A method and device for controlling the power of a fan in a tunnel

Abstract

The application discloses a kind of driving roadway inside ventilator power control method and device, it is related to mine driving roadway one ventilation and three prevention intelligent control technical field, comprising: using multiple wireless multi-parameter sensors arranged at different positions of driving roadway, obtain the concentration value of multiple gases in the air in driving roadway, using pre-constructed fuzzy depth neural network, process the concentration value of multiple gases in the air, output different voltage level values corresponding to multiple wireless multi-parameter sensors, using pre-set multiple-empowerment multiple-index decision analysis algorithm based on grey correlation analysis, evaluate multiple wireless multi-parameter sensors at different positions, obtain the comprehensive score of each wireless multi-parameter sensor, using the voltage level value corresponding to the wireless multi-parameter sensor with the highest score, control the power of ventilator in the region where wireless multi-parameter sensor is located.The method can output the ventilator control result of a certain region according to the actual evaluation result of sensor.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for ventilation, fire prevention, and gas control in mine tunneling roadways, and more specifically to a method and device for controlling the power of ventilation fans in tunneling roadways. Background Technology

[0002] Currently, the common method for air monitoring in tunnel boring machines is to use wired sensors or handheld instruments to collect samples at fixed points or periodically, and then transmit the collected data to a ground monitoring center for analysis and processing. However, this method has the following drawbacks:

[0003] First, traditional single-parameter sensors cannot meet the multi-dimensional and comprehensive data acquisition needs of mine roadways. Using multiple single sensors for gas data acquisition results in insufficient data integration and analysis capabilities, leading to poor monitoring accuracy and reliability. Furthermore, wired sensors require extensive cabling and signal lines, increasing costs and maintenance difficulty, and are susceptible to signal interference or interruption due to the harsh underground environment. Additionally, wired sensor deployment often relies on measuring tapes or visual inspection, resulting in large range errors, overlapping monitoring ranges between sensors, and weak data representativeness. Handheld instruments require manual operation and carrying, which is inefficient and poses errors and safety hazards. These problems lead to low coverage, untimely warnings, and delayed prevention and control in existing monitoring systems. Coal mine accidents often require manual judgment and handling, which is prone to misjudgment and delays, compromising safety. Second, ventilation during shaft construction and roadway excavation... Decision-making is often simplistic and crude, failing to adapt to changes in the working environment. The prevalence of unreliable ventilation and oversized controls is widespread. Mine tunneling control methods suffer from slow response, low precision, and simplistic logic, making it impossible to analyze and make decisions based on monitored gas data. This results in local fans operating at rated power for extended periods, wasting energy and accumulating when methane and smoke exceed limits. Traditional safety monitoring methods cannot meet the safety requirements under the unique conditions of mines, cannot reflect the gas status of tunneling roadways in real time, and are ill-suited for preventing and handling accidents involving excessive methane and smoke levels. Consequently, current traditional ventilation system control methods suffer from slow response and low precision, hindering timely and effective handling of emergencies. Summary of the Invention

[0004] The present invention aims to at least solve one of the technical problems existing in the prior art. To this end, a first aspect of the present invention proposes a method for controlling the power of a ventilation fan in a tunneling roadway, comprising:

[0005] Multiple wireless multi-parameter sensors deployed at different locations in the tunnel were used to obtain the concentration values ​​of various gases in the air within the tunnel.

[0006] By utilizing a pre-built fuzzy deep neural network located within multiple wireless multi-parameter sensors, the concentration values ​​of various gases in the air are processed, and the different voltage level values ​​corresponding to the multiple wireless multi-parameter sensors are output.

[0007] Using a pre-set multi-weighted multi-index decision analysis algorithm based on grey relational analysis, multiple wireless multi-parameter sensors at different locations are evaluated to obtain the comprehensive score of each wireless multi-parameter sensor.

[0008] The power of the ventilation fan in the area where the wireless multi-parameter sensor is located is controlled by using the voltage level value corresponding to the highest-scoring wireless multi-parameter sensor.

[0009] Furthermore, by utilizing multiple wireless multi-parameter sensors deployed at different locations within the tunnel, the concentration values ​​of various gases in the air within the tunnel are obtained, including:

[0010] Using three wireless multi-parameter sensors located in the gas mixing zone, return flow zone, and return airway of the tunneling roadway, the concentration values ​​of various gases in the air were collected, including CH4 concentration, smoke concentration, O2 concentration, and CO concentration.

[0011] Furthermore, by utilizing a pre-built fuzzy deep neural network located within multiple wireless multi-parameter sensors, the concentration values ​​of various gases in the air are processed, and different voltage level values ​​corresponding to the multiple wireless multi-parameter sensors are output, including:

[0012] Determine the input variables, output variables, and their respective membership function parameters of the fuzzy deep neural network;

[0013] Based on the input and output variables, determine the input layer, output layer, and hidden layer of the fuzzy deep neural network, as well as the weight matrix and first bias vector from the input layer to the hidden layer, and the weight matrix and second bias vector from the hidden layer to the output layer.

[0014] The fuzzy deep neural network is trained using random sample values. During training, the membership value of the output variable on its fuzzy subset and the cross loss entropy of the output variable value are used as the loss function. The chain rule is used to calculate the gradient of the output layer and the hidden layer and to perform backpropagation.

[0015] The concentration values ​​of various gases in the air collected by multiple wireless multi-parameter sensors are input into their respective trained fuzzy deep neural networks, which output corresponding voltage level values.

[0016] Furthermore, using a pre-set multi-weighted, multi-index decision analysis algorithm based on grey relational analysis, multiple wireless multi-parameter sensors at different locations are evaluated to obtain a comprehensive score for each wireless multi-parameter sensor, including:

[0017] Multiple objective weighting algorithms are used to process the subjective scoring matrix of multiple indicators of multiple pre-constructed wireless multi-parameter sensors to obtain different weight vectors for each wireless multi-parameter sensor.

[0018] The weight vectors of each wireless multi-parameter sensor are aggregated using the weight aggregation method of grey relational analysis to obtain the comprehensive weight vector of each wireless multi-parameter sensor.

[0019] A multi-index decision analysis algorithm and a weighted average method are used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the score of each wireless multi-parameter sensor.

[0020] Furthermore, various objective weighting algorithms are used to process the subjective scoring matrices of multiple indicators from pre-constructed wireless multi-parameter sensors to obtain different weight vectors for each wireless multi-parameter sensor, including:

[0021] Based on the differences in conditions of wireless multi-parameter sensors, the importance indicators of several wireless multi-parameter sensors are determined. The importance indicators include detection location, durability, reliability and response time.

