Mine intelligent ventilation combined optimal adjustment and optimal control method and system

Abstract

The invention discloses a mine intelligent ventilation combined optimal adjustment and optimal control method and system. The method comprises the steps that S1, multi-source mine ventilation data are collected through a distributed multi-mode sonic sensor by adopting a gridding deployment strategy; s2, performing fusion optimization on the multi-source mine ventilation data by using a trust function theory to obtain comprehensive state data; s3, on the basis of the comprehensive state data, obtaining optimal preliminary regulation and control actions of different mine equipment through a self-adaptive reinforcement learning model; and S4, based on the optimal preliminary regulation and control action, introducing a roadway structure safety degree as a dynamic constraint factor, combining with an improved pigeon inspired optimization algorithm to obtain a final regulation and control action, and realizing underground ventilation control based on the final regulation and control action. The safety degree of the roadway structure is incorporated into a fitness function, so that the safety of the roadway structure is fully considered while the ventilation effect and the energy consumption are optimized by the algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of mine control, and particularly to an intelligent ventilation combined optimization and control method and system for mines. Background Art

[0002] In mine exploitation operations, the ventilation system plays a crucial role in ensuring underground safety production. Mine ventilation not only needs to ensure that the air quality underground meets the standards, effectively dilute and discharge harmful gases such as gas, but also maintain suitable operating environment temperature and humidity. With the increase in the depth and breadth of mine exploitation, the ventilation system faces more challenges. On the one hand, the underground ventilation network structure is becoming increasingly complex, the ventilation requirements in different regions vary significantly, and it changes continuously under the influence of factors such as geological conditions and mining progress. On the other hand, the operating state of ventilation equipment also changes over time, and equipment failures may cause ventilation abnormalities. If not detected and processed in a timely and accurate manner, it is easy to trigger safety accidents.

[0003] At present, the existing ventilation control methods have deficiencies in ensuring the safety of roadway structures. The changes in wind pressure and wind speed generated by ventilation may affect the stability of roadways, threatening the safety of underground workers and equipment. Moreover, the existing mine ventilation control technologies are difficult to comprehensively consider multiple factors such as ventilation effect, energy consumption, and roadway structure safety. Therefore, an intelligent ventilation combined optimization and control method and system for mines are proposed herein. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art and achieve the above object, the present invention proposes the following technical solutions: An intelligent ventilation combined optimization and control method and system for mines, including: An intelligent ventilation combined optimization and control method and system for mines, the method steps are as follows: S1: Collect multi-source mine ventilation data through a grid deployment strategy using distributed multi-modal acoustic sensors; S2: Use the theory of belief functions to fuse and optimize the multi-source mine ventilation data to obtain comprehensive state data; S3: Based on the comprehensive state data, obtain the optimal preliminary control actions for different mine equipment through an adaptive reinforcement learning model; S4: Based on the optimal preliminary control actions, introduce the roadway structure safety factor as a dynamic constraint factor and combine it with an improved pigeon flock optimization algorithm to obtain the final control actions, and realize underground ventilation control based on the final control actions; Among them, the improved pigeon flock optimization algorithm is realized by using a chaotic map to initialize the pigeon flock positions in the pigeon flock optimization algorithm, introducing a diversity feedback mechanism during the update process of the pigeon flock positions, and considering the roadway structure safety factor.

[0005] The process of collecting multi-source mine ventilation data is as follows: Deploy multi-modal acoustic wave sensors, with the adjacent sensor nodes spaced apart in the horizontal direction and apart in the vertical direction to form a spatial grid. Each sensor node is equipped with a low-power acoustic wave transmitting and receiving module, and each sensor node collects the acoustic wave signals in the mine according to the set sampling frequency The collected acoustic wave signals are discrete signal sequences, and the discrete signal sequences include wind speed data, roadway structure acoustic signals, and equipment rotation speed signals.

[0006] The process of using the theory of belief functions to fuse and optimize multi-source mine ventilation data is as follows: Define an identification framework, which includes "ventilation equipment normal" and "ventilation equipment failure"; Extract the features of the multi-source mine ventilation data collected from the distributed acoustic wave sensor array. Convert the time-domain signal to the frequency-domain signal through the fast Fourier transform, extract the spectral features of the signal, and construct an evidence body according to the extracted spectral features. The evidence body includes ventilation equipment normal and ventilation equipment failure; According to the spectral features, compare with the pre-set fault feature threshold to determine the basic probability assignment function. For the spectral feature data of all sensor nodes, each node constructs an evidence body based on its own collected data, and uses the Dempster-Shafer combination rule to obtain the belief assignment data; Based on the belief assignment data, in the order of the propositions in the proposition set, arrange the belief assignment data of each proposition in sequence to form a data sequence, and this data sequence is the comprehensive status data.

