A method for predicting air quality

By using environmental data and historical concentration data of mine tunnels in neural networks for training, the problem of low air quality prediction accuracy is solved, improving prediction accuracy and reducing resource waste.

CN118629537BActive Publication Date: 2024-11-19SHAANXI FENGHUA TIMES ENVIRONMENTAL ENG CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411117440.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-19
Estimated Expiration
2044-08-15

AI Technical Summary

Technical Problem

When predicting air quality in mine tunnels, the accuracy of regional prediction without historical data is low and it is difficult to effectively improve.

Method used

By obtaining the current moment particle parameter data of the mine tunnel measurement points, combining environmental data and historical measurement points and prediction points concentration data, the neural network is trained, and the impact of environmental data and motion time is considered to improve the prediction accuracy.

Benefits of technology

It improves the accuracy of air quality prediction at the target prediction point, reduces resource waste, and meets the stability requirements of air quality prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118629537B_ABST
    Figure CN118629537B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of air quality prediction, and to a method for predicting air quality, comprising: obtaining parameter data of particulate matter at a measurement point according to a target acquisition frequency, including concentration data and environmental data, inputting the data into a trained neural network, and outputting a predicted value, wherein the target acquisition frequency is obtained according to the fluctuation degree of historical concentration data; if the predicted value is greater than a set standard value, the air quality of the target prediction point after a set time is poor; the training set of the neural network is: the input data of the training set is the parameter data of the historical measurement point, the output label is the concentration data of the historical prediction point at the moment #imgabs0#; #imgabs1#, #imgabs2# are the acquisition moments of the concentration data of the input data, S is the length of the lane, #imgabs3# is the wind speed of the historical measurement point, and #imgabs4# is the wind speed of the historical prediction point at the moment #imgabs5#. The present invention improves the accuracy of air quality prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of air quality prediction and a method for predicting air quality. Background Art

[0002] The air in mine tunnels usually contains particulate matter with a certain concentration. It is of great significance to predict the air quality in mine tunnels. For example, by predicting the air quality in the tunnel, the ventilation system can be optimized to ensure that the air quality in the tunnel meets the requirements.

[0003] Optimizing the ventilation system provides enough fresh air and removes particulate matter to maintain a safe working environment.

[0004] Relevant prior art, such as the patent application document with application publication number CN113011660A, discloses an air quality prediction method, system and storage medium. The method includes training a neural network based on air quality data within a historical time period to obtain a trained neural network, and then predicting the air quality at a future time based on the air quality data at a current time.

[0005] In the prior art, in a mine tunnel, the above method can be used to predict the concentration of particulate matter at a certain position in the tunnel at a current moment based on the concentration data of particulate matter at the position at a future moment.

[0006] However, when predicting the air quality of a certain place in a mine tunnel, there may be no historical data of the air quality of this place, and there may only be historical data of the air quality of other places in the mine tunnel. At this time, the accuracy of predicting the air quality of this place by the above method is low.

[0007] Based on this, it is particularly important to improve the accuracy of predicting air quality in mine tunnels. Summary of the invention

[0008] In order to solve the above-mentioned technical problem of low accuracy in predicting air quality, the present invention proposes a method for predicting air quality.

[0009] A method for predicting air quality, comprising:

[0010] The parameter data of particulate matter at the current moment of the measurement point in the mine tunnel is obtained according to the acquisition frequency. The parameter data includes concentration data and environmental data. The parameter data is input into the trained neural network to output the concentration prediction value of the target prediction point. If the prediction value is greater than the set standard value, it is considered that the air quality of the target prediction point after the set time is poor.

[0011] The training process is:

[0012] Obtain historical concentration data of particulate matter at multiple moments at historical measurement points and historical prediction points, input the training set into the neural network for training, and obtain a trained neural network;

[0013] The training set includes multiple sample data. For one sample data, it includes input data and its corresponding output label. The input data is the parameter data of the historical measurement point, and the output label is the parameter data of the historical prediction point. Concentration data at the time, where:

[0014]

[0015] is the collection time of the concentration data of the input data, S is the length of the lane between the historical measurement point and the historical prediction point, is the wind speed of the historical measurement point at the time of collection, The wind speed at the historical prediction point at the time of collection.

