CN110555570B - Intelligent prediction method and device for gas concentration of mine limited space disaster - Google Patents
Intelligent prediction method and device for gas concentration of mine limited space disaster Download PDFInfo
- Publication number
- CN110555570B CN110555570B CN201910871668.6A CN201910871668A CN110555570B CN 110555570 B CN110555570 B CN 110555570B CN 201910871668 A CN201910871668 A CN 201910871668A CN 110555570 B CN110555570 B CN 110555570B
- Authority
- CN
- China
- Prior art keywords
- harmful gas
- toxic
- frequency component
- sequence
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Operations Research (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a method and a device for intelligently predicting the concentration of gas of a mine limited space disaster, belonging to the field of mine safety, wherein the method comprises the following steps: preprocessing the monitoring data of the gas concentration of the mine confined space disaster; carrying out improved empirical mode decomposition on the processed data; and predicting the concentration of the disaster gas in the limited space of the mine based on the IEMD-WNN prediction model by using the processed data. The method can effectively make up for the special endpoint benefit of the traditional empirical mode decomposition method in the process of envelope line finding and overcome the serious error of the traditional algorithm by adopting the improved empirical mode decomposition method.
Description
Technical Field
The invention belongs to the field of mine safety, and particularly relates to intelligent prediction of gas concentration of a mine confined space disaster, in particular to an intelligent prediction method and device combining an improved empirical mode decomposition method and a wavelet neural network prediction model.
Background
The underground mining of the metal mine is carried out in a limited roadway space, is limited by the narrow dimension of the roadway, and is less in the roadway communicated with the ground surface, so that convection of air inside and outside the roadway is difficult to form, various pollutants formed in the mining process and toxic and harmful substances generated by self-oxidation of ore rocks are difficult to diffuse rapidly, acute or chronic poisoning of workers is caused, even life is endangered, and serious threat is brought to safety production of enterprises.
The conventional prediction method has certain limitation on the prediction of a nonlinear time sequence, the intelligent prediction method such as Neural Network (NN) has the defects of overfitting, easy falling into local optimization and the like, so that the generalization capability of the conventional prediction method is low, the application of the conventional prediction method is limited to a certain extent, and the gray Model prediction method (Grey Model, GM) is suitable for the condition that original data accords with an exponential distribution rule and fluctuation is not severe and supports the Vector Regression prediction method (SVR), and the main defect is that the determination of Model parameters has no uniform standard and the influence on the precision of a prediction result is large.
The time series of the concentration of the poisonous and harmful gas in the metal mine records the change value of the concentration of the poisonous and harmful gas in continuous time, and cannot reflect the change process of the concentration of the poisonous and harmful gas caused by the combined action of various influencing factors, such as equipment failure, wind speed and wind pressure change and the concentration change of the poisonous and harmful gas caused by other human factors. Due to the combined action of the factors, the concentration change of the poisonous and harmful gas in the mine has the characteristics of randomness, uncertainty, high nonlinearity and the like. Empirical mode decomposition is a signal decomposition method for decomposing a complex nonlinear and non-stationary signal into the sum of a finite number of intrinsic mode functions according to frequency difference, and a traditional algorithm is easy to generate serious errors in envelope calculation, has special endpoint benefits, and causes errors of decomposition results as the errors are gradually transmitted to the inside of data.
Disclosure of Invention
Aiming at the defects or improvement requirements of the prior art, the invention provides an intelligent prediction method and device for the concentration of the gas of the mine limited space disaster, so that the technical problem that the prediction precision of the gas of the mine limited space disaster has certain limitation is solved.
In order to achieve the above object, according to an aspect of the present invention, there is provided a method for intelligently predicting a gas concentration of a disaster in a limited space of a mine, including:
(1) Processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
(2) Determining the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence, and decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term;
(3) Respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension, and then respectively obtaining a predicted value of the high-frequency component, a predicted value of the low-frequency component and a predicted value of the trend term as input quantities of a wavelet neural network prediction model;
(4) And adding the predicted value of the high-frequency component, the predicted value of the low-frequency component and the predicted value of the trend item to obtain the predicted value of the concentration of the toxic and harmful gas.