[0022] A subjective scoring matrix is ​​constructed based on the subjective scores of the importance of multiple wireless multi-parameter multiple sensors under each importance indicator. Then, a normalization formula is used to normalize the different importance indicators of the subjective scoring matrix to obtain a standardized score matrix.

[0023] The standardized score matrix is ​​processed using the analytic hierarchy process to obtain the first weight vector of the importance index of each wireless multi-parameter sensor.

[0024] The entropy weight algorithm is used to process the standardized score matrix to obtain the second weight vector of the importance index of each wireless multi-parameter sensor.

[0025] The standardized score matrix is ​​processed using an indicator importance algorithm based on the correlation between indicators to obtain the third weight vector of the importance indicators of each wireless multi-parameter sensor.

[0026] The standard deviation algorithm is used to process the standardized score matrix to obtain the fourth weight vector of the importance index of each wireless multi-parameter sensor.

[0027] Furthermore, the weight aggregation method of grey relational analysis is used to aggregate the different weight vectors of each wireless multi-parameter sensor to obtain the comprehensive weight vector of each wireless multi-parameter sensor, including:

[0028] Based on the weight aggregation method of grey relational analysis, the first weight vector, the second weight vector, the third weight vector, the fourth weight vector and the fifth weight vector are respectively used with multiple grey relational degree values ​​of the ideal solution;

[0029] By utilizing multiple grey relational degree values, the relative importance weights of various objective weighting algorithms for each wireless multi-parameter sensor are obtained;

[0030] Based on the relative importance weights of various objective weighting algorithms, the comprehensive weights of the various objective weighting algorithms are obtained.

[0031] Furthermore, a multi-index decision analysis algorithm and a weighted average method are used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the score of each wireless multi-parameter sensor, including:

[0032] The TOPSIS algorithm is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the first set of scores after sorting multiple wireless multi-parameter sensors.

[0033] The VIKOR algorithm is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the second set of scores after sorting multiple wireless multi-parameter sensors.

[0034] The ELECTRE algorithm is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the third set of scores after sorting multiple wireless multi-parameter sensors.

[0035] The weighted summation method is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the fourth set of scores after sorting multiple wireless multi-parameter sensors.

[0036] The weighted average method is used to process the sorted first group of scores, the second group of scores, the third group of scores, and the fourth group of scores to obtain the weighted average score of each wireless multi-parameter sensor.

[0037] Furthermore, using the voltage level value corresponding to the highest-scoring wireless multi-parameter sensor, the power of the ventilation fan in the area where the wireless multi-parameter sensor is located is controlled, including:

[0038] The required output power of the ventilator is determined by using the voltage level value corresponding to the highest-scoring wireless multi-parameter sensor. The voltage level values ​​include low voltage, medium voltage, and high voltage, which correspond to the first power value, the second power value, and the third power value, respectively.

[0039] The first power value, the second power value, and the third power value are sent as control commands to the fan power control device.

[0040] The fan power control device adjusts the output power of the fan in real time.

[0041] In another aspect, the present invention provides a power control device for a ventilation fan in a tunneling roadway, comprising:

[0042] The data acquisition module is used to acquire the concentration values ​​of various gases in the air inside the tunnel by using multiple wireless multi-parameter sensors deployed at different locations in the tunnel.

[0043] The first processing module is used to process the concentration values ​​of various gases in the air using a pre-built fuzzy deep neural network located in multiple wireless multi-parameter sensors, and output different voltage level values ​​corresponding to the multiple wireless multi-parameter sensors.

[0044] The second processing module is used to evaluate multiple wireless multi-parameter sensors at different locations using a pre-set multi-weighted multi-index decision analysis algorithm based on grey relational analysis, and obtain the comprehensive score of each wireless multi-parameter sensor.

[0045] The output module is used to control the power of the ventilation fan in the area where the wireless multi-parameter sensor is located, based on the voltage level value corresponding to the highest-scoring wireless multi-parameter sensor.

[0046] This invention provides a method and apparatus for controlling the power of a ventilation fan in a tunneling roadway. Compared with the prior art, its advantages are as follows:

[0047] This invention employs a fuzzy deep neural network algorithm combining fuzzy logic and deep learning. The optimization method is adjusted to improve adaptability and efficiency, successfully solving complex nonlinear problems and achieving high-precision data fitting and prediction. The fuzzy deep neural network algorithm performs fuzzy calculations on monitored gas data. Based on a multi-weighted multi-MCDM method using grey relational analysis, multiple sensors are weighted and determined. Based on the determination results, one sensor output value is selected to control local ventilation. This invention integrates multiple objective weighting and MCDM methods, overcoming the defects and biases of single methods. It utilizes grey relational analysis and weighted average methods to aggregate results, improving consistency and comparability. Furthermore, this method allows adjustment of the weights of each MCDM method to adapt to comprehensive evaluations under different conditions. It can also precisely control ventilation fans within a region based on sensor evaluation results. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0049] Figure 1 A flowchart of a method for controlling the power of a ventilation fan in a tunneling roadway, provided as an embodiment of the present invention;

[0050] Figure 2This is a schematic diagram of a power control device for a ventilation fan in a tunnel, provided as an embodiment of the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] This specification provides the operational steps for the methods described in the embodiments or flowcharts, but may include more or fewer operational steps based on conventional or non-inventive labor. In actual system or server product execution, the methods shown in the embodiments or drawings may be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0053] Figure 1 A flowchart of a ventilation fan power control method in a tunneling roadway provided by an embodiment of the present invention is shown below. Figure 1 As shown, the method includes:

[0054] S101. Using multiple wireless multi-parameter sensors deployed at different locations in the tunnel, the concentration values ​​of various gases in the air within the tunnel are obtained.

[0055] It should be noted that traditional single-parameter sensors cannot meet the multi-dimensional and comprehensive data acquisition needs of mine roadways. Using multiple single sensors for gas data acquisition results in insufficient data integration capabilities, poor monitoring accuracy and reliability. Deploying sensors in mine roadways by relying on measuring tapes or visual inspection leads to large range errors, overlapping monitoring ranges between sensors, and weak representativeness of monitoring data. Therefore, multiple wireless multi-parameter sensors are selected and deployed at different locations in the roadway. Using a laser ranging and alignment device, the monitoring point locations are accurately positioned, improving monitoring coverage and accuracy. Each sensor has computing capabilities, reducing the data processing workload of the host computer and improving the accuracy and efficiency of commands, thus meeting the multi-dimensional and comprehensive data acquisition needs of mine roadways.