[0016] The environmental data of the roadway has a certain influence on the concentration of particulate matter at the prediction point. Considering the influence of environmental data on the prediction results can improve the accuracy of the prediction value. For example, environmental data may include the area of ​​the ventilation opening in the roadway, where the area of ​​the ventilation opening has a certain relationship with the concentration data at the prediction point. In a sample data, not only the environmental data of the roadway and the corresponding collection time are included, but also the The concentration data of historical measurement points is also included. The concentration data of the historical prediction point at time. Ideally, the collection time The concentration data of historical measurement points are The concentration data of the historical prediction points at each time are basically the same. This factor can improve the accuracy of the concentration data of the target prediction point predicted by the trained neural network. The concentration data of the historical measurement points and the time of collection The correlation of the concentration data of the historical prediction point is weak. If a sample data of the training includes the collection time The concentration data of the historical measurement points and the time of collection If the concentration data of the historical prediction points are missing, the accuracy of predicting the concentration data of the target prediction points through the trained neural network is low.

[0017] Optionally, the acquisition frequency is a target acquisition frequency, and the target acquisition frequency is:

[0018] For a collection frequency, a sequence of historical concentration data of particulate matter at historical measurement points is obtained, and the fluctuation index r of the sequence is:

[0019]

[0020]

[0021] Where n is the total number of data in the sequence, is the absolute value, is the floating index of the i-th data in the sequence, is the i-th data, is the average value of the floating index of a preset number of data before the i-th data;

[0022] For the fluctuation indicators corresponding to at least two acquisition frequencies, the acquisition frequency corresponding to the maximum value of the fluctuation indicators that are smaller than a preset threshold is used as the target acquisition frequency.

[0023] Considering that a low acquisition frequency may result in lower stability of the obtained concentration data, and may result in lower accuracy of the obtained concentration data. Considering that a too high acquisition frequency will waste certain resources, for example, when the acquisition frequency is too high, it will increase the energy consumption and maintenance cost of the equipment operation, especially the acquisition equipment that requires consumables. Based on this, the present invention obtains a target acquisition frequency based on the fluctuation index of the historical concentration data, wherein the larger the fluctuation index, the worse the stability of the corresponding data. The stability of the data collected at the target acquisition frequency can meet the requirements of subsequent predictions, and collecting data at the target acquisition frequency can reduce unnecessary waste of resources.

[0024] Optionally, the historical measurement point is the mine tunnel measurement point.

[0025] Optionally, the historical prediction point is the target prediction point.

[0026] Optionally, the particle size range of the particulate matter at any collection frequency is a preset range.

[0027] Optional, , is the movement time corresponding to the one sample data, and the set time is the average of the movement times corresponding to at least two sample data.

[0028] The movement time is involved in the training process of the neural network, so the set time can be obtained based on the movement time, which makes the subsequent prediction effect better.

[0029] Optionally, the environmental data includes: the length of the tunnel between the measurement point and the prediction point, the area of ​​all ventilation openings in the tunnel, and the wind speed at the measurement point collection time.

[0030] The length of the lane between the measurement point and the prediction point, the area of ​​all ventilation openings in the lane, and the wind speed at the measurement point acquisition time are related to the concentration data of the prediction point, and are related to the accuracy of the prediction. Based on this, if the environmental data includes the above data, the accuracy of subsequent predictions can be improved.

[0031] Optionally, the environmental data also includes: the average of the area of ​​the tunnel cross-section at the measurement point position and the area of ​​the tunnel cross-section at the prediction point position and the wind speed at the prediction point collection time, wherein, for any environmental data, the preset point collection time corresponding to any environmental data is the same as the measurement point collection time corresponding to any environmental data.

[0032] The area of ​​the tunnel cross section at the measurement point is related to the mean of the area of ​​the tunnel cross section at the prediction point, the wind speed at the prediction point, and the concentration data at the prediction point, and is also related to the accuracy of the prediction. Based on this, if the environmental data includes the above data, the accuracy of subsequent predictions can be improved.

[0033] Optionally, the neural network is a long short-term memory artificial neural network.

[0034] Beneficial effects of the present invention:

[0035] The present invention trains a neural network through the environmental data of the lane, the concentration data of the particulate matter at the historical measurement point, and the concentration data of the particulate matter at the historical prediction point to obtain a trained neural network. The influence of the environmental data on the prediction accuracy is taken into account, and the influence of the movement time on the prediction accuracy is also taken into account, thereby improving the accuracy of the prediction of the air quality at the target prediction point.

