nodes in the gas extraction system. Sensors at each node collect real-time information on gas concentration, flow, temperature, pressure, and oxygen content. The gas waste heat prediction model extracts spatial, attention, and temporal features from the raw data and uses these features to predict the availability and optimal utilization of gas waste heat. Spatial features are extracted using a convolutional neural network (CNN), attention features are generated using the self-attention mechanism of the Transformer model, and temporal features are modeled using a long short-term memory (LSTM) network. These three features work together to comprehensively capture the dynamic changes in parameters such as gas flow, concentration, and temperature. When the predicted gas concentration exceeds the upper safe explosion limit, the model identifies the gas as being in the high concentration range. If the gas concentration is between the lower and upper safe explosion limits, it is identified as being in the medium concentration range; if the gas concentration is below the lower explosion limit, it is identified as being in the low concentration range. The model selects corresponding calorific value utilization rules based on different gas concentration ranges, determining whether to increase or decrease waste heat recovery. For example, in the high concentration range, the energy density of gas is higher, and the system will choose to call the high calorific value utilization rule to ensure maximum heat recovery; in the low concentration range, a lower calorific value utilization rule will be applied to avoid excessive recovery and cause the system load to be too high.

[0033] After determining the gas concentration level, the corresponding calorific value utilization rules are queried through an ontologically structured expert knowledge base. The expert knowledge base includes preset calorific value recovery strategies for different concentration ranges. These rules not only cover waste heat utilization methods at different concentrations but also take into account factors such as safety, energy efficiency, and environmental benefits. For example, in high-concentration ranges, the knowledge base may recommend the use of high-efficiency heat exchangers and waste heat boilers to maximize gas waste heat recovery; while in low-concentration ranges, the system recommends a mild recovery strategy to avoid unnecessary energy waste.

[0034] Based on the preliminary control reference value, the system will construct a decision scoring function. The purpose of this scoring function is to evaluate different control strategies and select the best option. The design of the scoring function takes into account multiple aspects, including the ratio of energy recovery efficiency to the theoretical maximum value, the margin between equipment operating parameters and safety thresholds, the difference between the economic value of recovered heat and energy consumption costs, and the amount of carbon emissions reduced by recovering heat. Each factor has a different weight in the decision-making, and the final weighted comprehensive evaluation index will be used to determine the optimal control strategy. For example, under a specific operating condition, if the energy recovery efficiency is high but the equipment safety is poor, the recovery amount will be reduced to ensure safety, and vice versa.

[0035] To make the decision-making process more intelligent, the system adopts a hierarchical multi-agent reinforcement learning framework. The master agent is responsible for breaking down the overall task and assigning tasks to sub-agents. Each sub-agent is responsible for optimizing a different subsystem, such as the heat exchange system, boiler system, and circulation system. Through reinforcement learning, the sub-agents continuously adjust their strategies to achieve system-level collaborative optimization. In this process, the reinforcement learning reward signal comes from the decision scoring function, and the system learns the optimal state-action strategy through training. For example, in a practical application, when the master agent assesses that the current gas concentration is high, it may instruct the sub-agent to increase the load on the heat exchange system while reducing the combustion intensity of the boiler to optimize overall waste heat recovery. Based on the collaborative optimization state-action strategy, the system generates multiple sets of candidate control parameters. The system will make different adjustments under different temperature environments. During the utilization of waste heat from gas, if the ambient temperature underground in the coal mine is higher than the upper limit of comfort, the system will give priority to lowering the ambient temperature through ventilation and cooling to ensure the comfort of the working environment; if the ambient temperature is lower than the lower limit of comfort, the system will give priority to heating solutions; when the ambient temperature is within the comfortable range, the system will balance the power generation and heating solutions to optimize energy utilization.

[0036] All generated sets of control parameters are further optimized using evolutionary multi-objective optimization based on super-volume metrics. This optimization method incorporates a reference point adaptive adjustment mechanism and maintains a strategy through decision space diversity, ensuring a balance between different objectives. For example, in some cases, the system may choose to balance increasing energy recovery efficiency with reducing carbon emissions. Ultimately, the specific control values ​​for the heat exchanger, waste heat boiler, circulation pump, and control valves are determined based on the gas extraction operating conditions, resulting in the final control strategy.