[0056] S102. Using a pre-built fuzzy deep neural network located in multiple wireless multi-parameter sensors, the concentration values ​​of multiple gases in the air are processed, and the different voltage level values ​​corresponding to the multiple wireless multi-parameter sensors are output.

[0057] In this invention, each sensor is equipped with a fuzzy deep neural network to perform fuzzy calculations on the monitored gas data. Based on a preset fuzzy algorithm running in the deep neural network, the required output power level of the ventilator is determined, and the calculation results are sent to their respective monitoring system hosts. Each monitoring system host communicates with its respective sensor via LoRa, NB-IoT, RS485, and fiber optic communication. These four communication methods enable switching between different communication methods in various complex environments, ensuring at least two communication methods can be switched between each other. The monitoring system adopts a distributed structure, with each sensor having computing capabilities, reducing the data calculation tasks of the monitoring system host and improving the level of intelligence.

[0058] S103. Using a pre-set multi-weighted multi-index decision analysis algorithm based on grey relational analysis, evaluate multiple wireless multi-parameter sensors at different locations and obtain the comprehensive score of each wireless multi-parameter sensor.

[0059] In step S103, the multi-weighted index decision analysis algorithm based on grey relational analysis, which is mounted on the monitoring system host, is used for decision analysis. Based on the calculation results of the fuzzy deep neural network of the sensors, the ventilation fan is intelligently controlled, which further improves the control accuracy and response speed. The multi-weighted MCDM method based on grey relational analysis assigns weights to the fuzzy calculation results of multiple sensors. This method integrates multiple objective weighting and MCDM methods to overcome the defects and biases of single methods. It uses grey relational analysis and weighted average method to aggregate the results, improving consistency and comparability. This method can adjust the weights of each MCDM method to adapt to the comprehensive evaluation of different situations. It can also output commands that are closer to the actual ventilation needs based on the index scores of the actual difference conditions of the sensors.

[0060] S104. Using the voltage level value corresponding to the wireless multi-parameter sensor with the highest score, control the power of the ventilator in the area where the wireless multi-parameter sensor is located.

[0061] In step S104, after the monitoring system host sets the indicators of each sensor, the monitoring system host will follow the instructions transmitted by the optimal sensor to control the ventilator. As the sensors are used, the staff will change the indicators of each group of sensors in real time, and the control strategy will also change accordingly.

[0062] In one possible implementation, multiple wireless multi-parameter sensors deployed at different locations within the tunnel are used to acquire the concentration values ​​of various gases in the air within the tunnel, including:

[0063] Using three wireless multi-parameter sensors located in the gas mixing zone, return flow zone, and return airway of the tunneling roadway, the concentration values ​​of various gases in the air were collected, including CH4 concentration, smoke concentration, O2 concentration, and CO concentration.

[0064] In the embodiments provided by the present invention, the wireless multi-parameter sensor is an intrinsically safe wireless multi-parameter sensor, which includes: a CH4 sensor, a CO sensor, an O2 sensor, a CO2 sensor and a smoke sensor, used to monitor the concentrations of CH4, CO, O2, CO2 and smoke in the tunnel respectively;

[0065] The intrinsically safe wireless multi-parameter sensor also includes: an ultrasonic temperature and velocity sensor, used to monitor wind speed and temperature field in tunnels;

[0066] The RS-485 module, NB-IoT module, LoRa module, and Bluetooth module are used to implement wired and wireless data transmission methods, respectively; the fuzzy deep neural network module performs preliminary fuzzy calculations on the monitored gas data through a gas concentration fuzzy deep neural network algorithm.

[0067] The display unit is used to display the operating status and monitoring data of the intrinsically safe wireless multi-parameter sensor.

[0068] The air outlet, air pump, and air inlet are used to enable the intrinsically safe wireless multi-parameter sensor to draw gas into the air chamber for detection, as well as the internal gas circulation.

[0069] Audible and visual alarms and voice alarms are used to sound an alarm when an abnormal gas condition is detected.

[0070] Micro wind turbines, wind generators, and energy storage units are used to convert wind power in tunnels into electrical energy and to replenish the energy storage units.

[0071] A laser ranging and alignment device is installed on the ultrasonic temperature and velocity sensor probe to accurately position the monitoring points.

[0072] The sound-light-voice-alarm system consists of a sound-light alarm and a voice alarm working together to provide voice and sound-light alarms based on four conditions: excessive gas, excessive smoke, excessive carbon monoxide, and lack of oxygen.

[0073] In one possible implementation, a pre-built fuzzy deep neural network located within multiple wireless multi-parameter sensors is used to process the concentration values ​​of multiple gases in the air, outputting different voltage level values ​​corresponding to the multiple wireless multi-parameter sensors, including:

[0074] Determine the input variables, output variables, and their respective membership function parameters of the fuzzy deep neural network;

[0075] Based on the input and output variables, determine the input layer, output layer, and hidden layer of the fuzzy deep neural network, as well as the weight matrix and first bias vector from the input layer to the hidden layer, and the weight matrix and second bias vector from the hidden layer to the output layer.

[0076] The fuzzy deep neural network is trained using random sample values. During training, the membership value of the output variable on its fuzzy subset and the cross loss entropy of the output variable value are used as the loss function. The chain rule is used to calculate the gradient of the output layer and the hidden layer and to perform backpropagation.

[0077] The concentration values ​​of various gases in the air collected by multiple wireless multi-parameter sensors are input into their respective trained fuzzy deep neural networks, which output corresponding voltage level values.

[0078] In the embodiments provided by this invention, input variables and output variables, as well as their membership function parameters, are defined. There are four input variables: CH4 concentration, smoke concentration, O2 concentration, and CO concentration. There is one output variable: voltage. The membership function parameter matrix for the input variables is A, with a size of 12*3, and the membership function parameter matrix for the output variable is B, with a size of 3*3. A Gaussian function is used as the membership function, and the Gaussian function is:

[0079]

[0080] Where c is the central value and σ is the standard deviation.

