A linear temperature measurement inversion method for goaf based on GA-BP neural network

Abstract

The present invention provides a goaf linear temperature measurement inversion method based on GA-BP neural network, which belongs to the field of goaf coal spontaneous combustion monitoring, including: monitoring and collecting the temperature and gas concentration data of the goaf through a bundle tube monitoring system and an optical fiber temperature measurement system; using a BP neural network to learn and analyze the monitoring data at different times, and judging the coal spontaneous combustion state of the goaf according to the analysis results; the monitoring data includes monitoring time, gas concentration, temperature and monitoring position coordinates; if the coal spontaneous combustion is in a temperature rise state, the goaf fire source inversion model is started to invert the goaf fire source position. The present invention forms a goaf coal spontaneous combustion gas concentration and temperature joint perception network by burying a bundle tube monitoring system and an optical fiber temperature measurement system in the goaf on both sides of the working face inlet and return air lanes, realizing multi-point continuous perception of goaf coal spontaneous combustion, and accurately predicting the temperature of the central ignition point of the goaf coal, thereby reducing the occurrence of mine fire accidents.

Description

Technical Field

[0001] The invention belongs to the technical field of goaf coal spontaneous combustion monitoring, and in particular relates to a goaf linear temperature measurement inversion method based on a GA-BP neural network. Background Art

[0002] Spontaneous combustion of coal in goaf is one of the most important forms of mine fires. Fires caused by spontaneous combustion of coal account for 90% of the total number of mine fires. Since it mainly occurs in hidden places such as goafs in coal mines, it brings great difficulties to its firefighting and prevention. From the cause of spontaneous combustion of coal fires, the accumulation of heat after coal oxidation and heat release is the main cause of spontaneous combustion of coal in goafs. Spontaneous combustion of coal involves the generation and transfer of heat. It is the process of heat generation and transfer in porous media such as floating coal, and it is the result of the game between heat generation and heat dissipation. In-depth research on the mechanism of spontaneous combustion of coal in goafs and analysis of the basic theory of fire prevention and extinguishing technology are important basic supports for preventing and controlling mine fire accidents.

[0003] The prediction of coal spontaneous combustion in goaf usually uses two methods: index gas concentration and temperature measurement. The index gas concentration measurement method uses the coal spontaneous combustion program temperature rise experiment to establish the relationship between coal oxidation spontaneous combustion index gas and temperature based on the oxidation products and high-temperature decomposition products produced during the development of coal spontaneous combustion, determine the early warning indicators and thresholds, and use the bundle tube monitoring system to judge the degree of coal spontaneous combustion in goaf. Affected by the leakage flow field in the goaf, this method has problems such as low monitoring sensitivity in the early stage of coal spontaneous combustion and difficulty in locating high-temperature areas.

[0004] The temperature measurement method uses thermocouple temperature sensors, infrared temperature measurement, optical fiber temperature measurement, and wireless self-organizing network temperature measurement to directly measure the temperature of coal. It has the characteristics of less external interference and accurate measurement. However, due to the poor thermal conductivity of coal, the temperature conduction speed is lower than the gas flow speed. Therefore, the number and location of sensors are the key to detecting high-temperature areas in goafs. Xing Zhen et al. proposed distributed optical fiber temperature detection technology to monitor high-temperature points of fire sources and underground close-range exhaust beam detection technology to monitor gas abnormal points. Due to the restrictions of goaf roof collapse and mine pressure, it is difficult to arrange wired temperature sensors, and the signal transmission distance of wireless self-organizing network temperature measurement technology is short, making it difficult to promote the temperature measurement method on site. Therefore, a method combining temperature measurement with gas concentration can be used to give play to the advantages of fast reaction speed of gas concentration monitoring, intuitive temperature measurement, and high accuracy.

[0005] Domestic and foreign scholars have done a lot of research on the detection of spontaneous combustion fire sources in coal seams, but there are some shortcomings. For example, the method of pre-buried temperature sensors in goaf areas is not only time-consuming and labor-intensive, but also economically unfeasible for long-term and large-scale use. It will also affect production and is not conducive to the realization of high production and efficiency. Infrared thermal temperature measurement technology can only detect the temperature of the solid surface, and the internal temperature cannot be measured. Radon fire zone positioning is mainly carried out on the ground and is suitable for large-scale underground fire areas, while the measurement error of a local location in the goaf of the mining face is large. The previous numerical simulation method often has a large difference between the calculated fire source location and the actual situation due to the model error and unreasonable assumptions of boundary conditions. The geological radar method is subject to too many interference factors, and the measurement effect is not ideal. The gas composition method is better in determining the degree of spontaneous combustion of coal, but it is more difficult to locate the fire source due to the influence of ventilation. In response to the above problems, this project considers establishing a spontaneous combustion model of the goaf through basic scientific laws such as conservation of mass and energy, and using similar simulation methods to accurately measure the thermal physical parameters of the loose coal body, and combining optical fiber temperature measurement technology to realize the goaf temperature field distribution of the model inversion.