[0037] In a specific embodiment, the process of performing the step of querying the expert knowledge base organized based on the ontology structure according to the gas waste heat grade classification data may specifically include the following steps: The experience in handling different gas concentration ranges in the actual coal mine production environment is structured and organized. By establishing the correlation between equipment control rules, safety and explosion prevention rules, energy efficiency optimization rules, and emergency handling rules, a knowledge base for gas waste heat utilization is obtained. Input the gas waste heat classification data into the knowledge retrieval system, and quickly match similar historical working conditions when the underground gas concentration fluctuates to obtain a treatment plan; Conduct safety review of the treatment plan, identify and eliminate dangerous operations in the plan according to mine safety production standards, and obtain operating instructions that comply with coal mine safety regulations; Convert the operating instructions into control parameters for the heat exchange system, waste heat boiler system, and circulation system. When the gas concentration suddenly increases, reduce the waste heat boiler combustion intensity. When the gas concentration gradually decreases, increase the heat exchanger flow rate to obtain the operating parameters of each system. A cascaded utilization process for coal mine gas waste heat is constructed based on the operating parameters of each system. When the gas grade is high, the power generation system is prioritized; when the gas grade is medium, industrial heat is prioritized; when the gas grade is low, domestic hot water and heating systems are prioritized, resulting in a cascaded utilization plan. The cascade utilization plan is adapted to the actual situation of the coal mine. By considering the mine production plan, seasonal heat demand and equipment maintenance cycle, the theoretical parameters are actually corrected to obtain preliminary control reference values.

[0038] Specifically, the experience gained from handling different gas concentration ranges in the actual production environment of coal mines is structured and organized to ensure that actual production experience can be converted into system-operable rules, thereby generating a knowledge base for gas waste heat utilization. This knowledge base includes equipment control rules, safety and explosion prevention rules, energy efficiency optimization rules, and emergency response rules. These rules are organized using an ontological structure, ensuring that appropriate response strategies can be quickly retrieved in different situations. For example, in areas with high gas concentrations, equipment control rules may indicate increasing the load on the heat exchanger, while when the gas concentration is low, it is recommended to reduce the operating intensity of the heat exchanger to reduce unnecessary energy consumption. The development of these control rules relies on the analysis of historical operating conditions and the summary of experience, allowing the knowledge base to cover a variety of different operating conditions and response measures.

[0039] Once the gas waste heat classification data is obtained, it is input into the knowledge retrieval system. When the underground gas concentration fluctuates, the system quickly matches similar historical operating conditions based on real-time data to obtain a treatment plan. This process is actually a query process for historical data. By comparing the similarity of current parameters such as gas concentration, flow rate, and temperature with historical data, the system can quickly identify waste heat utilization solutions that have been successfully applied under similar operating conditions. For example, if the current gas concentration is 5.2%, the system will retrieve heat exchange strategies that have been successfully used in this concentration range in the past. This real-time matching based on historical data ensures that the system avoids human intervention when handling the current situation, improving the efficiency and accuracy of decision-making.

[0040] After matching treatment plans are identified, they undergo a safety review. The unique nature of coal mine production requires that all operations comply with strict safety standards, making a safety review of each treatment plan essential. This review process relies on mine safety standards, and the system identifies and eliminates potentially dangerous operations from the plans based on these standards. For example, a treatment plan might recommend increasing the load on the heat exchanger when gas concentrations are high, which could cause the system to overload and increase the risk of explosion. In this case, the system will modify the plan, reduce the load on the heat exchanger, and implement other safety measures to ensure that all operations comply with coal mine safety regulations.

[0041] The reviewed operating instructions are further converted into control parameters for the heat exchange system, waste heat boiler system, and circulation system. These control parameters specifically indicate the operating status of each device under the current operating conditions. For example, if gas concentration suddenly increases, the system will instruct the waste heat boiler to reduce combustion intensity to avoid excessive heat generation, thereby reducing energy waste and the risk of overheating. Conversely, as gas concentration gradually decreases, the system will increase the flow rate in the heat exchanger to more effectively recover and utilize heat. These control parameters are derived from a comprehensive analysis of real-time and historical data, ensuring optimized equipment operation.