[0081] The membership function parameter matrix of the input variables can be obtained:

[0082] A = ([[0.5,0.1,0.1],#CH4 concentration is low;

[0083] [1,0.2,0.2],#CH4 concentration is normal;

[0084] [1.5,0.3,0.3], #High CH4 concentration;

[0085] [2,0.4,0.4],# Low smoke concentration;

[0086] [4,0.8,0.8],# Smoke concentration is normal;

[0087] [6,1.2,1.2],#High smoke concentration;

[0088] [18,3.6,3.6], #Low O2 concentration;

[0089] [20,4,4],#O2 concentration is normal;

[0090] [22,4.4,4.4], #High O2 concentration;

[0091] [0.0018,0.00036,0.00036], #low CO concentration;

[0092] [0.0024, 0.00048, 0.00048], #CO concentration is normal;

[0093] [0.003, 0.0006, 0.0006]])# High CO concentration;

[0094] B = ([[300,60,60],#low voltage;

[0095] [380,76,76],#Voltage is normal;

[0096] [460,92,92]])# High voltage;

[0097] Define the structure of a fuzzy deep neural network, including the number of neurons in the input layer, hidden layer, and output layer. The hidden layer has 81 neurons, and the output layer has 1 neuron.

[0098] The 81 neurons in the hidden layer refer to the number of fuzzy rules. Each neuron corresponds to a fuzzy rule, such as "if the CH4 concentration is high, the smoke concentration is high, the O2 concentration is low, and the CO concentration is high, then the voltage is high". The output value of each neuron is the activation degree of the fuzzy rule under the current input, which indicates the degree of adaptation of the fuzzy rule to the current input. The larger the value, the more adapted the rule is. The output neuron refers to the value of the output variable, i.e., the voltage. The output value of the output neuron is the final output value obtained from the membership degree of the output variable on each fuzzy subset according to the fuzzy inference and defuzzification methods.

[0099] Define the weight matrix and bias vector for the fuzzy deep neural network and randomly initialize their values. The weight matrix from the input layer to the hidden layer is W1, with a size of 81*12, and the bias vector is b1, with a size of 81*1; the weight matrix from the hidden layer to the output layer is W2, with a size of 1*81, and the bias vector is b2, with a size of 1*1. The weight matrix W1 and the bias vector b1 represent the connection parameters from the input layer to the hidden layer.

[0100] The weight matrix W2 and the bias vector b2 represent the connection parameters from the hidden layer to the output layer. They are both generated using a random number generator and updated according to the learning algorithm.

[0101] Define the learning rate and the number of iterations. The learning rate is 0.01, and the number of iterations is 1000.

[0102] Perform an iterative loop, in each iteration:

[0103] Randomly select a sample. Use the randi function to randomly select the index k of the sample, and obtain the input variable x and the output variable y based on the index k.

[0104] Perform forward propagation to compute the outputs of the input layer, hidden layer, and output layer:

[0105] (1) Calculate the output of the input layer, i.e., the membership matrix U of the input variables, which is 1*12. Iterate for each input variable and for each fuzzy subset to obtain the corresponding membership function parameter a, and use the gaussmf function to calculate the corresponding membership degree.

[0106] The membership degree of the input variable on each fuzzy subset is determined using a Gaussian function as the membership function, and the formula is as follows:

[0107]

[0108] u ij x represents the membership degree of the i-th input variable on the j-th fuzzy subset. i Let a represent the value of the i-th input variable. ij and a ij +1 represents the membership function parameter of the i-th input variable on the j-th fuzzy subset.

[0109] (2) Calculate the output of the hidden layer, i.e. the activation matrix z of the fuzzy rule, which has a size of 1*81, and use the ReLU function as the activation function.

[0110] The activation degree of the fuzzy rule under the current input is calculated using the minimum operator as the connector between the conditions in the fuzzy rule, and the formula is as follows:

[0111]

[0112] Among them, z k u represents the activation level of the k-th fuzzy rule under the current input. iki It represents the membership degree of the i-th input variable on the ki-th fuzzy subset.

[0113] The activation matrix z is the output of the hidden layer, representing the activation level of each fuzzy rule under the current input. The activation matrix z can be calculated as follows:

[0114] z = [ReLU(u·W1+b1)]

[0115] (3) Calculate the output of the output layer, i.e. the membership matrix v of the output variables, which is 1*3 in size, and use the tanh function as the activation function.

[0116] The membership degree of the output variable on each fuzzy subset is determined using the maximum operator as the connector between conclusions in the fuzzy rule. The formula is as follows:

[0117] v j =max(z) k ·w kj )

[0118] Among them, v j z represents the membership degree of the output variable on the j-th fuzzy subset. k w represents the activation level of the k-th fuzzy rule under the current input. kj This represents the element in the k-th row and j-th column of the weight matrix from the hidden layer to the output layer.

[0119] Calculate the value of the output variable y hat The result is the deblurred image. The weighted average method is used as the deblurring method, and the formula is:

[0120]

[0121] Among them, y hat This indicates the value of the output variable, v j b represents the membership degree of the output variable on the j-th fuzzy subset. j This represents the numerical value of the output variable on the j-th fuzzy subset.

[0122] Calculate the loss value, specifically the cross-entropy loss.

[0123]

[0124] Where L represents the loss value, y represents the value of the output variable, and v j This represents the membership degree of the output variable on the j-th fuzzy subset.

[0125] Perform backpropagation to calculate the gradients of the output layer and hidden layers;

[0126] (1) Calculate the gradient dv of the output layer, which has a size of 1*3, and use the chain rule to calculate the gradient.

[0127]

[0128] Among them, dv j It represents the partial derivative of the loss value with respect to the membership degree of the output variable in the j-th fuzzy subset.

[0129] (2) Calculate the gradient dz of the hidden layer, which has a size of 1*81, and use the chain rule to calculate the gradient.

[0130]

[0131] Among them, dzk It represents the partial derivative of the loss value with respect to the activation degree of the k-th fuzzy rule.

[0132] Update the weight matrix and bias vector using the Adam algorithm as the optimization method.

[0133] m t =β1m t-1 +(1-β1)g t

[0134]

[0135]

[0136]

[0137]

[0138] Where, m t v t g t θ t Let β1, β2, and α represent the first-order moment estimate, the second-order moment estimate, the bias-corrected first-order moment estimate, the bias-corrected second-order moment estimate, the gradient, and the parameters, respectively. Let represent the hyperparameters, and t represent the number of iterations.

[0139] In one possible implementation, a pre-set multi-weighted, multi-index decision analysis algorithm based on grey relational analysis is used to evaluate multiple wireless multi-parameter sensors at different locations, obtaining a comprehensive score for each wireless multi-parameter sensor, including:

[0140] Multiple objective weighting algorithms are used to process the subjective scoring matrix of multiple indicators of multiple pre-constructed wireless multi-parameter sensors to obtain different weight vectors for each wireless multi-parameter sensor.