[0006] The most important thing in fire prevention and extinguishing in coal seam goaf areas of mines is fire prevention, and the key to fire prevention is early monitoring of coal temperature to prevent spontaneous combustion of residual coal. Only by understanding the changes in the temperature of residual coal in real time can we deal with various spontaneous combustion problems of residual coal freely and completely kill the mine fire in the cradle. Since BP neural network has some shortcomings in use, such as slow convergence speed, easy to fall into local minimum points and difficult to determine the network topology structure, and the characteristics of genetic algorithm itself can just make up for these shortcomings, the study of distributed optical fiber sensing technology of BP neural network optimized by genetic algorithm has important research significance for preventing spontaneous combustion of residual coal in goaf areas. Summary of the invention

[0007] In order to overcome the above-mentioned deficiencies in the prior art, the present invention provides a goaf area linear temperature measurement inversion method based on GA-BP neural network.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A linear temperature measurement inversion method for goaf based on GA-BP neural network, wherein bundle tube monitoring systems and optical fiber temperature measurement systems are buried on both sides of the air inlet tunnel and the return air tunnel of the working face of the goaf, forming a joint perception network of spontaneous combustion gas concentration and temperature of coal in the goaf, realizing multi-point continuous perception of spontaneous combustion of coal in the goaf, and predicting the temperature of the central ignition point of the coal in the goaf, the method comprises the following steps:

[0010] Step 1: monitoring and collecting the temperature and gas concentration data of the goaf area through the bundle tube monitoring system and the optical fiber temperature measurement system;

[0011] Step 2: Use BP neural network to study and analyze the monitoring data at different times, and judge the spontaneous combustion state of coal in the goaf according to the analysis results;

[0012] Step 3: If the coal spontaneous combustion is in a temperature rise state, the goaf fire source inversion model is started to invert the goaf fire source position.

[0013] Preferably, the fire development status includes three conditions: temperature rise, stability and temperature drop.

[0014] Preferably, when the goaf is in a temperature rise state, the goaf fire source inversion model is started to invert the location of the fire source, otherwise it returns to continue monitoring.

[0015] Preferably, the monitoring data includes monitoring time, gas concentration, temperature and monitoring location coordinates.

[0016] Preferably, the fire source inversion process of the goaf fire source inversion model is:

[0017] Step 3.1, model construction: establish a goaf fire source inversion model with the highest temperature point in the goaf as the objective function, the temperature of the air inlet and return tunnel as the decision variable, and the gas concentration at the bundle tube monitoring point, the distribution law of the "three zones" in the goaf, the distribution law of the temperature field in the goaf, and the distribution law of the leakage flow field as the constraint conditions;

[0018] Step 3.2, model simplification: To simplify the computational complexity of the algorithm, combined with the "three-zone" distribution law of the goaf of the working face, the initial population of the GA algorithm only calculates the position in the inclination direction of the working face, with 0 representing no fire source and 1 representing fire source. A single fire source is used, that is, there is only one valid value 1 in each population, and the rest are 0; the population size control is determined by the inclination length of the working face, and the general accuracy is set to 1m;

[0019] Step 3.3, model calculation: Use the strong global search capability of the GA algorithm to optimize the BP neural network results and calculate the optimal solution of the model, which is the location of the fire source in the goaf.

[0020] Preferably, the solution process of the GA algorithm fire source inversion model is:

[0021] Step 3.3.1, algorithm initialization: according to the BP neural network learning results, get the optimized initial weights and thresholds; randomly generate a fire source, encode the random initialization parameters, and use the fitness value to represent the optimized objective function;

[0022] Step 3.3.2, fitness value calculation: The fitness value is the temperature value monitored at different times. According to the air leakage flow field law and temperature change gradient, the fitness value at the optical fiber linear monitoring position in the goaf is calculated;

[0023] Step 3.3.3, algorithm iteration: Execute step 3.3.4 by continuously iterating the error between the training sample network output and the actual output;

[0024] Step 3.3.4, termination condition judgment: The termination condition is that the algorithm exceeds the set iteration step, or the calculation result accuracy meets the set requirements; if the fitness value reaches the set level or triggers other termination conditions, the population is the fire source location;

[0025] Step 3.3.5, population evolution: Perform population selection, crossover, and mutation operations according to the GA algorithm. This process requires ensuring that each generation of population has a single fire source, and then execute step 3.3.2 to perform the evolution operation of the new generation of population.