[0036] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0038] Figure 1 is a flow chart schematically illustrating a method for predicting air quality according to an embodiment of the present invention;

[0039] Figure 2 The figure schematically shows a mine tunnel, measurement point positions and target prediction point positions according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In some processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as S1, S2, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", "A1", "B1", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do they limit "A1", "B1", "first" and "second" to be different types; in other words, the serial numbers themselves for the components in this article, such as "A1", "B1", "first", "second", etc., are only used to distinguish the objects described, and do not have any order or technical meaning. For example: the first feature of an image is the feature of the image, where the first has no technical meaning and is only used to distinguish objects. The role of adding "first" includes: facilitating narration, understanding and explanation, and facilitating citing different objects, which can make the article more concise and clear. For example: the first feature and the second feature are different features.

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0042] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0043] The air in the mine tunnels usually contains high concentrations of particles, and the diameters of these particles may vary greatly. These particles are harmful to people in the tunnels. In order to reduce the harm of these particles to people in the tunnels, the concentration of these particles can be predicted. Specifically, the concentration of particles in one place in the tunnel can be combined with the concentration data of particles in another place to predict the concentration of particles.

[0044] The present invention predicts the concentration of particulate matter in the lane based on a neural network. Specifically, based on the concentration data of the measurement point at the current moment, the concentration of particulate matter at the target prediction point after a certain period of time at the current moment is predicted. The said period of time will be described in detail later.

[0045] Figure 1 FIG. 1 is a flow chart schematically showing a method for predicting air quality according to an embodiment of the present invention. Figure 1 The specific instructions are as follows.

[0046] Step S1: Obtain parameter data of particulate matter at a measurement point in a mine tunnel at the current moment according to a collection frequency, wherein the parameter data includes concentration data and environmental data.

[0047] Figure 2 Schematic diagram of a mine tunnel, measurement point locations and target prediction point locations according to an embodiment of the present invention. Figure 2 As shown, 201 is the position of the measurement point, 202 is the position of the target prediction point, and 203 is the lane.

[0048] The present invention realizes the prediction of the particle concentration at the target prediction point based on the parameter data of the mine tunnel measurement point.

[0049] The environmental data is the environmental data of the tunnel between the measuring point and the target prediction point of the mine tunnel. For example, the environmental data includes the wind speed at the measuring point at the current moment.

[0050] Environmental data affects the accuracy of predictions. Collecting environmental data and then inputting it into a trained neural network to get a predicted value can improve the accuracy of predictions.

[0051] The particle size range of the particles can be any range, for example, the particle size range can be less than 10 microns.

[0052] The acquisition frequency may be any acquisition frequency, that is, the acquisition frequency is not limited.

[0053] Implementers can set the collection frequency according to specific scenarios, for example, the collection frequency can be once every 30 seconds.

[0054] It should be noted that the method for collecting particulate matter concentration data is an existing technology and will not be described in detail.

[0055] Step S2: Train the neural network to obtain a trained neural network.

[0056] Obtain historical concentration data of particulate matter at at least two moments of the historical measurement point. For example, the total number of the historical concentration data may be 1000. The historical measurement point may be any position point in the mine tunnel, that is, the position of the historical measurement point is not limited.

[0057] Obtain historical concentration data of particulate matter at at least two moments at the historical prediction point. For example, the total number of the historical concentration data may be 1000. The historical prediction point may be a location point in the mine tunnel other than the historical measurement point.

[0058] It should be noted that in order to improve the effect of training the neural network, the distance between the historical measurement point and the historical prediction point should be relatively large, for example, the distance can be 400 meters.

[0059] In order to improve the prediction accuracy of neural network and the generalization ability of neural network model, the environmental data of the lane between the historical measurement points and the historical prediction points need to be used as part of the training set to train the neural network.

[0060] The training set for training the neural network includes at least two sample data, wherein for one sample data, the sample data includes corresponding input data and corresponding output label, and the input data is collected at the time The parameter data of the historical measurement points, including the collection time The concentration data of the historical measurement points and the time of collection The environmental data of the lane between the historical measurement point and the historical prediction point is output as follows: Concentration data at the historical prediction point at the moment.

[0061] in , , is the movement time corresponding to the sample data, is the collection time of the concentration data of the input data corresponding to the sample data, S is the length of the lane between the historical measurement point and the historical prediction point, The collection time of the historical measurement point The wind speed, The collection time of the historical prediction point wind speed.