[0141] The weight vectors of each wireless multi-parameter sensor are aggregated using the weight aggregation method of grey relational analysis to obtain the comprehensive weight vector of each wireless multi-parameter sensor.

[0142] A multi-index decision analysis algorithm and a weighted average method are used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the score of each wireless multi-parameter sensor.

[0143] In the embodiments provided by the present invention, the multi-weighted MCDM method (GWA-CDA) based on grey relational analysis is a comprehensive decision analysis method based on grey relational analysis and weighted average method. It integrates multiple objective weighting methods and MCDM methods, and uses grey relational analysis and weighted average method as aggregation tools for decision analysis.

[0144] In one possible implementation, a variety of objective weighting algorithms are used to process the subjective scoring matrix of multiple indicators of multiple pre-constructed wireless multi-parameter sensors to obtain different weight vectors for each wireless multi-parameter sensor, including:

[0145] Based on the differences in conditions of wireless multi-parameter sensors, the importance indicators of several wireless multi-parameter sensors are determined. The importance indicators include detection location, durability, reliability and response time.

[0146] A subjective scoring matrix is ​​constructed based on the subjective scores of the importance of multiple wireless multi-parameter multiple sensors under each importance indicator. Then, a normalization formula is used to normalize the different importance indicators of the subjective scoring matrix to obtain a standardized score matrix.

[0147] The standardized score matrix is ​​processed using the analytic hierarchy process to obtain the first weight vector of the importance index of each wireless multi-parameter sensor.

[0148] The entropy weight algorithm is used to process the standardized score matrix to obtain the second weight vector of the importance index of each wireless multi-parameter sensor.

[0149] The standardized score matrix is ​​processed using an indicator importance algorithm based on the correlation between indicators to obtain the third weight vector of the importance indicators of each wireless multi-parameter sensor.

[0150] The standard deviation algorithm is used to process the standardized score matrix to obtain the fourth weight vector of the importance index of each wireless multi-parameter sensor;

[0151] The standardized score matrix is ​​processed using the mean-weighting algorithm to obtain the fifth weight vector of the importance index for each wireless multi-parameter sensor.

[0152] In the embodiments provided by the present invention,

[0153] Step 1: Assign the first, second, and third sensors A, B, and C, respectively. In practical use, first, based on the sensor differences, obtain the importance index for each of the three sensors. Under operating conditions, assign a subjective score of 1-9 to each sensor for each index, from smallest to largest, representing the expected importance of the sensor for that index. Obtain the subjective scoring matrix, where i and j are any natural numbers from 1, 2, 3, 4, 5.

[0154]

[0155] Based on the subjective scoring matrix, vector normalization is used, i.e. x ij Let x′ represent the element in the i-th row and j-th column of the subjective rating matrix.ij This represents the element in the i-th row and j-th column of the standardized score matrix after vector normalization.

[0156] This yields the standardized score matrix:

[0157] Group Monitoring locations Durability reliability Response time A 0.8018 0.3665 0.6554 0.6554 B 0.5345 0.8552 0.5735 0.5735 C 0.2673 0.3665 0.4915 0.4915

[0158] In one possible implementation, the weight vectors of each wireless multi-parameter sensor are aggregated using the weight aggregation method of grey relational analysis to obtain the comprehensive weight vector of each wireless multi-parameter sensor, including:

[0159] Based on the weight aggregation method of grey relational analysis, the first weight vector, the second weight vector, the third weight vector, the fourth weight vector and the fifth weight vector are respectively used with multiple grey relational degree values ​​of the ideal solution;

[0160] By utilizing multiple grey relational degree values, the relative importance weights of various objective weighting algorithms for each wireless multi-parameter sensor are obtained;

[0161] Based on the relative importance weights of various objective weighting algorithms, the comprehensive weights of the various objective weighting algorithms are obtained.

[0162] In the embodiments provided by this invention, the weight determination is performed according to the following steps:

[0163] Calculate the weights based on all conditions and the standardized matrix above.

[0164] (1) Analytic Hierarchy Process (AHP) to calculate the first weight vector W1: The weights are determined based on the subjective judgment of experts or decision-makers on the importance of the indicators. The rationality of the judgment is verified by constructing a standardized score matrix and calculating the consistency ratio. For example, if the monitoring location is 3 times more important than the response time, 3 times more important than the reliability, and 6 times more important than the durability, the steps are as follows:

[0165] Establish a judgment matrix:

[0166] Monitoring locations Durability reliability Response time Monitoring locations 1 6 3 3 Durability 1 / 6 1 1 / 2 1 / 2 reliability 1 / 3 2 1 1 Response time 1 / 3 2 1 1

[0167] Calculate the weight vector W1

[0168]

[0169] A i It is the element in the i-th row of the judgment matrix, and n is the dimension of the judgment matrix.

[0170] index Monitoring locations Durability False alarm rate Response time Weight 0.5455 0.0909 0.1818 0.1818

[0171]

[0172] Where n is the dimension of the judgment matrix, A is the judgment matrix, W is the weight vector, AW is the judgment matrix multiplied by the weight vector, and the largest eigenvalue represents the largest eigenvalue of the judgment matrix. The closer its value is to n, the better the consistency of the judgment matrix.

[0173]

[0174]

[0175] Where n is the dimension of the judgment matrix, λ max The largest eigenvalue is , and the consistency index represents the degree of inconsistency in the judgment matrix. The smaller the value, the better the consistency of the judgment matrix. CI is the consistency index, and RI is the random consistency index, whose value is found in the table based on the size of n. The consistency ratio indicates whether the consistency of the judgment matrix is ​​within an acceptable range. If its value is less than 0.1, the consistency test is considered passed; otherwise, the elements of the judgment matrix need to be adjusted. CR is calculated to be 0.03, therefore the consistency of the judgment matrix is ​​acceptable.

[0176] (2) Calculation of the second weight vector W2 using the entropy weight method: Based on the sensor's score for each indicator, calculate the information entropy of each indicator, then subtract the information entropy from 1 to obtain the information benefit of each indicator, and finally normalize to obtain the second weight vector of each indicator. The calculation results are shown in the table below:

[0177] formula:

[0178]

[0179] d j =1-e j

[0180]

[0181] in, It is a constant. e is the standardized value of the i-th sensor under the j-th index. j d is the information entropy of the j-th indicator. j The information benefit of the j-th indicator, W 2j is the weight of the j-th indicator in W2, and n is the number of indicators.