[0042] Based on the operating parameters of these devices, a cascaded utilization process for coal mine gas waste heat will be established. Depending on the gas grade, waste heat will be prioritized for allocation to different systems. For example, when gas concentrations are high, the system prioritizes high-grade heat energy for power generation. When gas concentrations are medium, the heat energy is allocated to industrial heating. When gas concentrations are low, the heat energy is used for heating or domestic hot water systems. This cascaded utilization plan ensures maximum utilization of each type of heat energy while avoiding waste of high-grade heat energy.

[0043] The cascade utilization plan is then adapted based on the actual coal mine production situation. For example, in winter, coal mines may face higher heating demand, so the cascade utilization plan needs to be adjusted to prioritize the use of heat energy for heating systems. During peak production periods, waste heat is more likely to be used for power generation and industrial heat. In addition, the maintenance cycle of equipment is also an important factor in adjusting the plan. For equipment undergoing maintenance, the system automatically dispatches heat energy to other available equipment to ensure maximum utilization of waste heat. Therefore, all cascade utilization plans will be dynamically adjusted based on factors such as the coal mine's actual production plan, seasonal heat demand, and equipment maintenance cycle to optimize thermal energy utilization efficiency.

[0044] These adjusted cascade utilization plans will generate preliminary control reference values, which will serve as the basis for the next optimization decision. In actual application, assuming that the coal mine is currently facing a gas concentration of 5.0%, the system selects a control strategy suitable for the current concentration range by comparing historical data and real-time data. After a safety review, the system decided to increase the flow rate of the heat exchanger by 10% and reduce the load of the waste heat boiler by 5%. These control parameters will be adjusted based on the system's evaluation to ensure that the operation of each link can achieve optimal gas waste heat utilization while ensuring safety. This process demonstrates how the artificial intelligence-based gas waste heat utilization method can achieve efficient energy utilization and safety assurance in the coal mine production process through data analysis, knowledge base support and intelligent optimization.

[0045] In a specific embodiment, the process of executing step S104 may specifically include the following steps: The control strategy is converted into a hierarchical control instruction sequence, and the adjustment parameters of each device are obtained by establishing a three-level control architecture of device layer, control layer and management layer; Conduct coal mine safety production constraint checks on adjustment parameters, compare parameter values ​​with safety limits and apply safety margins to obtain safety control instructions that meet explosion protection requirements; A heat exchanger parameter control scheme is constructed based on safety control instructions. The heat exchange efficiency is controlled by adjusting the inlet and outlet temperature difference and flow parameters. When the gas concentration fluctuates, the heat exchange rate is dynamically adjusted to obtain the heat exchanger operation curve. Formulate a waste heat boiler parameter control plan based on the heat exchanger operating curve, and obtain the boiler system operating curve by coordinating the three key parameters of boiler load, feed water flow, and steam pressure; Design a circulation system control scheme based on the heat exchanger operating curve and the boiler system operating curve. Ensure quantitative fluid delivery by adjusting the circulation pump speed and pipe network pressure to obtain the circulation system operating curve. The heat exchanger operating curve, boiler system operating curve and circulation system operating curve are systematically integrated and coordinated, and a cascade utilization implementation plan is obtained according to the principle of using high-grade thermal energy for power generation, medium-grade thermal energy for industrial heat, and low-grade thermal energy for heating.

[0046] Specifically, the control strategy is broken down from high-level to low-level, forming adjustment parameters. For example, when the system determines that gas concentration is within a dangerous range based on real-time data and predictive models, the control strategy initiates appropriate countermeasures, such as reducing the load on the heat exchanger or adjusting the operating intensity of the waste heat boiler. These control instructions are precisely translated into specific operating parameters for each device. These device adjustment parameters include, but are not limited to, the inlet and outlet temperature difference of the heat exchanger, flow rate, load on the waste heat boiler, and speed of the circulating pump. Every parameter change is strictly inspected and confirmed in accordance with safety production standards. When setting adjustment parameters, the system performs safety production constraint checks to ensure that all control instructions do not exceed coal mine safety thresholds. Specifically, this safety check compares each parameter value against pre-set safety limits. For example, if the gas concentration limit is set at 5%, if the system predicts that the gas concentration will exceed this value, a safety control instruction will be triggered, automatically adjusting the device's operating status to prevent overload or other hazards. Furthermore, a certain safety margin is applied to ensure safe and stable operation of the equipment even under unusual operating conditions.