Preferably, step (1) comprises:
judging abnormal data in the original concentration monitoring time sequence of the toxic and harmful gas by a historical data judgment method and a Grabbs criterion judgment method, and selecting target abnormal data;
and processing the target abnormal data by adopting a moving average line method to obtain an estimation value of each target abnormal data according to the statistical characteristics of the data in the preset range of each target abnormal data, and replacing the target abnormal data with the corresponding estimation value.
Preferably, in the step (2), the target toxic and harmful gas concentration sequence is decomposed into a high frequency component, a low frequency component and a trend term, and the method comprises the following steps:
respectively adding a characteristic wave at two ends of the target toxic and harmful gas concentration sequence by a wavelet neural network method to obtain a continuation toxic and harmful gas concentration sequence;
determining local extreme points of each signal in the extended toxic and harmful gas concentration sequence, and performing cubic spline interpolation processing on each extreme point to obtain an upper envelope line, a lower envelope line and a mean value envelope line of each signal;
and selecting a target signal according with the intrinsic mode function according to the upper envelope line, the lower envelope line and the mean envelope line of each signal, and decomposing the target signal into the sum of a limited intrinsic mode function and a residual quantity to obtain a high-frequency component, a low-frequency component and a trend item of the target toxic and harmful gas concentration sequence.
Preferably, before step (3), the method further comprises: the training step of the wavelet neural network prediction model comprises the following steps:
processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term, and respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence;
classifying samples after phase space reconstruction, determining training samples and test samples, training a wavelet neural network prediction model by the training samples, and continuously correcting parameters of the wavelet neural network prediction model according to errors of a prediction output result and an expected output result so as to continuously approximate the prediction output result of the wavelet neural network prediction model to the expected output result to obtain a trained wavelet neural network prediction model;
and inputting the test sample into the trained wavelet neural network prediction model to judge the prediction precision of the trained wavelet neural network prediction model.
Preferably, from l =2 -1 log 2 And m determines the number of hidden layer neurons in the wavelet neural network prediction model, wherein l is the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence, and m is the number of hidden layer neurons.
According to another aspect of the invention, an intelligent prediction device for gas concentration of a mine limited space disaster is provided, which comprises:
the first preprocessing module is used for processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
the first decomposition module is used for determining the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence and decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term;
the prediction module is used for respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension, and then respectively obtaining a prediction value of the high-frequency component, a prediction value of the low-frequency component and a prediction value of the trend term as input quantities of a wavelet neural network prediction model;
and the prediction result determining module is used for adding the prediction value of the high-frequency component, the prediction value of the low-frequency component and the prediction value of the trend item to obtain the concentration prediction value of the toxic and harmful gas.
Preferably, the first preprocessing module comprises:
the judging module is used for judging the abnormal data in the original concentration monitoring time sequence of the toxic and harmful gas by a historical data judging method and a Grabbs criterion judging method and selecting target abnormal data;
and the first preprocessing submodule is used for processing the target abnormal data by adopting a moving average line method so as to obtain an estimation value of each target abnormal data according to the statistical characteristics of the data in the preset range of each target abnormal data and replace the target abnormal data with the corresponding estimation value.
Preferably, the first decomposition module comprises:
the continuation module is used for adding a characteristic wave at each of two ends of the target toxic and harmful gas concentration sequence through a wavelet neural network method to obtain a prolonged toxic and harmful gas concentration sequence;
the envelope acquisition module is used for determining local extreme points of each signal in the extended toxic and harmful gas concentration sequence, and after cubic spline interpolation processing is carried out on each extreme point, an upper envelope, a lower envelope and a mean envelope of each signal are obtained;
and the first decomposition submodule is used for selecting a target signal according with the intrinsic mode function according to the upper envelope line, the lower envelope line and the mean envelope line of each signal, and decomposing the target signal into the sum of a limited number of intrinsic mode functions and a residual amount so as to obtain a high-frequency component, a low-frequency component and a trend item of the target toxic and harmful gas concentration sequence.