[0062] It should be noted that each sample data corresponds to an input data and an output label. The sample data, input data and output label correspond to each other.

[0063] The implementer can set the total number of sample data according to the specific scenario. In one embodiment, the total number of sample data can be 800.

[0064] For example, the 5th acquisition moment of the historical measurement point corresponds to a sample data, and the input data of the sample data includes the concentration data and environmental data of the historical measurement point at the 5th acquisition moment. If the movement time corresponding to the 5th acquisition moment is 100 seconds, the output label of the sample data is the concentration data of the historical prediction point 100 seconds after the 5th acquisition moment. For example, if the 5th acquisition moment is the 5th second, the output label of the sample data is the concentration data of the historical prediction point at the 105th second. It should be noted that if there is no concentration data at the 105th second historical prediction point, the concentration data of a certain acquisition moment of the historical prediction point after the 105th second is selected as the output label of the sample data, wherein there is no acquisition moment between the acquisition moment and the 105th second. In one embodiment, if there is no concentration data at the 105th second historical prediction point, the concentration data at the acquisition moment of the historical prediction point closest to the 105th second is selected as the output label of the sample data.

[0065] It should be noted that the collection time of the historical measurement point concentration data corresponds one-to-one with the sample data.

[0066] At this point, the training set is obtained.

[0067] The training set is input into the neural network for training to obtain a trained neural network.

[0068] In one embodiment, the loss function of the neural network is a mean square error loss function, and if the loss function is less than 0.005, the training is considered complete.

[0069] In another embodiment, if the number of training times reaches a preset number of training times, the training is considered completed. The implementer can set the preset number of training times according to the specific scenario. For example, the preset number of training times can be 500.

[0070] It should be noted that the meaning of movement time is: for any particle at a historical measurement point at any time, it will reach the historical prediction point after approximately the movement time corresponding to that moment. In other words, if the influence of interference factors such as vents on the concentration of particles in the roadway is not considered, that is, in an ideal situation, the concentration data of the historical measurement point at that moment is basically equal to the concentration data of the historical prediction point after the corresponding movement time. Therefore, considering the movement time can improve the accuracy of subsequent air quality predictions.

[0071] Step S3: Input the parameter data of the particulate matter at the current moment of the mine tunnel measurement point into the trained neural network, and output the concentration prediction value of the target prediction point.

[0072] It should be noted that the concentration prediction value is the concentration data prediction value of the particulate matter at the target prediction point after the set time at the current moment.

[0073] In one embodiment, it can be known through the training process of the neural network that the set time can be obtained according to the movement time.

[0074] It should be noted that the environmental data input into the trained neural network corresponds to the environmental data used to train the neural network, that is, the environmental data contains the same parameters. For example, the wind speed at the measurement point is a parameter.

[0075] Step S4: If the predicted value is greater than the set standard value, it is considered that the air quality of the target prediction point after the set time at the current moment is poor.

[0076] For example, if the set time is 200 seconds, the predicted value is greater than the set standard value, and the current time is the 0th second, it is considered that the air quality at the target prediction point at the 200th second is poor.

[0077] Implementers can set the standard value according to specific scenarios. For example, if the collected particle size is less than 10 microns, the set standard value can be 150 micrograms per cubic meter.

[0078] In one embodiment, the ratio of the predicted value to the set standard value is calculated and recorded as an evaluation index of air quality. The larger the evaluation index is, the worse the air quality at the target prediction point after the set time at the current moment is.

[0079] In one embodiment, the acquisition frequency corresponding to the mine tunnel measurement point at the current moment is the target acquisition frequency. The following describes the process of obtaining the target acquisition frequency.

[0080] Within the set acquisition time, multiple historical concentration data of particulate matter at any acquisition frequency of the historical measurement point are acquired, wherein the multiple historical concentration data are acquired continuously and the multiple historical concentration data constitute a historical concentration data sequence. For example, the acquisition frequency may be once every 20 seconds.

[0081] The implementer can set the set collection time according to the specific scenario. For example, the set collection time can be 60 minutes.

[0082] Calculate the volatility index of the historical concentration data series. The expression of the volatility index r of the historical concentration data series is:

[0083]

[0084]

[0085] Where n is the total number of data in the historical concentration data series, is the absolute value, is the floating index of the ith data in the historical concentration data sequence, is the i-th data, is the mean of the floating index of the preset number of data before the i-th data. It is called the floating reference value of the i-th data, where the preset number of data are collected continuously and there is no data between the preset number of data and the i-th data.