[0026] The linear temperature measurement inversion method of goaf area based on GA-BP neural network provided by the present invention has the following beneficial effects:

[0027] Based on the gas bundle pipe monitoring system in the goaf area, the present invention designs and develops a bundle pipe intelligent multi-source sampling device, proposes a multi-point continuous monitoring technology for goaf gas, and combines the optical fiber line temperature measurement technology. By burying the bundle pipe monitoring system and the optical fiber temperature measurement system in the goaf area on both sides of the inlet and return air lanes of the working face, a joint perception network of the spontaneous combustion gas concentration and temperature of the goaf area coal is formed, and the multi-point continuous perception of the spontaneous combustion of the goaf area coal is realized, and the temperature of the central ignition point of the coal in the goaf area is accurately predicted, thereby reducing the occurrence of mine fire accidents;

[0028] The present invention studies the spatiotemporal distribution characteristics of thermal conduction in porous media and heat transfer in gas transport in goafs, constructs a joint perception model of gas concentration and temperature of spontaneous combustion of coal in goafs using a GA-BP neural network, and establishes a multi-point continuous perception algorithm for spontaneous combustion of coal in goafs that jointly monitors temperature and gas concentration. This algorithm is of great significance for monitoring and locating spontaneous combustion of coal in goafs and improving prevention of spontaneous combustion of coal. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiment of the present invention and its design scheme, the following briefly introduces the drawings required for this embodiment. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0030] Figure 1 It is a BP neural network model;

[0031] Figure 2 A flow chart of a method for linear temperature measurement inversion in goaf area based on GA-BP neural network provided in an embodiment of the present invention;

[0032] Figure 3 This is a schematic diagram of the joint sensing network of coal spontaneous combustion temperature and gas concentration in goaf;

[0033] Figure 4 This is the layout diagram of the monitoring system for the air intake and return air lanes of the working face in the goaf area. DETAILED DESCRIPTION

[0034] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0035] Example 1

[0036] The present invention provides a linear temperature measurement inversion method for goaf area based on GA-BP neural network, such as Figure 4 As shown in the figure, bundle tube monitoring systems and optical fiber temperature measurement systems are buried on both sides of the air inlet and return air lanes of the working face in the goaf, forming a joint sensing network for the concentration and temperature of spontaneous combustion gas in the goaf, realizing multi-point continuous sensing of spontaneous combustion of coal in the goaf, and predicting the temperature of the central ignition point of the coal in the goaf, such as Figure 2 and Figure 3 As shown, the method specifically comprises the following steps:

[0037] Step 1: Monitor and collect the temperature and gas concentration data of the goaf area through the bundle tube monitoring system and the optical fiber temperature measurement system;

[0038] Step 2: Use BP neural network to study and analyze the monitoring data at different times, and judge the spontaneous combustion state of coal in the goaf according to the analysis results; Figure 1 As shown, the monitoring data includes monitoring time, gas concentration, temperature and monitoring location coordinates;

[0039] Step 2.1: Construct GA-BP neural network model

[0040] BP neural network is a kind of neural network that can deduce the relationship between unknown input and output as long as more data can be obtained, thus realizing the analysis and prediction of any data. Figure 1 As shown in the figure, the BP neural network is divided into three parts, including the input layer, the output layer, and the hidden layer between the input layer and the output layer. The neuron nodes of the input layer are transmitted forward to all the neuron nodes of the hidden layer. After being processed by the hidden layer, they are finally transmitted by the neuron nodes of the output layer. The BP neural network is divided into two tasks, forward propagation of working signals and reverse propagation of error signals. The working signals and error signals are continuously circulated and modified until satisfactory results are obtained, and then they can be stopped, training is completed, and the results are output.

[0041] Genetic algorithm is a search algorithm used to solve optimization problems. It is a type of evolutionary algorithm. It takes all individuals in a population as the object and uses randomization technology to guide efficient search of an encoded parameter space. Basic operations include mutation, crossover and selection to generate high-quality solutions to optimization and search problems.

[0042] Combined with the practical problems of BP neural network structure and parameter optimization, traditional improvement methods include additional momentum terms, adaptive learning rate and adjustable activation function.