[0007] The process of using the Dempster-Shafer combination rule to obtain the belief assignment data is as follows: Based on the evidence bodies of any two sensor nodes, fuse the evidence bodies to obtain the fused basic probability assignment. Traverse the evidence bodies of all sensor nodes for Dempster-Shafer combination rule fusion to obtain the corresponding basic probability assignment, and combine the basic probability assignments of all sensor nodes to obtain the belief assignment data.

[0008] The adaptive reinforcement learning model uses a deep neural network as the basic architecture, introduces the local perception and weight sharing ideas of the convolutional neural network in the hidden layer, and updates the strategy based on the double deep Q network algorithm to adapt to the complex non-linear dynamic characteristics of the mine ventilation system; The adaptive reinforcement learning model includes an input layer, multiple hidden layers, and an output layer.

[0009] The process of obtaining the preliminary regulation actions is as follows: The input layer of the adaptive reinforcement learning model receives comprehensive state data, which is converted into a comprehensive vector representing dimensions and then transmitted to the hidden layer; The hidden layer consists of multiple neurons, which are connected by weights. The activation function is used to perform a non-linear transformation on the output of the neurons, and the ideas of local perception and weight sharing in the convolutional neural network are introduced. Convolutional modules are set at intervals to optimize the convolution operation on the local area of the data; The output layer outputs the regulation action scores of different devices. The number of neurons corresponds to the number of device actions to be regulated, and the probabilities of the regulation actions of different devices are output; The policy is updated through the double deep Q-network algorithm. The double deep Q-network includes a current network and a target network. The Q function is defined, and the Q function includes the parameters of the deep neural network. The current network is used to select actions, and the target network is used to evaluate the action value. The parameters of the current network are updated by minimizing the loss function and the experience replay mechanism is adopted. After multiple rounds of training, the parameters of the current network are updated. When the training converges, the optimal preliminary regulation strategy under different comprehensive states is obtained.

[0010] The process of obtaining the roadway structure safety degree is as follows: Safety sensors are used to monitor the deformation and stress parameters of the roadway. The safety sensors include displacement sensors and stress sensors; The deformation and stress parameters are combined with the geomechanics model to obtain the roadway structure safety index, and the roadway structure safety degree is quantified through an empirical model based on the structure safety index.

[0011] The process of obtaining the final regulation action by introducing the roadway structure safety degree as a dynamic constraint factor and combining it with the improved pigeon flock optimization algorithm is as follows: The pigeon flock optimization algorithm is introduced, and the positions of the pigeon flock are initialized using chaotic mapping Let the size of the pigeon flock be , and each pigeon represents a regulation action combination. The vector dimension is the same as the number of parameters of the regulation action, and the positions of the initial pigeon flock within the parameter value range are randomly generated; The advantages and disadvantages of each regulation action combination are evaluated through a fitness function. For each regulation action combination, its fitness value is calculated according to the fitness function, and the position where the regulation action combination with the highest fitness is located is found, which is the global optimal position; The flying direction of the pigeon flock is adjusted according to the global optimal position, and the flying speed is obtained, and the position is updated according to the flying speed; A diversity feedback mechanism is introduced during the position update process. According to the correction amount determined by the roadway structure safety degree and the regulation action adjustment rules, when the position of the pigeon causes the roadway structure safety degree to be lower than the preset safety threshold, its position is corrected through the correction amount; Repeatedly calculate the fitness function, update the positions and velocities of the pigeon flock until the iterative termination condition is met, i.e., the fitness function converges or the maximum number of iterations is reached. The finally obtained global optimal position is the global optimal position. The corresponding combination of control actions is the final control action.

[0012] The process of determining the correction amount according to the roadway structure safety degree and the control action adjustment rule is as follows: Based on the roadway structure safety threshold and the actual roadway structure safety degree corresponding to the pigeon position, obtain the roadway structure safety degree index; Construct a control action adjustment rule. The control action adjustment rule includes the rotational speed of the ventilator and the opening degree of the ventilation valve. Set the correction amount of the rotational speed of the ventilator as and the correction amount of the opening degree of the ventilation valve as. The correction amount of the rotational speed of the ventilator is obtained based on the current combination of control actions considering the minimum step size of the rotational speed adjustment of the ventilator, and the correction amount of the opening degree of the ventilation valve is obtained based on the current combination of control actions considering the minimum step size of the opening degree adjustment of the ventilation valve: Determine the final correction amount based on the correction amount of the rotational speed of the ventilator and the correction amount of the opening degree of the ventilation valve.

[0013] The implementation process of the diversity feedback mechanism is as follows: Calculate the diversity index of the pigeon flock. Use the Euclidean distance to measure the average distance between pigeons. When the diversity index is lower than the preset threshold, randomly perturb the positions of the pigeons; The diversity index is obtained by calculating the Euclidean distance between any two pigeons, then adding up all the Euclidean distances between pairs of pigeons and dividing by the total number of pigeon pairs.