[0086] The implementer may set the preset number according to the specific scenario. For example, the preset number may be an empirical value of 9.

[0087] It should be noted that if the corresponding fluctuation index is calculated only based on the data in the historical concentration data sequence, the first data in the sequence has no corresponding floating index. Based on this, it is assumed that the k+1 concentration data before the data in the historical concentration data sequence are known, k is the preset number, wherein there is no concentration data between the k+1 concentration data and the data in the historical concentration data sequence, the k+1 concentration data is collected continuously, and the collection frequency of the k+1 concentration data is the same as the collection frequency of the data in the historical concentration data sequence.

[0088] In another embodiment, the first k+1 concentration data in the historical concentration data sequence are removed to obtain a reference sequence, and the fluctuation index of the reference sequence is calculated, and the fluctuation index of the reference sequence is used as the fluctuation index of the historical concentration data sequence.

[0089] In one embodiment, the floating index of the second data in the historical concentration data sequence is used as the floating index of the first data, and the floating reference value of the k+1th data in the historical concentration data sequence is used as the floating reference value of the first to kth data.

[0090] Further explanation: The larger the value, the lower the stability of the concentration data collected at this collection frequency, and the greater the possibility that the collected concentration data has a low accuracy. The reason is that under normal circumstances, the difference between the particle concentrations at adjacent collection moments is small, and the fluctuation index corresponding to the collected concentration data is small.

[0091] The method for obtaining the volatility indicators corresponding to other collection frequencies is the same as the method for obtaining the above-mentioned volatility indicators and will not be repeated here.

[0092] The fluctuation indexes corresponding to all the acquisition frequencies are obtained by traversing, and the acquisition frequency corresponding to the maximum value of the fluctuation indexes less than the preset threshold is taken as the target acquisition frequency.

[0093] The implementer can set the preset threshold according to the specific scenario. For example, the preset threshold can be an empirical value of 0.15.

[0094] In another embodiment, the expression of the fluctuation index r corresponding to any acquisition frequency is: ,in, is a positive correlation normalization function, for example, the positive correlation normalization function can be a positive correlation maximum and minimum normalization function. It is called reference value B1. The reference value B1 corresponding to each acquisition frequency is calculated. The maximum value in the maximum and minimum normalization function is the maximum value of all reference values ​​B1, and the minimum value in the maximum and minimum normalization function is the minimum value of all reference values ​​B1. The traversal operation can obtain the fluctuation index corresponding to each acquisition frequency; the acquisition frequency corresponding to the maximum value of the fluctuation index whose fluctuation index is less than the preset fluctuation threshold is taken as the target acquisition frequency. The implementer can set the preset fluctuation threshold according to the specific scenario. For example, the preset fluctuation threshold can take the empirical value of 0.7.

[0095] The detailed description is given below.

[0096] The greater the acquisition frequency, the greater the possibility that the data collected at the acquisition frequency is more stable, that is, the greater the possibility that the difference between adjacent concentration data collected is smaller. Moreover, at a certain wind speed, the distribution of particulate matter in the air of the tunnel may be uneven. If the acquisition frequency is faster, the concentration data of the collected particulate matter is more likely to be stable. The faster the wind speed in the tunnel, the more likely the concentration data of the collected particulate matter is less stable.

[0097] It is understandable that training the neural network based on more stable concentration data is more effective.

[0098] However, if the acquisition frequency is too high: it will lead to too much concentration data, resulting in too much calculation when processing the concentration data, which will waste certain computing resources; it will increase the energy consumption and maintenance costs of equipment operation, especially for acquisition equipment that requires consumables; it may cause the performance stability of the acquisition equipment to be poor during continuous operation, thereby affecting the accuracy of the collected data.

[0099] Therefore, appropriately increasing the acquisition frequency can improve the accuracy of air quality prediction. The stability of the data collected at the target acquisition frequency can meet the requirements of subsequent predictions, and collecting data at the target acquisition frequency can reduce unnecessary waste of resources.

[0100] In one embodiment, the particle size range of the above-mentioned particulate matter is a set range. The implementer can set the set range according to the specific scenario. The set range can be less than 10 microns.

[0101] In one embodiment, the above-mentioned historical measurement points may be the mine tunnel measurement points.

[0102] In one embodiment, the historical prediction point may be the target prediction point.