[0043] Step 2.2: Sample dataset preprocessing

[0044] When the existing training function is not ideal, other methods need to be selected to optimize the structure and parameters of the BP neural network. Genetic algorithm is such an effective method. Using genetic algorithm to optimize BP neural network can effectively make up for the shortcomings of traditional BP neural network.

[0045] These methods are all implemented in MATLAB as training functions, that is, different options can be selected in the parameter list of the MATLAB function. This implementation method is simple and convenient, but it also limits the scope of selection and cannot reasonably and effectively select a better method according to the actual problem that needs to be solved.

[0046] Step 3: If the coal spontaneous combustion is in a temperature rise state, the goaf fire source inversion model is started to invert the goaf fire source position.

[0047] In this embodiment, the fire development state includes three conditions: temperature rise, stability, and temperature drop. When the goaf is in a temperature rise state, the goaf fire source inversion model is started to invert the location of the fire source, otherwise it returns to continue monitoring.

[0048] Specifically, the fire source inversion process of the goaf fire source inversion model is as follows:

[0049] Step 3.1, model construction: establish a goaf fire source inversion model with the highest temperature point in the goaf as the objective function, the temperature of the air inlet and return tunnel as the decision variable, and the gas concentration at the bundle tube monitoring point, the distribution law of the "three zones" in the goaf, the distribution law of the temperature field in the goaf, and the distribution law of the leakage flow field as the constraint conditions;

[0050] Step 3.2, model simplification: To simplify the computational complexity of the algorithm, combined with the "three-zone" distribution law of the goaf of the working face, the initial population of the GA algorithm only calculates the position in the inclination direction of the working face, with 0 representing no fire source and 1 representing fire source. A single fire source is used, that is, there is only one valid value 1 in each population, and the rest are 0; the population size control is determined by the inclination length of the working face, and the general accuracy is set to 1m;

[0051] Step 3.3, model calculation: Use the strong global search capability of the GA algorithm to optimize the BP neural network results and calculate the optimal solution of the model, which is the location of the fire source in the goaf.

[0052] Furthermore, in this embodiment, the solution process of the fire source inversion model of the GA algorithm is:

[0053] Step 3.3.1, algorithm initialization: according to the BP neural network learning results, get the optimized initial weights and thresholds; randomly generate a fire source, encode the random initialization parameters, and use the fitness value to represent the optimized objective function;

[0054] Step 3.3.2, fitness value calculation: The fitness value is the temperature value monitored at different times. According to the air leakage flow field law and temperature change gradient, the fitness value at the optical fiber linear monitoring position in the goaf area is calculated;

[0055] Step 3.3.3, algorithm iteration: Execute step 3.3.4 by continuously iterating the calculation of the error between the training sample network output and the actual output;

[0056] Step 3.3.4, termination condition judgment: The termination condition is that the algorithm exceeds the set iteration step, or the calculation result accuracy meets the set requirements; if the fitness value reaches the set level or triggers other termination conditions, the population is the fire source location; if the fitness value does not meet the termination condition, execute step 3.3.5;

[0057] Step 3.3.5, population evolution: Perform population selection, crossover, and mutation operations according to the GA algorithm. This process requires ensuring that each generation of population has a single fire source, and then execute step 3.3.2 to perform the evolution operation of the new generation of population.

[0058] The global search capability of the GA algorithm is used to optimize the BP neural network structure, obtain the optimized initial weights and thresholds, encode the randomly initialized parameters, use the fitness value to represent the optimized objective function, and use the continuous iterative calculation of the error between the training sample network output and the actual output to make the fitness value reach the set level or trigger other termination conditions. The GA-BP neural network is used to find the optimal solution, so that all solutions are finally within the optimal solution range.

[0059] The linear temperature measurement inversion method for goaf area based on GA-BP neural network disclosed in the present invention is designed and developed on the basis of goaf area gas bundle tube monitoring system, and a bundle tube intelligent multi-source sampling device is designed and developed, and a multi-point continuous monitoring technology for goaf area gas is proposed. In combination with the optical fiber line temperature measurement technology, a bundle tube monitoring system and an optical fiber temperature measurement system are buried in the goaf on both sides of the working face inlet and return air lanes to form a joint perception network of goaf area coal spontaneous combustion gas concentration and temperature, realize multi-point continuous perception of goaf area coal spontaneous combustion, and accurately predict the temperature of the central ignition point of the coal in the goaf, thereby reducing the occurrence of mine fire accidents.