[0047] After receiving safety control instructions that meet explosion-proof requirements, the system constructs a parameter control plan for the heat exchanger based on these instructions. This control plan involves adjusting the inlet and outlet temperature differential and flow rate to control heat exchange efficiency. When gas concentration fluctuates, the system dynamically adjusts the heat exchange rate. For example, when gas concentration increases, the system may reduce the heat exchanger flow rate to avoid excessive heat recovery. Conversely, when gas concentration is low, the system increases the flow rate to improve waste heat recovery efficiency. Through this adjustment process, the system can reflect changes in gas concentration in real time and adjust the heat exchanger's operating status accordingly to ensure efficient heat recovery.

[0048] Based on the heat exchanger's operating curve, the system develops a control plan for the waste heat boiler. Waste heat boiler operation requires the coordination of multiple parameters, including boiler load, feedwater rate, and steam pressure. These parameters interact to determine the boiler's thermal output. For example, in conditions of high gas concentration, the boiler may need to reduce combustion intensity to minimize heat generation, while in conditions of lower concentrations, the boiler load can be increased to maximize heat recovery. By coordinating these key parameters, the system can precisely control the boiler's operating status, avoid excessive fuel consumption, and improve energy efficiency.

[0049] The system also designed a control scheme for the circulation system based on the operating curves of the heat exchanger and boiler. The circulation system's task is to ensure that the fluid is delivered to each device in a measured quantity by adjusting the circulation pump's speed and pipe network pressure. Flow stability is crucial for heat recovery, making the regulation of the circulation system particularly critical. For example, when gas concentration is high, the system may reduce flow to reduce equipment burden and ensure safe and stable operation; when concentration is low, the system may increase flow to optimize the heat recovery process. Through these dynamic adjustments, the circulation system ensures the coordinated operation of the heat exchange and boiler systems, making the waste heat recovery process more efficient.

[0050] By integrating and coordinating the operating curves of the heat exchanger, boiler system, and circulation system, the system develops a tiered utilization plan based on the principle of using high-grade heat energy for power generation, medium-grade heat energy for industrial heat, and low-grade heat energy for heating. This plan ensures optimal utilization of different types of heat energy. For example, when gas concentration is high and heat energy is high-grade, the system prioritizes heat energy for power generation, thereby providing higher energy output for the coal mine. Meanwhile, when gas concentration is low, low-grade heat energy is used for heating and hot water supply, optimizing overall energy distribution.

[0051] In a specific embodiment, the process of converting the control strategy into a hierarchical control instruction sequence may specifically include the following steps: Decompose the control strategy into control levels, and obtain the three-level control function definition by determining the responsibility boundaries of real-time control tasks, process control tasks, and optimization scheduling tasks; The device layer communication topology is established based on the three-level control function definition, and the device layer data interaction framework is obtained by connecting sensors and actuators to the local control unit through the fieldbus network; The control layer control algorithm is constructed based on the data interaction framework of the equipment layer. The real-time control strategy is obtained by converting the relationship between gas concentration and equipment parameters into improved fuzzy adaptive PID control parameters. Design a management-level scheduling strategy based on real-time control strategies, and obtain a time-segmented execution plan by decomposing long-term optimization objectives into a sequence of short-term control objectives. Perform control instruction conversion on the time-segment execution plan, and obtain the device control instruction set by converting the abstract control strategy into parameter setting values ​​that can be executed by specific devices; A three-level inter-control system communication mechanism is established based on the device control instruction set. By defining the uplink data aggregation process and the downlink instruction distribution process, the adjustment parameters of each device are obtained.

[0052] Specifically, the control strategy is decomposed into three levels: real-time control tasks, process control tasks, and optimized scheduling tasks. These three tasks are clearly assigned responsibilities. By clearly defining the boundaries of each task's responsibilities, the control functions at each level are efficiently executed. Real-time control tasks are primarily responsible for responding to immediately changing data and performing direct feedback control. Process control tasks address system operational stability over the medium to long term and coordinate equipment within the system. Optimal scheduling tasks globally optimize long-term energy usage to ensure optimal allocation of resources.