[0103] In one embodiment, the particle size range of the particles at the historical measurement point at any acquisition frequency is a preset range. The implementer can set the preset range according to the specific scenario. For example, the particle size range of the particles at the current moment of the acquired mine tunnel measurement point is less than 10 microns, and the preset range can be less than 50 microns, where the particle size corresponding to the preset range is a larger value.

[0104] Because when the wind speed is low, the particles with larger diameters may settle due to their own gravity, so the concentration of particles with larger diameters is more affected by wind speed, etc. At a certain collection frequency, the concentration data of particles with larger diameters is more likely to be less stable. When the wind speed is high, the concentration of particles with larger diameters is more likely to be high. Conversely, when the wind speed is low, the concentration of particles with larger diameters is more likely to be low.

[0105] Therefore, if the target collection frequency corresponding to the larger particle size particles meets the requirements, the stability of the concentration data of the smaller particle size particles collected at the target collection frequency usually also meets the requirements.

[0106] In one embodiment, the set time is the average of the exercise time corresponding to at least two sample data, wherein the sample data corresponds to the exercise time one by one.

[0107] In one embodiment, the average of all movement times is calculated and recorded as the set time.

[0108] In another embodiment, the set time The expression is: ,in, is the wind speed at the measuring point in the mine tunnel at the current moment, is the wind speed at the target prediction point at the current moment, It is the length of the mine tunnel between the measured point and the target prediction point.

[0109] In one embodiment, the environmental data includes: the length of the tunnel between the measurement point and the prediction point, the area of ​​all ventilation openings in the tunnel, and the wind speed at the measurement point collection time.

[0110] In one embodiment, the environmental data also includes: the average of the area of ​​the tunnel cross-section at the measurement point and the area of ​​the tunnel cross-section at the prediction point and the wind speed at the prediction point collection time, wherein for any environmental data, the preset point collection time is the same as the measurement point collection time.

[0111] For example, when making a prediction, the environmental data includes: the length of the tunnel between the mine tunnel measurement point and the target prediction point, the area of ​​all ventilation openings in the tunnel, the wind speed at the mine tunnel measurement point at the current moment, the average of the area of ​​the tunnel cross-section at the mine tunnel measurement point and the area of ​​the tunnel cross-section at the target prediction point, and the wind speed at the prediction point at the current moment.

[0112] For example, when training a neural network, the environmental data of any sample data in the data set includes: the length of the alley between the historical measurement point and the historical prediction point, the area of ​​all vents in the alley and the wind speed at the historical measurement point at the collection time corresponding to the sample data, the average of the area of ​​the alley cross-section at the historical measurement point and the area of ​​the alley cross-section at the historical prediction point, and the wind speed at the historical prediction point at the collection time.

[0113] It should be noted that the measurement points and prediction points used for explanation below correspond to each other. For example, historical measurement points correspond to historical prediction points, and mine tunnel measurement points correspond to target prediction points.

[0114] Further explanation. The length of the lane between the measurement point and the prediction point has a certain relationship with the prediction accuracy of the particle concentration. The longer the length of the lane, the greater the possibility that the prediction accuracy is lower. Therefore, when predicting the concentration of particle matter at the target prediction point, in order to improve the prediction accuracy, the length of the lane needs to be considered.

[0115] There may be vents in the tunnel for exchanging air with the outside world. Because the area of ​​the vents is related to the concentration of particulate matter at the prediction point, when predicting the concentration of particulate matter at the target prediction point, in order to improve the accuracy of the prediction, it is necessary to consider the area of ​​the vents in the tunnel. For example, when the concentration of particulate matter at the measurement point is large, if the area of ​​the vents in the tunnel between the measurement point and the prediction point is also large, the concentration of particulate matter at the measurement point may be small after a period of time when it reaches the prediction point. The reason is that some particulate matter may be discharged from the tunnel through the vents.

[0116] When the wind speed at the measurement point is high, the accuracy of the concentration of particles collected at the measurement point may be low. When the wind speed at the prediction point is high, the accuracy of the concentration of particles collected at the prediction point may be low. Both the measurement point and the prediction point can be called the collection point. The reason is that when the wind speed at the collection point is high, the concentration of particles near the collection point will fluctuate greatly, thereby increasing the uncertainty of the collection; when the wind speed at the collection point is high, some particle collection equipment may not be able to effectively capture particles. Moreover, at a certain wind speed, the smaller the particles, the less likely they are to be captured by the machine that collects the particle concentration; the greater the wind speed, the less likely the particles are to be captured by the machine that collects the particle concentration. In short, there is a certain relationship between the wind speed at the collection point and the accuracy of the collected particle concentration. Therefore, when predicting the concentration of particles at the target prediction point, in order to improve the accuracy of the prediction, the wind speed at the measurement point and the prediction point needs to be considered.