[0182] (3) Calculate the third weight vector W3 based on the correlation between indicators: Calculate the third weight vector for each indicator based on the correlation between each indicator and other indicators. The specific formula is as follows:

[0183]

[0184]

[0185] Where, ρ jk It is the correlation coefficient between the j-th indicator and the k-th indicator, I j The importance of the j-th indicator, W 3j is the weight of the j-th indicator in W3, and m is the number of indicators.

[0186] (4) Calculate the fourth weight vector (W4) based on the standard deviation. Calculate the fourth weight vector for each indicator based on the variance of the data for each indicator. The specific formula is as follows:

[0187]

[0188] Among them, s j W is the standard deviation of the data under the j-th indicator. 4j is the weight of the j-th indicator in W4, and m is the number of indicators.

[0189] (5) Calculation of the fifth weight vector W5 using the mean method: Calculate the fifth weight vector for each indicator based on the average value of the data for each indicator. The specific formula is as follows:

[0190]

[0191] in W is the average value of the data under the j-th indicator. 5j is the weight of the j-th indicator in W5, and n is the number of indicators.

[0192] Step 3: Weighted Aggregation Method (GWAM) based on Grey Relational Analysis: The relative importance of each weighting method is determined based on the grey relational degree between its weight vector and the ideal solution. Then, a weighted average method is used to calculate the comprehensive weight vector. The specific steps are as follows:

[0193] (1) Determine the ideal solution in

[0194]

[0195]

[0196] (2) Calculate the grey relational degree for each weighting method;

[0197] The grey relational degree G1 of the AHP method, the grey relational degree G2 of the entropy weight method, the grey relational degree G3 of the CRITIC method, the grey relational degree G4 of the standard deviation algorithm, and the grey relational degree G5 of the average weight algorithm are as follows:

[0198]

[0199] G1=0.99, G2=0.99, G3=0.99, G4=0.98, G5=0.99

[0200] Where ρ is the resolution coefficient, usually taken as 0.5, G i It refers to the grey relational degree of each weighting method, i = 1, 2, 3, 4, 5. It is a numerical value that reflects the similarity or difference between data sequences. Its value range is [0,1]. The closer it is to 1, the more similar the data sequences are and the higher the degree of correlation; conversely, the lower the degree of correlation, the lower the degree of correlation.

[0201] (3) Calculate the relative importance of the weighting method and calculate the comprehensive weight vector

[0202]

[0203] The relative importance of each weighting method is:

[0204] r1=r2=r3=r4=r5=0.20

[0205] r i The values ​​represent the importance of each weight vector in the overall weight vector, i = 1, 2, 3, 4, 5.

[0206] The comprehensive weight vector is

[0207]

[0208] W 合 = (0.386, 0.17, 0.23, 0.20) T

[0209] In one possible implementation, a multi-index decision analysis algorithm and a weighted average method are used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the score of each wireless multi-parameter sensor, including:

[0210] The TOPSIS algorithm is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the first set of scores after sorting multiple wireless multi-parameter sensors.

[0211] The VIKOR algorithm is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the second set of scores after sorting multiple wireless multi-parameter sensors.

[0212] The ELECTRE algorithm is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the third set of scores after sorting multiple wireless multi-parameter sensors.

[0213] The weighted summation method is used to process the comprehensive weight vector of each wireless multi-parameter sensor to obtain the fourth set of scores after sorting multiple wireless multi-parameter sensors.

[0214] The weighted average method is used to process the sorted first group of scores, the second group of scores, the third group of scores, and the fourth group of scores to obtain the weighted average score of each wireless multi-parameter sensor.

[0215] In the embodiments provided by this invention, (1) the TOPSIS method: based on the distance between each sensor and the positive ideal solution and the negative ideal solution, the relative proximity of each sensor is calculated, and the sensors are sorted according to the magnitude of the relative proximity. The specific steps are as follows:

[0216] First, determine the positive and negative ideal solutions. The positive ideal solution is the maximum value of the data for each indicator, and the negative ideal solution is the minimum value of the data for each indicator. That is: A * = (9,7,8,8), A - =(3,3,6,6)

[0217] Then, using the Euclidean distance formula, calculate the distance between each sensor and the positive and negative ideal solutions.

[0218]

[0219]

[0220] in, It is the distance between the i-th sensor and the ideal solution. It is the distance between the i-th sensor and the negative ideal solution. represents the maximum and minimum values ​​in the j-th column of the standardized score matrix, respectively.

[0221] Calculate the relative proximity of each sensor. Use the relative proximity formula:

[0222]

[0223] Finally, the scores are sorted according to their relative similarity to obtain the first set of scores.

[0224]

[0225] (2) VIKOR method: Based on the distance between each sensor and the ideal solution and the group utility value, calculate the comprehensive index value of each sensor, and sort them according to the magnitude of the comprehensive index value. The specific steps are as follows:

[0226] First, determine the ideal solution. The ideal solution is the maximum value of the data for each indicator. That is: A* = (9, 7, 8, 8)

[0227] Next, calculate the distance between each sensor and the ideal solution, and the grouping utility value. Using the following formula, we obtain:

[0228]

[0229]

[0230] Among them, S i R is the distance between the i-th sensor and the ideal solution. i It is the grouping utility value of the i-th sensor. Let represent the maximum and minimum values ​​in the j-th column of the standardized score matrix, respectively. Based on the calculations, the distance and group utility values ​​are obtained:

[0231] Calculate the overall performance value for each sensor using the following formula:

[0232]

[0233] Among them, Q i S is the comprehensive index value of the i-th sensor, v is a weighting parameter between 0 and 1, usually taken as 0.5. min S max R min R max These are the minimum and maximum values ​​of distance and group utility, respectively.

[0234] Obtain the comprehensive index values, sort them, and obtain the second set of scores:

[0235]

[0236] (2) ELECTRE method: Determine the support and ranking of each sensor based on its performance in various indicators. The specific steps are as follows:

[0237] First, determine the preference threshold q for each indicator. j Inductive threshold p j , rejection threshold v j

[0238]

[0239]

[0240]

[0241] The preference threshold q for each indicator j Inductive threshold p j , rejection threshold v j , where w Giα is the weight of the i-th indicator obtained by weight aggregation in grey relational analysis. α, β and γ are three global parameters, which represent the minimum difference required for preference, induction and rejection, respectively. These parameters can be set according to the specific problem and the decision-maker's preferences. Generally, the values ​​are between 0 and 1. They represent how much difference the sensor has on a certain indicator to generate preference, induction and rejection, respectively.

[0242] Then calculate the preference strength between each pair of sensors.