[0053] Based on these three levels of control function definitions, the system constructs a communication topology for the device layer. By connecting various sensors and actuators to the fieldbus network, the device layer enables data exchange with local control units. Sensors such as gas concentration sensors, flow meters, and thermometers transmit real-time data collected via the fieldbus network to the upper-level control unit, ensuring precise control instructions for each device. This data transmission method enables the device layer to respond to system needs in real time, while ensuring timely data processing through an efficient communication architecture. The control layer design further optimizes the device adjustment process. At the control layer, the system uses an improved fuzzy adaptive PID control algorithm to calculate device adjustment parameters based on the relationship between gas concentration and various device parameters. The PID control algorithm is a commonly used feedback control strategy that calculates the error between the current state and the desired state and adjusts the control variable to gradually approach the target. Fuzzy logic further enhances the system's robustness in the face of uncertainty, building upon the traditional PID algorithm. This control algorithm enables the system to dynamically adjust device parameters based on real-time data. For example, when gas concentration changes, parameters such as the flow rate of the heat exchanger and the combustion intensity of the waste heat boiler will be adjusted based on real-time data to ensure optimal energy recovery.

[0054] Based on real-time control strategies, management designed a scheduling strategy that decomposes long-term optimization objectives into a sequence of short-term control objectives to generate time-phased execution plans. This scheduling strategy, through periodic objective decomposition, ensures flexible adjustments based on actual conditions within each control cycle. For example, if the current gas concentration is high, the short-term goal may focus on reducing the waste heat boiler load and improving the efficiency of the heat exchanger. However, when the gas concentration is low, the goal may be to increase the flow rate of the heat exchanger to maximize heat recovery.

[0055] After the time-segment execution plan is generated, the control instructions are further converted into parameter settings that can be executed by the equipment. This conversion process depends on the specific control objectives and the response characteristics of the equipment. For example, if the system instruction requires increasing the flow rate of a heat exchanger, the corresponding control instruction will be specific to the adjustment values ​​of temperature, pressure, and flow rate. With these converted parameter settings, the equipment can operate within a precise control range, thereby achieving optimal heat recovery. Based on the equipment control instruction set, the system establishes a communication mechanism between the three levels of control systems. The upstream data aggregation process aggregates all sensor data from the equipment layer and transmits it to the control layer for real-time analysis and decision-making. The downstream instruction distribution process sends the specific instructions generated by the control layer to the equipment layer, ensuring that the equipment operates according to the optimized control parameters. Through this close coordination between upstream and downstream, the system achieves precise control and adjustment, ensuring efficient energy recovery and preventing safety hazards.

[0056] The above describes the method for utilizing waste heat from gas based on artificial intelligence in the embodiment of the present application. The following describes the system for utilizing waste heat from gas based on artificial intelligence in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the gas waste heat utilization system based on artificial intelligence includes: Processing module 201 is used to collect and process sensor parameters of the gas extraction system in real time to obtain a standardized data set; Modeling module 202, configured to input the standardized data set into a fused convolutional Transformer neural network for analysis and modeling to obtain a gas waste heat prediction model; An analysis module 203 is used to perform decision analysis on gas waste heat utilization based on the gas waste heat prediction model and a preset expert knowledge base to obtain a control strategy; The control module 204 is used to control the parameters of the heat exchanger, waste heat boiler, circulation pump, and control valve equipment according to the control strategy to obtain a cascade utilization execution plan.

[0057] By integrating these components and employing AI algorithms to dynamically analyze data such as gas concentration, flow rate, and temperature, the system can effectively identify and predict the availability and optimal utilization of waste heat from gas. Through precise model training, the system can adjust waste heat recovery strategies in real time and optimize control strategies based on factors such as gas concentration, flow rate, and temperature. This significantly improves waste heat utilization efficiency and avoids the inefficiency and energy waste associated with traditional empirical control methods. The AI ​​algorithm also dynamically adjusts system operating parameters based on the correlation between real-time and historical data, ensuring that each device (such as heat exchangers, waste heat boilers, and circulating pumps) achieves optimal performance under varying operating conditions. For example, when gas concentration fluctuates, the system automatically adjusts parameters such as heat exchanger flow rate and waste heat boiler load. When gas concentration is high, the system reduces the heat exchanger flow rate and waste heat boiler load to prevent equipment overload and ensure safety. When gas concentration is low, the system increases the heat exchanger flow rate to improve waste heat recovery efficiency. In this scenario, artificial intelligence technology can process large amounts of sensor data in real time and make precise decisions, ensuring maximum gas waste heat recovery without overloading or inefficient equipment. By precisely controlling equipment operation, the risk of gas explosions caused by improper operation is avoided. For example, the system dynamically adjusts control strategies based on a gas concentration prediction model. When gas concentrations exceed safety thresholds, emergency measures are swiftly implemented, effectively mitigating safety risks. Furthermore, the system uses real-time data feedback to adjust and immediately identify and address any anomalies, ensuring the entire recovery process remains within a safe and controllable range.