[0014] An intelligent ventilation joint optimization and control system for mines, comprising: Data acquisition module: Collect multi-source mine ventilation data through distributed multi-modal acoustic wave sensors using a grid deployment strategy; Fusion optimization module: Use the theory of belief functions to fuse and optimize multi-source mine ventilation data to obtain comprehensive status data; Reinforcement learning module: Based on the comprehensive status data, obtain the optimal preliminary control actions for different mine equipment through an adaptive reinforcement learning model; Control optimization module: Based on the optimal preliminary control actions, introduce the roadway structure safety degree as a dynamic constraint factor and combine it with an improved pigeon flock optimization algorithm to obtain the final control actions, and realize underground ventilation control based on the final control actions.

[0015] The present invention has the following beneficial effects: In the present invention, first, by extracting features from multi-source mine ventilation data, using the fast Fourier transform to convert the time-domain signal into a frequency-domain signal and extract spectral features, and constructing an evidence body to fuse data using the Dempster-Shafer synthesis rule, the information of multiple sensor nodes can be integrated, reducing data uncertainty and error. For example, when abnormal data appears in some sensors due to environmental interference, by fusing the evidence bodies of other sensors, the interference can be effectively excluded, and a more reliable trust degree distribution data of the normal or faulty ventilation equipment can be obtained; Secondly, an adaptive reinforcement learning model based on a deep neural network and introducing the ideas of local perception and weight sharing of the convolutional neural network is used, combined with the double deep Q-network algorithm for policy update, so that the model can better adapt to the complex non-linear dynamic characteristics of the mine ventilation system, and can learn in real time according to the comprehensive state data and generate optimal preliminary control actions for different mine equipment; Finally, by introducing the roadway structure safety factor as a dynamic constraint factor and combining with an improved pigeon flock optimization algorithm to obtain the final control action, the improved pigeon flock optimization algorithm initializes the positions of the pigeon flock using chaotic mapping, increasing the diversity and ergodicity of the initial population and improving the search ability of the algorithm. During the position update process, a diversity feedback mechanism is introduced to avoid the algorithm falling into a local optimal solution and ensure that a better combination of control actions can be searched. At the same time, by incorporating the roadway structure safety factor into the fitness function, while optimizing the ventilation effect and energy consumption, the safety of the roadway structure is fully considered. When the position of the pigeon (combination of control actions) causes the roadway structure safety factor to be lower than the preset threshold, the position is adjusted through a reasonable correction amount, ensuring the safety of underground workers and equipment. Brief Description of the Drawings

[0016] Figure 1 It is a method step diagram of a mine intelligent ventilation joint optimization and control method and system proposed by the present invention.

[0017] Figure 2 It is a system block diagram of a mine intelligent ventilation joint optimization and control method and system proposed by the present invention. Detailed Embodiments

[0018] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0019] Embodiment 1: As Figure 1 shown, a mine intelligent ventilation joint optimization and control method proposed by the present invention, the method steps include: S1: Collect multi-source mine ventilation data through a grid deployment strategy using distributed multi-modal acoustic wave sensors; In the mine roadway, taking roadway intersections, bends, and working faces as core nodes, construct a three-dimensional stereo perception network. Deploy multi-modal acoustic wave sensors in the three-dimensional stereo perception network. The multi-modal acoustic wave sensors include The adjacent sensor nodes are spaced in the horizontal direction and in the vertical direction, forming a regular spatial grid. Each sensor node is equipped with a low-power acoustic wave transmitting and receiving module. Each sensor node samples the acoustic wave signals in the mine at a set sampling frequency ; Specifically, the collected acoustic wave signals are discrete signal sequences, and the discrete signal sequences include wind speed data, roadway structure acoustic signals, and equipment rotation speed signals; Wind speed data: The wind speed signals collected by the multi-modal acoustic wave sensors can be used to analyze the wind speed information at different positions in the mine. For example, at roadway intersections, bends, etc., the wind speed will change due to the change of the roadway structure. Accurately obtaining the wind speed data can determine whether the ventilation is smooth in subsequent decision-making. If the wind speed is too low, it may lead to risks such as gas accumulation; Roadway structure acoustic signals: When acoustic waves propagate in the roadway, reflection, refraction, etc. will occur due to the roadway structure (such as intersections, bends). The signals collected by the acoustic sensors contain roadway structure information; Equipment rotation speed signals: The change of the equipment rotation speed will change the sound characteristics emitted, and the signals collected by the acoustic sensors can indirectly reflect the rotation speed situation.