[0117] When the mean of the area of ​​the cross section of the lane at the measurement point position and the area of ​​the cross section of the lane at the prediction point position is large, the mean is called mean A1, and the area of ​​the vent in the lane has a small impact on the predicted value of the particulate matter. The greater the difference between mean A1 and the area of ​​the vent, the smaller the impact of the area of ​​the vent in the lane on the predicted value of the particulate matter. Therefore, when predicting the concentration of particulate matter at the target prediction point, in order to improve the accuracy of the prediction, it is necessary to consider mean A1.

[0118] It should be noted that the above-mentioned environmental data of the lanes are only exemplary, and the implementer can set the environmental data according to the specific scenario. For example, the environmental data may also include the power of the ventilator in the lane between the measurement point and the prediction point. It is understandable that the power of the ventilator in the lane has a certain relationship with the concentration of particulate matter at the prediction point.

[0119] In one embodiment, the neural network may be a long short-term memory artificial neural network (Long Short-Term Memory).

[0120] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0121] As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or." Additionally, the articles "a" and "an" as used in this application and the appended claims are generally understood to mean "one or more" unless specified otherwise or clear from context to be directed to a singular form.

[0122] Where the terms "includes," "having," "having," "having," or variations thereof are used in the detailed description or claims, such terms are intended to be inclusive in a manner similar to the term "comprising."

[0123] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for predicting air quality, characterized in that: include: The parameter data of particulate matter at the current moment of the measurement point in the mine tunnel is obtained according to the acquisition frequency. The parameter data includes concentration data and environmental data. The parameter data is input into the trained neural network to output the concentration prediction value of the target prediction point. If the prediction value is greater than the set standard value, it is considered that the air quality of the target prediction point after the set time is poor. The training process is: Obtain historical concentration data of particulate matter at multiple moments at historical measurement points and historical prediction points, input the training set into the neural network for training, and obtain a trained neural network; The training set includes multiple sample data. For one sample data, it includes input data and its corresponding output label. The input data is the parameter data of the historical measurement point, and the output label is the parameter data of the historical prediction point. The concentration data at the time, , is the collection time of the concentration data of the input data, S is the length of the lane between the historical measurement point and the historical prediction point, is the wind speed of the historical measurement point at the time of collection, The wind speed at the historical prediction point at the time of collection; The acquisition frequency is the target acquisition frequency, and the target acquisition frequency is: For a collection frequency, a sequence of historical concentration data of particulate matter at historical measurement points is obtained, and the fluctuation index r of the sequence is: , , where n is the total number of data in the sequence, is the absolute value, is the floating index of the i-th data in the sequence, is the i-th data, is the average value of the floating index of a preset number of data before the i-th data; For the fluctuation indicators corresponding to at least two acquisition frequencies, the acquisition frequency corresponding to the maximum value of the fluctuation indicators that are smaller than a preset threshold is used as the target acquisition frequency.

2. A method for predicting air quality according to claim 1, characterized in that: The historical measurement points are the mine tunnel measurement points.

3. A method for predicting air quality according to claim 1, characterized in that: The historical prediction point is the target prediction point.

4. A method for predicting air quality according to claim 1, characterized in that: is the movement time corresponding to the one sample data, and the set time is the average of the movement times corresponding to at least two sample data.

5. The method for predicting air quality according to claim 1, characterized in that: The environmental data include: the length of the tunnel between the measuring point and the prediction point, the area of ​​all ventilation openings in the tunnel, and the wind speed at the measuring point collection time.

6. A method for predicting air quality according to claim 5, characterized in that: The environmental data also includes: the average of the area of ​​the tunnel cross-section at the measurement point position and the area of ​​the tunnel cross-section at the prediction point position and the wind speed at the prediction point collection time, wherein, for any environmental data, the preset point collection time corresponding to any environmental data is the same as the measurement point collection time corresponding to any environmental data.

7. A method for predicting air quality according to claim 1, characterized in that: The neural network is a long short-term memory artificial neural network.