[0243]

[0244] Where I(x) ij -x kj q j p j v j ), is a function between 0 and 1, representing the superiority or inferiority of the i-th sensor relative to the k-th sensor on the j-th metric.

[0245] Then determine the support relation matrix, i.e.

[0246]

[0247] c and d are two cutting thresholds between 0 and 1, typically 0.7 and 0.3.

[0248] Finally, the support vector is determined, i.e.

[0249]

[0250] The third group of scores was C. E = (A: 2, B: 1, C: 0), ranking A is greater than B, and B is greater than C.

[0251] (4) Weighted Sum Method (WSM): This method calculates the weighted total score for each sensor based on its score and weight for each indicator, and then sorts the sensors according to their weighted total scores. The formula for the WSM method is as follows:

[0252]

[0253] Among them, S i It is the weighted total score of the i-th sensor, w j It is the weight of the j-th indicator, x ij It is the score of the i-th sensor under the j-th indicator, and n is the number of indicators.

[0254] The solutions are ranked according to their total scores, and a fourth set of scores is obtained.

[0255]

[0256]

[0257] Step 5: Use the weighted average method to synthesize the MCDM results.

[0258] (1) Assign a weight to each MCDM method to represent the reliability or importance of the method. Assume that this invention assigns a weight of 0.4 to the TOPSIS method, a weight of 0.3 to the VIKOR method, a weight of 0.2 to the ELECTRE method, and a weight of 0.1 to the WSM method.

[0259] (2) Based on the ranking results of each MCDM method, assign a score to each sensor to indicate the degree of performance of the sensor under that method. Assume that this invention assigns 10 points to the first place, 5 points to the second place, and 1 point to the third place.

[0260] (3) Calculate the weighted average score for each sensor and sort them according to the score. The specific formula is as follows:

[0261]

[0262] T i u represents the weighted average score or ranking of the i-th sensor. k s represents the weight of the k-th MCDM method. ik This represents the score or grade of the i-th sensor under the k-th MCDM method, where n represents the number of MCDM methods.

[0263] The final weighted average score for each sensor is: A = 10, B = 4.5, C = 1.

[0264] In summary, this method uses objective or subjective methods to determine the weight of each indicator, and employs grey relational analysis and weighted average as aggregation tools in its decision analysis. The final result, under these indicator conditions, shows that sensor A is the best, sensor B is the second best, and sensor C is the worst.

[0265] In one possible implementation, the power of the ventilation fan in the area where the wireless multi-parameter sensor is located is controlled using the voltage level value corresponding to the highest-scoring wireless multi-parameter sensor, including:

[0266] The required output power of the ventilator is determined by using the voltage level value corresponding to the highest-scoring wireless multi-parameter sensor. The voltage level values ​​include low voltage, medium voltage, and high voltage, which correspond to the first power value, the second power value, and the third power value, respectively.

[0267] The first power value, the second power value, and the third power value are sent as control commands to the fan power control device.

[0268] The fan power control device adjusts the output power of the fan in real time.

[0269] In another aspect, the present invention also provides a power control device 200 for a ventilation fan in a tunnel, such as... Figure 2 As shown, the device includes:

[0270] The data acquisition module 201 is used to acquire the concentration values ​​of various gases in the air inside the tunnel by using multiple wireless multi-parameter sensors arranged at different locations in the tunnel.

[0271] The first processing module 202 is used to process the concentration values ​​of various gases in the air using a pre-built fuzzy deep neural network located in multiple wireless multi-parameter sensors, and output different voltage level values ​​corresponding to the multiple wireless multi-parameter sensors.

[0272] The second processing module 203 is used to evaluate multiple wireless multi-parameter sensors at different locations using a pre-set multi-weighted multi-index decision analysis algorithm based on grey relational analysis, and obtain the comprehensive score of each wireless multi-parameter sensor.

[0273] Output module 204 is used to control the power of the ventilation fan in the area where the wireless multi-parameter sensor is located by using the voltage level value corresponding to the highest-scoring wireless multi-parameter sensor.

[0274] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0275] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method of controlling the power of a fan in a mine roadway, characterised by, The method comprises the following steps: obtaining concentration values of multiple gases in the air in the tunneling roadway by using multiple wireless multi-parameter sensors arranged at different positions in the tunneling roadway; processing the concentration values of the multiple gases in the air by using a pre-constructed fuzzy deep neural network in the multiple wireless multi-parameter sensors, and outputting different voltage level values corresponding to the multiple wireless multi-parameter sensors; evaluating the multiple wireless multi-parameter sensors at different positions by using a pre-set multi-attribute decision analysis algorithm based on grey correlation analysis, and obtaining a comprehensive score of each wireless multi-parameter sensor; controlling the power of the ventilator in the area where the wireless multi-parameter sensor is located by using the voltage level value corresponding to the wireless multi-parameter sensor with the highest score. The step of evaluating the multiple wireless multi-parameter sensors at different positions by using a pre-set multi-attribute decision analysis algorithm based on grey correlation analysis, and obtaining a comprehensive score of each wireless multi-parameter sensor comprises the following steps: processing a subjective scoring matrix of multiple attributes of the multiple wireless multi-parameter sensors pre-constructed by using multiple objective weighting algorithms, and obtaining different weight vectors of each wireless multi-parameter sensor; aggregating the different weight vectors of each wireless multi-parameter sensor by using a weight aggregation method of grey correlation analysis, and obtaining a comprehensive weight vector of each wireless multi-parameter sensor; processing the comprehensive weight vector of each wireless multi-parameter sensor by using a multi-attribute decision analysis algorithm and a weighted average method, and obtaining a score of each wireless multi-parameter sensor.

2. A method of controlling the power of a mine fan in a mine roadway as claimed in claim 1, characterised in that, The step of obtaining concentration values of multiple gases in the air in the tunneling roadway by using multiple wireless multi-parameter sensors arranged at different positions in the tunneling roadway comprises the following steps: collecting concentration values of multiple gases in the air by using three tunneling roadway wireless multi-parameter sensors arranged respectively in a gas mixing area, a backflow area and a return airway of the tunneling roadway, wherein the concentration values of the multiple gases include CH4 concentration values, smoke concentration values, O2 concentration values and CO concentration values.