[0058] above Figure 2 The artificial intelligence-based gas waste heat utilization system in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The artificial intelligence-based gas waste heat utilization equipment in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0059] Figure 3This is a schematic diagram of the structure of an artificial intelligence-based waste gas heat utilization device provided by an embodiment of the present invention. This artificial intelligence-based waste gas heat utilization device 300 may vary significantly depending on its configuration or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors), memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) storing application programs 333 or data 332. The memory 320 and storage media 330 may be either transient or persistent storage. The program stored in the storage medium 330 may include one or more modules (not shown), each of which may include a series of instructions for operating on the artificial intelligence-based waste gas heat utilization device 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, allowing the artificial intelligence-based waste gas heat utilization device 300 to execute the series of instructions stored in the storage medium 330 to implement the steps of the artificial intelligence-based waste gas heat utilization method described above.

[0060] The artificial intelligence-based waste heat utilization device 300 may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. It will be understood by those skilled in the art that Figure 3 The structure of the artificial intelligence-based gas waste heat utilization equipment shown does not constitute a limitation on the artificial intelligence-based gas waste heat utilization equipment provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0061] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the artificial intelligence-based gas waste heat utilization method.

[0062] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0063] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion 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 for enabling an artificial intelligence-based gas waste heat utilization device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0064] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for utilizing waste heat from gas based on artificial intelligence, characterized in that: The method comprises: Real-time collection and processing of sensor parameters of the gas extraction system to obtain a standardized data set; Inputting the standardized data set into the fused convolutional Transformer neural network for analysis and modeling to obtain a gas waste heat prediction model; Performing decision analysis on gas waste heat utilization based on the gas waste heat prediction model and a preset expert knowledge base to obtain a control strategy; According to the control strategy, the parameters of the heat exchanger, waste heat boiler, circulation pump and control valve equipment are adjusted and controlled to obtain a cascade utilization execution plan.

2. The method for utilizing waste heat from gas based on artificial intelligence according to claim 1, characterized in that: The real-time collection and processing of sensor parameters of the gas drainage system to obtain a standardized data set includes: Parameters are collected through sensor devices at key nodes in the gas extraction system to obtain raw data on gas concentration, flow, temperature, pressure, and oxygen content; The original data is subjected to a sliding window median filter to remove interference noise in a coal mine gas extraction environment to obtain preliminary filtered data; performing data smoothing on the preliminary filtered data to obtain smoothed data; The smoothed data is subjected to the 3σ principle to identify abnormal values ​​caused by sudden gas outburst and filled with the data by linear interpolation method to obtain a data sequence; The data sequence is normalized using the Z-Score normalization method to obtain normalization parameters.

3. The method for utilizing waste heat from gas based on artificial intelligence according to claim 1, characterized in that: The standardized data set is input into the fused convolutional Transformer neural network for analysis and modeling to obtain a gas waste heat prediction model, including: The standardized dataset is input into a convolutional neural network with three convolutional layers for processing. Each convolution layer uses 32, 64, and 128 convolution kernels of 3×3 size, respectively, to obtain the spatial characteristics of the gas concentration distribution. The spatial features are subjected to dimensionality reduction by performing a batch normalization layer and a 2×2 maximum pooling operation with a stride of 2 after each convolution layer to obtain a reduced-dimensional spatial feature representation; Input the reduced-dimensional spatial feature representation into a multi-head self-attention mechanism with 8 heads, with the hidden layer dimension set to 256 and the feedforward network dimension set to 1024, to obtain attention-weighted features; A 2×2 maximum pooling operation with a step size of 2 is applied to the attention weighted features to reduce memory consumption and obtain a simplified attention feature; The simplified attention features are input into a two-layer bidirectional LSTM structure, where the first layer contains 128 hidden units and the second layer contains 64 hidden units, to obtain the temporal variation features of gas flow and temperature; The gas waste heat prediction model is obtained by training a learnable attention network to calculate the weight coefficients of spatial features, attention features and temporal features, constructing a loss function for optimization training.