Preferably, the apparatus further comprises:
the second preprocessing module is used for processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
the second decomposition module is used for decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend item, and respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend item according to the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence;
the model training module is used for classifying the samples after the phase space reconstruction, determining training samples and testing samples, training a wavelet neural network prediction model by the training samples, and continuously correcting parameters of the wavelet neural network prediction model according to errors of a prediction output result and an expected output result so as to enable the prediction output result of the wavelet neural network prediction model to continuously approach the expected output result, thereby obtaining the trained wavelet neural network prediction model;
and the model verification module is used for inputting the test sample into the trained wavelet neural network prediction model so as to judge the prediction precision of the trained wavelet neural network prediction model.
Preferably, from l =2 -1 log 2 m determining the number of hidden layer neurons in the wavelet neural network prediction model, wherein l is the target toxicityAnd (3) reconstructing and embedding dimension of the harmful gas concentration sequence, wherein m is the number of hidden layer neurons.
According to another aspect of the present invention, there is provided a computer readable storage medium having stored thereon program instructions, which when executed by a processor, implement the intelligent prediction method for disaster gas concentration in a confined space of a mine as described in any one of the above.
In general, compared with the prior art, the above technical solution contemplated by the present invention can achieve the following beneficial effects: the invention adopts the improved empirical mode decomposition method to effectively make up the special endpoint benefit of the traditional empirical mode decomposition method in the envelope line search, overcomes the serious error of the traditional algorithm, reasonably predicts the change trend of the disaster gas concentration in a period of time in the future, improves the prediction precision while weakening random and uncertain information interference, is a feasible mine limited space disaster gas concentration intelligent prediction method, and has higher practical application value.
Drawings
Fig. 1 is a schematic flow diagram of an intelligent prediction method for disaster gas concentration in a limited space of a mine according to an embodiment of the present invention;
FIG. 2 is a flow chart of IEMD-WNN prediction of toxic and harmful gas concentration according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an intelligent prediction device for disaster gas concentration in a limited space of a mine according to an embodiment of the present invention;
FIG. 4 is a time series of raw CO gas concentrations provided by an embodiment of the present invention;
FIG. 5 is a time series of CO gas concentrations after pretreatment according to an embodiment of the present invention;
fig. 6 is a graph of IEMD decomposition results provided by an embodiment of the present invention;
fig. 7 is a diagram of high frequency components after IEMD decomposition according to an embodiment of the present invention;
fig. 8 is a diagram of a low-frequency component after IEMD decomposition according to an embodiment of the present invention;
fig. 9 is a trend chart after IEMD decomposition according to an embodiment of the present invention;
fig. 10 is a comparison graph of a predicted value and an actual value of a high-frequency component according to an embodiment of the present invention;
fig. 11 is a diagram comparing a predicted value and an actual value of a low-frequency component according to an embodiment of the present invention;
FIG. 12 is a comparison graph of the predicted value and the actual value of the trend term provided by the embodiment of the present invention;
fig. 13 is a comparison between a predicted value and an actual value of the CO gas concentration according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
In order to improve the prediction accuracy of the gas concentration of the mine confined space disaster, the invention provides the intelligent prediction method of the gas concentration of the mine confined space disaster, which effectively analyzes nonlinear information and has smaller prediction error and is based on an Improved Empirical Mode Decomposition (IEMD) -Wavelet Neural Network (WNN) prediction model.