3. A method of controlling the power of a mine fan in a mine roadway as claimed in claim 1 wherein, The step of processing the concentration values of the multiple gases in the air by using a pre-constructed fuzzy deep neural network in the multiple wireless multi-parameter sensors, and outputting different voltage level values corresponding to the multiple wireless multi-parameter sensors comprises the following steps: determining input variables, output variables and their respective membership function parameters of the fuzzy deep neural network; determining an input layer, an output layer and a hidden layer of the fuzzy deep neural network, and weight matrices and a first bias vector from the input layer to the hidden layer, and weight matrices and a second bias vector from the hidden layer to the output layer according to the input variables and the output variables; training the fuzzy deep neural network by using random sample values, wherein, in the training, a cross-entropy loss of membership values of the output variables on their fuzzy subsets and values of the output variables is used as a loss function, and a chain rule is used to calculate gradients of the output layer and the hidden layer and perform back propagation; inputting the concentration values of the multiple gases in the air collected by the multiple wireless multi-parameter sensors into the trained fuzzy deep neural network of each wireless multi-parameter sensor, and outputting corresponding multiple voltage level values.

4. A method of controlling the power of a mine fan in a mine roadway as claimed in claim 1 wherein, The subjective scoring matrix of the multiple indicators of the multiple wireless multi-parameter sensors is processed by using multiple objective weighting algorithms to obtain different weight vectors of each wireless multi-parameter sensor, including: According to the difference conditions of the wireless multi-parameter sensors, the importance indicators of the multiple wireless multi-parameter sensors are determined, and the importance indicators include detection site, durability, reliability, and response time; According to the subjective scoring matrix constructed by the importance subjective scoring of the multiple wireless multi-parameter sensors under each importance indicator, the different importance indicators of the subjective scoring matrix are normalized by using a normalization formula to obtain a standardized score matrix; The standardized score matrix is processed by using an analytic hierarchy process algorithm to obtain a first weight vector of the importance indicators of each wireless multi-parameter sensor; The standardized score matrix is processed by using an entropy weight algorithm to obtain a second weight vector of the importance indicators of each wireless multi-parameter sensor; The standardized score matrix is processed by using an index importance algorithm based on the correlation between indicators to obtain a third weight vector of the importance indicators of each wireless multi-parameter sensor; The standardized score matrix is processed by using a standard deviation algorithm to obtain a fourth weight vector of the importance indicators of each wireless multi-parameter sensor; The standardized score matrix is processed by using a mean method weight algorithm to obtain a fifth weight vector of the importance indicators of each wireless multi-parameter sensor.

5. A method of controlling the power of a mine fan in a mine roadway as claimed in claim 4 wherein, The different weight vectors of each wireless multi-parameter sensor are aggregated by using a weight aggregation method based on grey correlation analysis to obtain a comprehensive weight vector of each wireless multi-parameter sensor, including: According to the weight aggregation method based on grey correlation analysis, multiple grey correlation degree values of the first weight vector, the second weight vector, the third weight vector, the fourth weight vector, and the fifth weight vector with an ideal solution are respectively obtained; The relative importance weights of the multiple objective weighting algorithms of each wireless multi-parameter sensor are obtained by using the multiple grey correlation degree values; According to the relative importance weights of the multiple objective weighting algorithms, the comprehensive weights of the multiple objective weighting algorithms are obtained.

6. A method of controlling the power of a mine fan in a mine roadway as claimed in claim 1 wherein, The comprehensive weight vector of each wireless multi-parameter sensor is processed by using a multi-indicator decision analysis algorithm and a weighted average method to obtain a score of each wireless multi-parameter sensor, including: The comprehensive weight vector of each wireless multi-parameter sensor is processed by using a TOPSIS algorithm to obtain a first group of scoring values of the multiple wireless multi-parameter sensors after sorting; The comprehensive weight vector of each wireless multi-parameter sensor is processed by using a VIKOR algorithm to obtain a second group of scoring values of the multiple wireless multi-parameter sensors after sorting; The comprehensive weight vector of each wireless multi-parameter sensor is processed by using an ELECTRE algorithm to obtain a third group of scoring values of the multiple wireless multi-parameter sensors after sorting; The comprehensive weight vector of each wireless multi-parameter sensor is processed by using a weighted summation method to obtain a fourth group of scoring values of the multiple wireless multi-parameter sensors after sorting; The weighted average score of each wireless multi-parameter sensor is obtained by using a weighted average method to process the first group of scoring values, the second group of scoring values, the third group of scoring values, and the fourth group of scoring values after sorting.

7. A method of controlling the power of a mine fan in a mine roadway as claimed in claim 1 wherein, The power of the ventilator in the area where the wireless multi-parameter sensor is located is controlled by using the voltage level value corresponding to the wireless multi-parameter sensor with the highest score, including: The output power required by the ventilator is determined by using the voltage level value corresponding to the wireless multi-parameter sensor with the highest score, and the voltage level value includes low voltage, medium voltage and high voltage, which correspond to the first power value, the second power value and the third power value, respectively; The first power value, the second power value and the third power value are sent to the ventilator power control device as control instructions; The output power of the ventilator is adjusted in real time by the ventilator power control device.

8. A power control device for a fan in a tunneling gallery, characterized in that It includes: The data acquisition module is used to acquire the concentration values of multiple gases in the air in the tunneling roadway by using multiple wireless multi-parameter sensors arranged at different positions in the tunneling roadway; The first processing module is used to process the concentration values of multiple gases in the air by using a pre-constructed fuzzy deep neural network located in the multiple wireless multi-parameter sensors, and output different voltage level values corresponding to the multiple wireless multi-parameter sensors; The second processing module is used to evaluate the multiple wireless multi-parameter sensors at different positions by using a pre-set multi-attribute decision analysis algorithm based on grey correlation analysis, and obtain the comprehensive score of each wireless multi-parameter sensor; The output module is used to control the power of the ventilator in the area where the wireless multi-parameter sensor is located by using the voltage level value corresponding to the wireless multi-parameter sensor with the highest score; The multiple wireless multi-parameter sensors at different positions are evaluated by using a pre-set multi-attribute decision analysis algorithm based on grey correlation analysis, and the comprehensive score of each wireless multi-parameter sensor is obtained, including: The subjective scoring matrix of multiple indicators of multiple wireless multi-parameter sensors pre-constructed is processed by using multiple objective weighting algorithms to obtain different weight vectors of each wireless multi-parameter sensor; The different weight vectors of each wireless multi-parameter sensor are aggregated by using the weight aggregation method of grey correlation analysis to obtain the comprehensive weight vector of each wireless multi-parameter sensor; The comprehensive weight vector of each wireless multi-parameter sensor is processed by using a multi-attribute decision analysis algorithm and a weighted average method to obtain the score of each wireless multi-parameter sensor.

Images