[0060] Based on the gas bundle pipe monitoring system in the goaf, the present invention combines the optical fiber line temperature measurement technology to propose a multi-point continuous monitoring technology for goaf gas. By burying the bundle pipe monitoring system and the optical fiber temperature measurement system in the goaf on both sides of the working face inlet and return air lanes, a joint perception network of spontaneous combustion gas concentration and temperature of coal in the goaf is formed, so as to realize multi-point continuous perception of spontaneous combustion of coal in the goaf, and accurately predict the temperature of the central ignition point of the residual coal in the goaf, thereby reducing the occurrence of mine fire accidents. The present invention studies the spatiotemporal distribution characteristics of thermal conduction of porous media in the goaf and gas transport and heat transfer, and uses the GA-BP neural network to construct a joint perception model of spontaneous combustion gas concentration and temperature of coal in the goaf, and establishes a multi-point continuous perception algorithm for spontaneous combustion of coal in the goaf with joint monitoring of temperature and gas concentration, which is of great significance for monitoring and positioning spontaneous combustion of coal in the goaf and improving prevention of coal spontaneous combustion.

[0061] The embodiments described above are only preferred specific implementation modes of the present invention, and the protection scope of the present invention is not limited thereto. Any simple changes or equivalent replacements of the technical solutions that can be obviously obtained by any technician familiar with the field within the technical scope disclosed in the present invention belong to the protection scope of the present invention.

Claims

1. A linear temperature measurement inversion method for goaf area based on GA-BP neural network, characterized in that: A bundle tube monitoring system and an optical fiber temperature measurement system are buried on both sides of the air inlet tunnel and the return air tunnel of the working face of the goaf, and the method comprises the following steps: Step 1: monitoring and collecting the temperature and gas concentration data of the goaf area through the bundle tube monitoring system and the optical fiber temperature measurement system; Step 2: Use BP neural network to study and analyze the monitoring data at different times, and judge the spontaneous combustion state of coal in the goaf according to the analysis results; Step 3: If the coal spontaneous combustion is in a temperature rise state, the goaf fire source inversion model is started to invert the goaf fire source position; The fire source inversion process of the goaf fire source inversion model is as follows: Step 3.1, model construction: establish a goaf fire source inversion model with the highest temperature point in the goaf as the objective function, the temperature of the air inlet and return tunnel as the decision variable, and the gas concentration at the bundle tube monitoring point, the "three zones" distribution law of the goaf, the temperature field distribution law of the goaf, and the leakage flow field distribution law as the constraint conditions; Step 3.2, model simplification: To simplify the computational complexity of the algorithm, combined with the "three-zone" distribution law of the goaf of the working face, the initial population of the genetic algorithm only calculates the position in the inclination direction of the working face, with 0 representing no fire source and 1 representing fire source. A single fire source is used, that is, there is only one valid value 1 in each population, and the rest are 0; Step 3.3, model calculation: Use the global search capability of the genetic algorithm to optimize the BP neural network results and calculate the optimal solution of the model, which is the location of the fire source in the goaf.

2. The method for linear temperature measurement inversion in goaf area based on GA-BP neural network according to claim 1 is characterized in that: The fire development states include temperature rise, stability and temperature drop.

3. The method for linear temperature measurement inversion in goaf area based on GA-BP neural network according to claim 2 is characterized in that: When the goaf is in a temperature-rising state, start the goaf fire source inversion model to invert the location of the fire source, otherwise return to continue monitoring.

4. The method for linear temperature measurement inversion in goaf area based on GA-BP neural network according to claim 1 is characterized in that: The monitoring data includes monitoring time, gas concentration, temperature and monitoring location coordinates.

5. The method for linear temperature measurement inversion in goaf area based on GA-BP neural network according to claim 1 is characterized in that: The solution process of the genetic algorithm fire source inversion model is: Step 3.3.1, algorithm initialization: according to the BP neural network learning results, get the optimized initial weights and thresholds; randomly generate a fire source, encode the random initialization parameters, and use the fitness value to represent the optimized objective function; Step 3.3.2, fitness value calculation: The fitness value is the temperature value monitored at different times. According to the air leakage flow field law and temperature change gradient, the fitness value at the optical fiber linear monitoring position in the goaf is calculated; Step 3.3.3, algorithm iteration: use the continuous iterative calculation of the error between the training sample network output and the actual output; Step 3.3.4, termination condition judgment: The termination condition is that the algorithm exceeds the set iteration step, or the calculation result accuracy meets the set requirements; if the fitness value reaches the set level or triggers other termination conditions, the population is the fire source location; if the fitness value does not meet the termination condition, execute step 3.3.5; Step 3.3.5: Population evolution: Perform population selection, crossover, and mutation operations based on genetic algorithms.

Images