[0020] S2: Use the theory of belief functions to fuse and optimize the multi-source mine ventilation data to obtain comprehensive state data; The theory of belief functions is a mathematical theory for dealing with uncertain information. Applying it to mine ventilation decision-making, first define a frame of discernment , and the frame of discernment is a set composed of all possible propositions. For the mine ventilation system, the frame of discernment includes, for example, "ventilation equipment is normal" and "ventilation equipment fails"; Extract the features of the multi-source mine ventilation data collected from the distributed acoustic wave sensor array. Convert the time-domain signal to the frequency-domain signal through the fast Fourier transform (FFT); Extract the spectral features of the signal. Let the extracted spectral features be , where For the characteristic dimension, an evidence body is constructed based on the extracted spectral features. For different propositions, the evidence body includes: Indicates that "the ventilation equipment is normal", Indicates that "the ventilation equipment fails"; According to the spectral features , compare with the pre-set fault feature threshold to determine the basic probability assignment function , for example: if the feature (the amplitude of a certain frequency component of the equipment running sound) exceeds the pre-set normal range threshold , then assign a higher degree of trust to the proposition of "ventilation equipment failure", and at the same time ; For the spectral feature data of all sensor nodes, each node constructs an evidence body based on its own collected data, and uses the Dempster-Shafer combination rule to obtain the trust degree assignment data; The process of obtaining the trust degree assignment data is as follows: Adopt a fusion algorithm (Dempster-Shafer combination rule), based on the evidence bodies of any two sensor nodes, fuse the evidence bodies to obtain the fused basic probability assignment , traverse the evidence bodies of all sensor nodes for Dempster-Shafer combination rule fusion to obtain the corresponding basic probability assignment , Using it as an index, combine the basic probability assignments of all sensor nodes together, and finally obtain the trust degree assignment data on the comprehensive state of the data of each sensor node in the mine ventilation; Specifically, in the mine ventilation scenario, multiple sensor nodes will obtain data from different positions and angles and construct evidence bodies. Each evidence body represents a local judgment on a certain aspect of the mine ventilation system. The purpose of the Dempster-Shafer combination rule is to integrate these scattered and local evidence bodies to obtain a judgment on the state of the mine ventilation system that comprehensively reflects all evidence information; Based on the trust degree assignment data, in the order of the propositions in the proposition set, arrange the trust degree assignment data (the trust degree values after quantization or standardization processing) of each proposition in sequence to form a data sequence, denoted as , where is the data index, and each data point in the data sequence represents the trust degree of the state of different sensor nodes, comprehensively reflecting the state possibilities of the data collected by the sensor nodes in the mine ventilation system in various aspects. This data sequence is the comprehensive state data; Specifically, the obtained trust assignment data after fusion form a data sequence , and this data sequence arranges the trust values in the order of propositions. Each value synthesizes the information of all sensor nodes regarding the corresponding proposition. In the data sequence, the trust value corresponding to the "ventilation equipment failure" proposition of a certain sensor node is obtained by fusing the evidence bodies of each node. It is no longer the judgment of a local node, but an overall assessment of the possibility of ventilation equipment failure after integrating the information of all nodes. Through such a data sequence form, the comprehensive data status of the mine ventilation system is presented intuitively and accurately.

[0021] S3: Based on the comprehensive status data, obtain the optimal preliminary control actions for different mine equipment through an adaptive reinforcement learning model; The adaptive reinforcement learning model uses a deep neural network (DNN) as the basic architecture, and introduces the local perception and weight sharing ideas of a convolutional neural network (CNN) in the hidden layer. Based on the Double Deep Q-Network (DDQN) algorithm, the policy is updated to adapt to the complex non-linear dynamic characteristics of the mine ventilation system. The adaptive reinforcement learning model includes an input layer, multiple hidden layers, and an output layer; The input layer of the adaptive reinforcement learning model receives the comprehensive status data. Let the dimension of the comprehensive status data be , and it is converted into a comprehensive vector representing the dimension, that is . The number of neurons in the input layer is the same as the dimension of the comprehensive status data, and the comprehensive status data is directly passed to the next hidden layer; The hidden layer consists of multiple neurons, and the neurons are connected by weights. The activation function is used to perform non-linear transformation on the output of the neurons. Let the input of the neurons in the -th hidden layer be , the output be , the weight matrix be , and the bias vector be , then:

[0022] Among them, is the activation function. Through the non-linear transformation of multiple hidden layers, the complex features in the comprehensive status data are gradually extracted; The specific process of introducing the local perception and weight sharing ideas of a convolutional neural network (CNN) in the hidden layer is as follows: In the hidden layer, convolutional modules are set at intervals to optimize the convolutional operation on the local area of the data, which is specifically expressed as: For one-dimensional comprehensive status data, a one-dimensional convolutional kernel is used, and the size is The convolution kernel slides on the data to perform convolution operations, that is , where is the convolution kernel weight, is the bias vector, is the output after convolution optimization; Specifically, through this optimization method of local perception and weight sharing ideas, the model's ability to capture local key information in the data is enhanced, while the number of parameters is reduced and the training efficiency is improved; The output layer outputs the regulation action scores of different devices. The number of neurons corresponds to the number of device actions to be regulated. Suppose there are devices to be regulated in the mine ventilation system, and each device has possible regulation actions. The number of neurons in the output layer is , and the output layer uses a linear activation function. The output is a vector representing the probabilities of the regulation actions of different devices, that is ; The process of policy update through the double deep Q-network algorithm is as follows: The adaptive reinforcement learning model interacts with the mine ventilation system environment. At each time step, the model receives the current comprehensive state data as input and outputs a regulation action vector . According to the regulation action vector, corresponding device regulation operations are performed in the mine ventilation system, and then the changes in the system state are observed to obtain new comprehensive state data and the corresponding reward value ; The reward value is defined according to the changes in the system state. For example, if the energy consumption of the ventilation system decreases and the air quality improves, a positive reward is given. If there are abnormalities in the ventilation equipment or the air quality deteriorates, a negative reward is given; Policy update is performed using the Double Deep Q-Network (DDQN). The double deep Q-network solves the overestimation problem that may occur in DQL on the basis of Deep Q-Learning (DQL), and defines the Q function , where are the parameters of the deep neural network, representing the cumulative reward expected to be obtained by performing an action in the state. The double deep Q-network includes a current network and a target network; The current network is used to select actions, and the target network is used to evaluate the action value. The parameters of the current network are updated by minimizing the loss function , and the loss function is defined as:

[0023] where , is the reward value,​ is the discount factor, is the next state, is the current network parameter, is the target network parameter, Adopt an experience replay mechanism. After multiple rounds of training, continuously extract samples from the experience replay buffer for training and update the current network parameters. When the training converges, the current network at this time has learned the optimal preliminary control strategies under different comprehensive states, denoted as , where is the number of indices of the control strategy; Example: Suppose there are 3 main devices to be controlled in a mine ventilation scenario: the main ventilator, the local ventilator, and the ventilation valve. The main ventilator has 3 control actions (increase the rotation speed, maintain the rotation speed, decrease the rotation speed), the local ventilator has 2 control actions (turn on, turn off), and the ventilation valve has 2 control actions (open wider, close smaller). Through the adaptive reinforcement learning model, it receives comprehensive state data in real time and outputs the optimal control result. For example, when the input comprehensive state data shows that the gas concentration increases and the temperature of equipment 1 is too high, the optimal control actions output by the current network are that the main ventilator maintains the rotation speed, the local ventilator is turned on, and the ventilation valve is opened wider. This set of actions can not only effectively reduce the gas concentration but also prevent the main ventilator from being damaged due to excessive acceleration, realizing the optimal control of the mine ventilation system in this state. In actual operation, as long as the real-time comprehensive state data of the mine ventilation system (including equipment status, air quality indicators, etc.) is input into this converged current network, the action combination output by the network will be executed as the optimal control action, thus realizing the intelligent control of the mine ventilation system; S4: Based on the optimal preliminary control actions, introduce the roadway structure safety factor as a dynamic constraint factor and combine it with the improved pigeon flock optimization algorithm to obtain the final control actions, and realize the underground ventilation control based on the final control actions; The process of obtaining the roadway structure safety factor is as follows: Use safety sensors (such as displacement sensors, stress sensors, etc.) to monitor the deformation and stress parameters of the roadway, combine the deformation and stress parameters and calculate through the geomechanics model to obtain the roadway structure safety index. The structure safety index includes the roadway wall displacement and the stress size index of the support structure. According to the roadway wall displacement and the stress size index in the structure safety index, quantify through the empirical model to obtain the quantified roadway structure safety value, that is, the roadway structure safety factor , and set a roadway structure safety threshold ; Specifically, perform normalization preprocessing on the initially collected safety sensor data, input the preprocessed displacement and stress data into a geomechanics model, and according to the calculation principle of the model, the obtained The value ranges from 0 to 1, and the higher the value, the safer the roadway structure is; The process of introducing the safety degree of the roadway structure as a dynamic constraint factor and combining it with the improved pigeon flock optimization algorithm to obtain the final regulation action is as follows: Introduce a pigeon flock optimization algorithm, use chaotic mapping to initialize the positions of the pigeon flock, increase the diversity and ergodicity of the initial population, and the chaotic mapping position formula is expressed as: , where, is the control parameter, is the previous position; Assume that the size of the pigeon flock in the pigeon flock optimization algorithm is , each pigeon represents a regulation action combination, the vector dimension is the same as the number of parameters of the regulation action, and the positions of the initial pigeon flock within the parameter value range are randomly generated based on chaotic mapping; Through a fitness function to evaluate the quality of each pigeon position (i.e., the regulation action combination), is the preliminary regulation strategy; The fitness function comprehensively considers the ventilation effect (the degree of reduction in gas concentration, the improvement of air quality), energy consumption, and the safety degree of the roadway structure. Assume that the target amount of gas concentration reduction is , the actual reduction amount is , the air quality improvement index is , the energy consumption is , then the fitness function can be expressed as:

[0024] where, is the safety degree of the roadway structure. This fitness function combines the safety degree factor of the roadway structure, takes the safety degree of the roadway structure as a dynamic constraint factor, and further increases the diversity considered by the fitness function; For each pigeon (i.e., the regulation action combination), calculate its fitness value according to the fitness function, and find the position where the regulation action combination with the highest fitness is located, which is the global optimal position ; According to the global optimal position, adjust the flight direction of the pigeon flock, obtain the flight speed and update the position according to the flight speed. Assume that the current position of the th pigeon is , the speed is , the global optimal position is , and the flight speed acquisition formula is:

[0025] where, is the learning factor; The position update formula is: , ensure that the updated position is within the range of parameter values; Introduce a diversity feedback mechanism during the position update process, specifically: Calculate the diversity index of the pigeon flock , use the Euclidean distance to measure the average distance between pigeons. When the diversity index is lower than the preset threshold, it indicates that the pigeon flock is too concentrated and prone to falling into a local optimum. At this time, randomly perturb the positions of the pigeons to achieve diversity feedback. The formula is expressed as:

[0026] Where is the perturbation coefficient, is a random number from the standard normal distribution, is the new current position; Among them, the diversity index is obtained by calculating the Euclidean distance between any two pigeons, then adding up all the Euclidean distances between pairs of pigeons, and dividing by the total number of pigeon pairs; Specifically, when the diversity index is relatively high, it means that the average distance between pigeons in the pigeon flock is large, the difference in the combination of control actions is large, the pigeon flock is widely distributed in the search space, and it can explore different regions more comprehensively, reducing the risk of falling into a local optimal solution. When the diversity index is lower than the preset threshold, it indicates that the average distance between pigeons is small, the pigeon flock is too concentrated, and they are all gathered in a certain local area of the search space, and it is very easy to fall into a local optimal solution. At this time, it is necessary to increase the diversity through random perturbation so that the algorithm can continue to effectively search for the global optimal solution; At the same time, consider the safety degree of the roadway structure to update the positions and speeds of the pigeon flock: Introduce a correction amount determined according to the safety degree of the roadway structure and the adjustment rules of control actions. When the position of the pigeon causes the safety degree of the roadway structure to be lower than the preset threshold , correct its position through the correction amount , where is the correction amount determined according to the safety degree of the roadway structure and the adjustment rules of control actions; The process of determining the correction amount according to the safety degree of the roadway structure and the adjustment rules of control actions is: Based on the safety threshold of the roadway structure , and the position of the pigeon , obtain the safety degree index of the roadway structure corresponding to the actual safety degree of the roadway structure at the position of the pigeon Define the safety degree index of the roadway structure :

[0027] Among them, the value reflects the relative degree to which the current control action combination (corresponding to the pigeon position ) deviates the safety degree of the roadway structure from the target threshold. The larger it is, the more serious the deviation. Construct a control action adjustment rule, and the control action adjustment rule includes the rotational speed of the ventilation fan in the control action and the opening degree of the ventilation valve , that is , based on the control action (the rotational speed of the ventilation fan and the opening degree of the ventilation valve ), set the correction amount of the rotational speed of the ventilation fan to be and the correction amount of the opening degree of the ventilation valve to be . The correction amount of the rotational speed of the ventilation fan is obtained based on the current control action combination considering the minimum step size of the rotational speed adjustment of the ventilation fan, and the correction amount of the opening degree of the ventilation valve is obtained based on the current control action combination considering the minimum step size of the opening degree adjustment of the ventilation valve; According to the roadway structure safety degree index and the control action adjustment rule, calculate the correction amount , that is: Obtain the correction amount of the rotational speed of the ventilation fan :

[0028] Among them, is the minimum step size of the rotational speed adjustment of the ventilation fan; Obtain the correction amount of the opening degree of the ventilation valve :

[0029] Among them, is the minimum step size of the opening degree adjustment of the ventilation valve; Based on the correction amount of the rotational speed of the ventilation fan being and the correction amount of the opening degree of the ventilation valve being obtain the final correction amount, expressed as . In this way, the adjustment range of the pigeon position (i.e., the control action combination) is associated with the safety degree of the roadway structure. The more serious the deviation of the safety degree, the greater the adjustment range; Repeatedly calculate the fitness function, update the pigeon flock position and speed until the iteration termination condition is met (reaching the maximum number of iterations ). The control action combination corresponding to the finally obtained global optimal position is the final control action introduced by taking the safety degree of the roadway structure as a dynamic constraint factor and combining the improved pigeon flock optimization algorithm; Convert the final regulation action into specific control instructions and send them to the underground ventilation equipment, control the ventilator to operate at the finally determined speed, and adjust the ventilation valve to the corresponding opening degree.