4. The method for utilizing waste heat from gas based on artificial intelligence according to claim 3, characterized in that: The decision analysis of gas waste heat utilization based on the gas waste heat prediction model and the preset expert knowledge base to obtain a control strategy includes: Obtaining a fusion prediction result of spatial features, attention features, and temporal features from the gas waste heat prediction model; determining a high-concentration interval when the predicted gas concentration is higher than the safe explosion upper limit; determining a medium-concentration interval when the predicted gas concentration is between the safe explosion lower limit and the upper limit; and determining a low-concentration interval when the predicted gas concentration is lower than the safe explosion lower limit, thereby obtaining gas waste heat grade classification data; According to the gas waste heat grade classification data, an expert knowledge base organized based on an ontological structure is queried. When the gas is in a high concentration range, a high calorific value utilization rule is called; when the gas is in a medium concentration range, a medium calorific value utilization rule is called; when the gas is in a low concentration range, a low calorific value utilization rule is called to obtain a preliminary control reference value. A decision scoring function is constructed for the preliminary control reference value, and a weighted comprehensive evaluation index is obtained by calculating the ratio of energy recovery efficiency to theoretical maximum value, the margin between equipment operating parameters and safety thresholds, the difference between the economic value of recovered heat and energy consumption cost, and reduced carbon emissions; The weighted comprehensive evaluation index is input into the hierarchical multi-agent reinforcement learning framework as a reward signal. The main agent decomposes the tasks and assigns them to sub-agents. The sub-agents are responsible for optimizing the heat exchange system, boiler system, and circulation system, and a collaboratively optimized state-action strategy is obtained. Based on the collaborative optimization state-action strategy, multiple sets of candidate control parameters are generated. When the ambient temperature in the coal mine is higher than the upper comfortable limit, the ventilation and cooling solution is given priority. When the ambient temperature is lower than the lower comfortable limit, the heating solution is given priority. When the ambient temperature is between the lower and upper comfortable limits, the power generation and heating solutions are balanced, thereby obtaining a multi-objective trade-off control parameter set. The multi-objective trade-off control parameter set is subjected to evolutionary multi-objective optimization based on the super volume index, and a reference point adaptive adjustment mechanism and a decision space diversity maintenance strategy are introduced. The control values ​​of the heat exchanger, waste heat boiler, circulating pump, and control valve are determined according to the gas extraction working conditions to obtain the control strategy.

5. The method for utilizing waste heat from gas based on artificial intelligence according to claim 4, characterized in that: The data of the gas waste heat level classification is queried based on the expert knowledge base organized based on the ontological structure, and when in a high concentration range, a high calorific value utilization rule is called; when in a medium concentration range, a medium calorific value utilization rule is called; when in a low concentration range, a low calorific value utilization rule is called, to obtain a preliminary control reference value, including: The experience in handling different gas concentration ranges in the actual coal mine production environment is structured and organized. By establishing the correlation between equipment control rules, safety and explosion prevention rules, energy efficiency optimization rules, and emergency handling rules, a knowledge base for gas waste heat utilization is obtained. The gas waste heat classification data is input into the knowledge retrieval system, and when the underground gas concentration fluctuates, similar historical working conditions are quickly matched to obtain a treatment plan; Conduct a safety review of the treatment plan, identify and eliminate dangerous operations in the plan based on mine safety production standards, and obtain operating instructions that comply with coal mine safety regulations; The operation instructions are converted into control parameters of the heat exchange system, the waste heat boiler system, and the circulation system. When the gas concentration suddenly increases, the combustion intensity of the waste heat boiler is reduced. When the gas concentration gradually decreases, the heat exchanger flow rate is increased to obtain the operating parameters of each system; Based on the operating parameters of each system, a cascade utilization process for coal mine gas waste heat is constructed. When the gas grade is high, the power generation system is prioritized; when the gas grade is medium, industrial heat is prioritized; when the gas grade is low, domestic hot water and heating systems are prioritized, thereby obtaining a cascade utilization plan. The cascade utilization scheme is adapted to the actual situation of the coal mine, and the theoretical parameters are actually corrected by considering the mine production plan, seasonal heat demand and equipment maintenance cycle to obtain the preliminary control reference value.