As shown in fig. 1 and fig. 2, in the embodiment of the present invention, a concentration monitoring time series is X = { X (t), t =1,2, …, n }, and the method for intelligently predicting the gas concentration of the mine confined space disaster includes the following steps:
s1: processing abnormal data of the original toxic and harmful gas concentration monitoring sequence to obtain a processed target toxic and harmful gas concentration sequence of X '= { X' (t), t =1,2, …, n };
in the embodiment of the invention, the preprocessing of the disaster gas concentration monitoring data in the limited space of the mine comprises judging the abnormal monitoring data and processing the abnormal monitoring data, and the judging of the abnormal monitoring data comprises a historical data judging method and a Grubbs criterion judging method.
The historical data discrimination method comprises the following steps: according to the mine limited space disaster gas database, judging the abnormal higher or lower value in the monitoring data based on the rules and characteristics of various monitoring data, firstly determining that no emergency occurs at the monitoring point, and then quickly and intuitively judging the abnormal monitoring data according to the characteristics of historical data.
The Grubbs criterion (Grubbs) discrimination method has low requirements on the sample size of the monitoring data. The time sequence of the toxic and harmful gas concentration of a certain monitoring point is X = { X (t), t =1,2, …, n }, and the abnormal value judgment step is as follows:
(1) Arranging x (t) into sequential statistics in ascending order, i.e.
x(1)≤x(2)≤…≤x(n)
(2) Calculating the lower side Grabs number g (1) and the upper side Grabs number g (n)
Wherein, the first and the second end of the pipe are connected with each other,
the sigma is the variance of the whole monitoring sample;
(3) Calculating the number G (n, alpha) of the Grubbs criterion according to the significance level alpha;
wherein, the significance level alpha can be determined according to actual needs, and is generally 0.05 or 0.01.
(4) And (3) abnormal value judgment:
if G (1) is more than or equal to G (n, alpha), x (1) is an abnormal value;
if G (n) is more than or equal to G (n, alpha), x (n) is an abnormal value;
(5) And repeating the steps until all abnormal values in the monitoring time sequence are judged.
The abnormal monitoring data processing mode is as follows: according to the statistical characteristics of the data in a period of time near the abnormal data points, the original abnormal monitoring data is replaced by the obtained estimation value through a statistical principle and a data analysis method.
Wherein, the moving average line method is as follows: if the time span of monitoring the abnormal data of the concentration of the toxic and harmful gas is short and generally does not exceed 12 hours, the abnormal monitoring data can be processed by adopting a moving average line method. The time series of poisonous and harmful gas concentrations formed by the poisonous and harmful gas concentration monitoring data is X = { X (t), t =1,2, …, n }, and if an abnormal value occurs at a certain time t = i, a moving average line value X (i) can be calculated instead of the abnormal value, and X (i) can be calculated by n (i) before the time t = i x Calculating the mean value of the data points to obtain:
s2: determining a reconstruction embedding dimension l of a target toxic and harmful gas concentration sequence X '= { X' (t), t =1,2, …, n }, namely the number l of input neurons of a wavelet neural network prediction model;
the reconstruction embedding dimension l of the target toxic and harmful gas concentration sequence X '= { X' (t), t =1,2, …, n } can be determined by using a G-P algorithm, and the specific method is not limited uniquely in the embodiment of the present invention.
S3: decomposing a target toxic and harmful gas concentration sequence X '= { X' (t), t =1,2, …, n } into a high-frequency component, a low-frequency component and a trend term;
the IEMD may be used to decompose the target toxic and harmful gas concentration sequence X '= { X' (t), t =1,2, …, n } into a high frequency component, a low frequency component, and a trend term, and specifically, what manner is used in the embodiment of the present invention is not limited uniquely.
The empirical mode decomposition is a signal decomposition method for decomposing a complex nonlinear and non-stationary signal into the sum of a finite number of intrinsic mode functions according to the difference of frequencies. Empirical mode decomposition of a signal is the process of envelope finding when moving average iteration is performed on the signal.