[0030] Embodiment 2: As Figure 2 shown, a mine intelligent ventilation combined optimization and control system includes: Data acquisition module: Collect multi-source mine ventilation data through a grid deployment strategy using distributed multi-modal acoustic sensors; Fusion and optimization module: Use the theory of belief functions to fuse and optimize multi-source mine ventilation data to obtain comprehensive status data; Reinforcement learning module: Based on the comprehensive status data, obtain the optimal preliminary regulation actions of different mine equipment through an adaptive reinforcement learning model; Regulation and optimization module: Based on the optimal preliminary regulation actions, introduce the roadway structure safety factor as a dynamic constraint factor and combine it with an improved pigeon flock optimization algorithm to obtain the final regulation actions, and realize underground ventilation control based on the final regulation actions.

[0031] In the application, several formulas involved are calculated by taking their numerical values after dimensionless, and the establishment of the formulas is obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. Some coefficients or weights in the formulas are set by those skilled in the art according to the actual situation, so no more details will be given here.

[0032] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution.

[0033] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent combined optimization regulation and control of mine ventilation, characterized in that, The method steps are as follows: S1: Collect multi-source mine ventilation data through a grid deployment strategy using distributed multi-modal acoustic wave sensors; S2: Use the theory of belief functions to fuse and optimize the multi-source mine ventilation data to obtain comprehensive status data; S3: Based on the comprehensive status data, obtain the optimal preliminary control actions for different mine equipment through an adaptive reinforcement learning model; S4: Based on the optimal preliminary control actions, introduce the roadway structure safety factor as a dynamic constraint factor and combine it with an improved pigeon flock optimization algorithm to obtain the final control actions, and implement underground ventilation control based on the final control actions; Among them, the improved pigeon flock optimization algorithm is realized by using chaotic mapping to initialize the pigeon flock position in the pigeon flock optimization algorithm, introducing a diversity feedback mechanism in the process of updating the pigeon flock position, and considering the roadway structure safety factor; The diversity feedback mechanism calculates the diversity index of the pigeon flock, uses the Euclidean distance to measure the average distance between pigeons, and randomly perturbs the positions of the pigeons when the diversity index is lower than the preset threshold.

2. The intelligent ventilation combined optimization regulation and control method for mine according to claim 1, characterized in that The process of collecting multi-source mine ventilation data is as follows: Deploy multimodal acoustic wave sensors, with the adjacent sensor nodes spaced at a distance of in the horizontal direction and in the vertical direction to form a spatial grid. Each sensor node is equipped with a low-power acoustic wave transmitting and receiving module, and each sensor node samples the acoustic wave signals in the mine at the set sampling frequency . The collected acoustic wave signals are discrete signal sequences, and the discrete signal sequences include wind speed data, roadway structure acoustic signals, and equipment rotation speed signals.

3. The intelligent ventilation combined optimal regulation and control method for mine according to claim 2, characterized in that, The process of using the theory of belief functions to fuse and optimize multi-source mine ventilation data is as follows: Define a recognition framework, which includes "ventilation equipment normal" and "ventilation equipment failure"; Extract the features of the multi-source mine ventilation data collected from the distributed acoustic wave sensor array, convert the time-domain signal into a frequency-domain signal through fast Fourier transform, extract the spectral features of the signal, and construct an evidence body based on the extracted spectral features. The evidence body includes ventilation equipment normal and ventilation equipment failure; According to the spectral features, compare with the preset fault feature threshold to determine the basic probability assignment function. For the spectral feature data of all sensor nodes, each node constructs an evidence body based on its own collected data, and uses the Dempster-Shafer combination rule to obtain the trust degree assignment data; Based on the trust degree assignment data, in the order of the propositions in the proposition set, arrange the trust degree assignment data of each proposition in sequence to form a data sequence, and this data sequence is the comprehensive status data.

4. A method for intelligent ventilation combined optimization and control in a mine according to claim 3, characterized in that The process of using the Dempster-Shafer combination rule to obtain the trust degree assignment data is as follows: Based on the evidence bodies of any two sensor nodes, fuse the evidence bodies to obtain the fused basic probability assignment. Traverse the evidence bodies of all sensor nodes for Dempster-Shafer combination rule fusion to obtain the corresponding basic probability assignment, and combine the basic probability assignments of all sensor nodes to obtain the trust degree assignment data.

5. The intelligent ventilation combined optimization regulation and control method for mine according to claim 4, characterized in that, The adaptive reinforcement learning model uses a deep neural network as the basic architecture, introduces the local perception and weight sharing ideas of the convolutional neural network in the hidden layer, and updates the strategy based on the double deep Q network algorithm to adapt to the complex non-linear dynamic characteristics of the mine ventilation system; The adaptive reinforcement learning model includes an input layer, multiple hidden layers and an output layer.