6. The method for utilizing waste heat from gas based on artificial intelligence according to claim 1, characterized in that: The control strategy is used to adjust parameters of the heat exchanger, waste heat boiler, circulation pump, and control valve equipment to obtain a cascade utilization execution plan, including: The control strategy is converted into a hierarchical control instruction sequence, and the adjustment parameters of each device are obtained by establishing a three-level control architecture of device layer, control layer and management layer; Conducting coal mine safety production constraint checks on the adjustment parameters, and obtaining safety control instructions that meet explosion-proof requirements by comparing the parameter values ​​with safety limits and applying safety margins; Based on the safety control instructions, a heat exchanger parameter control scheme is constructed to control the heat exchange efficiency by adjusting the inlet and outlet temperature difference and flow parameters. When the gas concentration fluctuates, the heat exchange rate is dynamically adjusted to obtain the heat exchanger operation curve; Formulate a waste heat boiler parameter control plan based on the heat exchanger operation curve, and obtain the boiler system operation curve by coordinating three key parameters: boiler load, feed water flow and steam pressure; Designing a circulation system control scheme based on the heat exchanger operating curve and the boiler system operating curve, and ensuring quantitative fluid delivery by adjusting the circulation pump speed and the pipe network pressure to obtain the circulation system operating curve; The heat exchanger operating curve, boiler system operating curve and circulation system operating curve are systematically integrated and coordinated, and the cascade utilization implementation plan is obtained according to the principle of using high-grade thermal energy for power generation, medium-grade thermal energy for industrial heat, and low-grade thermal energy for heating.

7. The method for utilizing waste heat from gas based on artificial intelligence according to claim 6, characterized in that: The control strategy is converted into a hierarchical control instruction sequence, and the adjustment parameters of each device are obtained by establishing a three-level control architecture of device layer, control layer and management layer, including: Decomposing the control strategy into control levels, and obtaining a three-level control function definition by determining the responsibility boundaries of real-time control tasks, process control tasks, and optimization scheduling tasks; Establishing a device layer communication topology based on the three-level control function definition, and obtaining a device layer data interaction framework by connecting sensors and actuators to the local control unit through a fieldbus network; A control layer control algorithm is constructed based on the device layer data interaction framework, and a real-time control strategy is obtained by converting the relationship between gas concentration and device parameters into improved fuzzy adaptive PID control parameters; Designing a management scheduling strategy based on the real-time control strategy, and obtaining a time-segmented execution plan by decomposing the long-term optimization goal into a sequence of short-term control goals; Performing control instruction conversion on the time-segment execution plan, and obtaining a device control instruction set by converting an abstract control strategy into parameter setting values ​​executable by a specific device; A three-level control system communication mechanism is established based on the device control instruction set, and the adjustment parameters of each device are obtained by defining an uplink data aggregation process and a downlink instruction distribution process.

8. A gas waste heat utilization system based on artificial intelligence, characterized in that: For implementing the artificial intelligence-based gas waste heat utilization method according to any one of claims 1 to 7, the artificial intelligence-based gas waste heat utilization system comprises: The processing module is used to collect and process the sensor parameters of the gas extraction system in real time to obtain a standardized data set; A modeling module is used to input the standardized data set into a fused convolutional Transformer neural network for analysis and modeling to obtain a gas waste heat prediction model; An analysis module, configured to perform decision analysis on gas waste heat utilization based on the gas waste heat prediction model and a preset expert knowledge base to obtain a control strategy; The control module is used to control the parameters of the heat exchanger, waste heat boiler, circulation pump, and control valve equipment according to the control strategy to obtain a cascade utilization execution plan.

9. A gas waste heat utilization device based on artificial intelligence, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the gas waste heat utilization method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor is enabled to execute the method for utilizing waste heat from gas based on artificial intelligence according to any one of claims 1 to 7.

Images