6. The intelligent ventilation combined optimal regulation and control method for mine according to claim 5, characterized in that, The process of obtaining the preliminary control actions is as follows: Based on the input layer of the adaptive reinforcement learning model, receive the comprehensive status data, convert it into a comprehensive vector representing the dimension, and then transmit it to the hidden layer; The hidden layer consists of multiple neurons, which are connected by weights. An activation function is used to perform a non-linear transformation on the output of the neurons, and the ideas of local perception and weight sharing in the convolutional neural network are introduced. Convolution modules are set at intervals to optimize the convolution operation on the local area of the data; The output layer outputs the regulation action scores of different devices. The number of neurons corresponds to the number of device actions to be regulated, and the probabilities of the regulation actions of different devices are output; The strategy is updated through the double deep Q-network algorithm. The double deep Q-network includes a current network and a target network. The Q function is defined, and the Q function includes the parameters of the deep neural network. The current network is used to select actions, and the target network is used to evaluate the action value. The parameters of the current network are updated by minimizing the loss function and an experience replay mechanism is adopted. After multiple rounds of training, the parameters of the current network are updated. When the training converges, the optimal preliminary regulation strategy under different comprehensive states is obtained.

7. A method for intelligent ventilation combined optimization regulation and control in a mine according to claim 6, characterized in that, The process of obtaining the roadway structure safety degree is as follows: Safety sensors are used to monitor the deformation and stress parameters of the roadway. The safety sensors include displacement sensors and stress sensors; The deformation and stress parameters are combined with the geomechanical model to obtain the roadway structure safety index, and the roadway structure safety degree is quantified through an empirical model based on the structure safety index.

8. The intelligent ventilation combined optimal regulation and control method for mine according to claim 7, characterized in that, The process of obtaining the final regulation action by introducing the roadway structure safety degree as a dynamic constraint factor and combining it with the improved pigeon flock optimization algorithm is as follows: The pigeon flock optimization algorithm is introduced, and the positions of the pigeon flock are initialized using chaotic mapping Let the size of the pigeon flock be , where each pigeon represents a combination of regulatory actions, the vector dimension is the same as the number of parameters of the regulatory actions, and the positions of the initial pigeon flock within the parameter value range are randomly generated; The advantages and disadvantages of each regulation action combination are evaluated through a fitness function. For each regulation action combination, its fitness value is calculated according to the fitness function, and the position where the regulation action combination with the highest fitness is located is found, which is the global optimal position; The flying direction of the pigeon flock is adjusted according to the global optimal position, and the flying speed is obtained. The position is updated according to the flying speed; A diversity feedback mechanism is introduced during the position update process. The diversity index in the diversity feedback mechanism is obtained by calculating the Euclidean distance between any two pigeons, then adding up the Euclidean distances between all pairs of pigeons and dividing by the total number of pigeon pairs; According to the correction amount determined by the roadway structure safety degree and the regulation action adjustment rule, when the position of the pigeon causes the roadway structure safety degree to be lower than the preset safety threshold, its position is corrected through the correction amount; Repeatedly calculate the fitness function, update the positions and velocities of the pigeon flock until the iterative termination condition is met, i.e., the maximum number of iterations is reached. The finally obtained global optimal position is the global optimal position. The corresponding combination of regulatory actions is the final regulatory action.

9. The intelligent ventilation combined optimization regulation and control method for mine according to claim 8, characterized in that, The process of determining the correction amount according to the roadway structure safety degree and the regulation action adjustment rule is as follows: Based on the roadway structure safety threshold and the actual roadway structure safety degree corresponding to the pigeon position, the roadway structure safety degree index is obtained; A regulation action adjustment rule is constructed. The regulation action adjustment rule includes the rotational speed of the ventilator and the opening degree of the ventilation valve. The rotational speed correction amount of the ventilator is set as and the opening degree correction amount of the ventilation valve is set as. The rotational speed correction amount of the ventilator is obtained based on the current regulation action combination considering the minimum step size of the rotational speed adjustment of the ventilator, and the opening degree correction amount of the ventilation valve is obtained based on the current regulation action combination considering the minimum step size of the opening degree adjustment of the ventilation valve: The final correction amount is determined based on the rotational speed correction amount of the ventilator and the opening degree correction amount of the ventilation valve.

10. An intelligent ventilation combined optimization and control system for a mine, which uses the method described in any one of claims 1 to 9, is characterized in that, It includes: Data acquisition module: Multi-source mine ventilation data is collected through distributed multi-modal acoustic wave sensors using a grid deployment strategy; Fusion and Optimization Module: Using the theory of belief functions to fuse and optimize multi-source mine ventilation data to obtain comprehensive status data; Reinforcement Learning Module: Based on the comprehensive status data, obtain the optimal preliminary control actions for different mine equipment through an adaptive reinforcement learning model; Regulation and Optimization Module: Based on the optimal preliminary control actions, introduce the roadway structure safety factor as a dynamic constraint factor and combine it with an improved pigeon flock optimization algorithm to obtain the final control actions, and implement underground ventilation control based on the final control actions.

Images