In an embodiment of the present invention, an improved empirical mode decomposition IEMD method is performed on preprocessed data, where the improved empirical mode decomposition IEMD method is: for a given signal X = { X (t), t =1,2, …, n }, a characteristic wave is added to each end of an original signal through a wavelet neural network learning algorithm, continuation at the end point is intelligently completed through self-learning, and specific steps of determining the continuation are as follows:
(1) Generating a learning sample matrix P according to a predetermined determination rule (e×h) And a target vector Q corresponding thereto (1×h) H is the number of learning samples, and e is the number of data points set during continuation;
(2) For omics Xi Yangben P e,i And a target vector Q i There is a corresponding weight vector W i And offset b i So that it satisfies:
n i ＝(W i ×P e,i )+b i (4)
a i ＝f(n i ) (5)
wherein f is a transfer function and n i As an input term of a transfer function, a i Representing the converted analog vector, i representing the ith learning sample;
(3) According to the analog vector set a (1×f) And a target vector Q (1×f) Determining a weight vector W and an offset b by adopting a least square method;
(4) Determining a first continuation value a at the boundary according to formula (4) and formula (5) based on the determined weight vector W and the offset b 1 '；
(5) With the first continuation value a 1 ' repeating the steps (2) to (4) for new boundary points until all continuation sequences are determined.
S4: performing phase space reconstruction on the decomposed high-frequency component, low-frequency component and trend term according to the reconstruction embedding dimension l, and then using the phase space reconstruction as the input quantity of a wavelet neural network prediction model to respectively obtain a prediction value of the high-frequency component, a prediction value of the low-frequency component and a prediction value of the trend term;
the phase space reconstruction aims to meet the requirement of the wavelet neural network prediction model on the input dimension.
In the embodiment of the present invention, the method further includes: training a wavelet neural network prediction model:
processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
decomposing the concentration sequence of the target poisonous and harmful gas into a high-frequency component, a low-frequency component and a trend term, and respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension of the concentration sequence of the target poisonous and harmful gas;
classifying the samples after the phase space reconstruction, determining training samples and test samples, training a wavelet neural network prediction model by the training samples, and continuously correcting parameters of the wavelet neural network prediction model according to errors of a prediction output result and an expected output result so as to continuously enable the prediction output result of the wavelet neural network prediction model to approach the expected output result, thereby obtaining the trained wavelet neural network prediction model;
and inputting the test sample into the trained wavelet neural network prediction model to judge the prediction precision of the trained wavelet neural network prediction model.
In the embodiment of the invention, the hidden layer of the wavelet neural network learning algorithm adopts the wavelet basis function as the excitation function, so that the neural network has higher prediction precision and faster convergence speed in classification and prediction, the hidden layer Sigmoid function in the BP neural network is replaced by the wavelet basis function as the excitation function, signals are transmitted in the forward direction, and errors are transmitted in the reverse direction, so that the organic integration of wavelet analysis and the neural network is realized.
The concentration value of the toxic and harmful gas in the metal mine is a group of data sequences which change along with time, and factors influencing the concentration value are more, so that the time sequence has non-stable and nonlinear dynamic characteristics. According to the chaos time sequence theory, an embedding dimension l of the chaos time sequence is obtained by using a G-P method, and then the neuron number of the wavelet neural network input layer is determined according to the embedding dimension l.
The number of hidden layer neurons of the wavelet neural network has a great influence on the prediction precision of the wavelet neural network: when the number of the neurons is too small, the network learning effect is poor, and the prediction precision is insufficient; when the number of the neurons is too large, the network learning time is long, and the phenomenon of overfitting is easy to occur. Formula (6) is employed as a method for determining hidden layer neurons.
l＝2 -1 log 2 m (6)
In the formula, l is the number of neurons in the input layer, and m is the number of neurons in the hidden layer.
Let y (1), y (2), …, y (n) be the prediction output of wavelet neural network, n be the number of output layer neurons, v ij And v jk For weight, when the input signal sequence is x (i) (i =1,2, … l), the hidden layer output is:
where h (j) is the output of the jth neuron in the hidden layer, ψ a,b Is a wavelet basis function, a j For wavelet basis function psi a,b Scaling factor of b j For wavelet basis function psi a,b The translation factor of (c).
Thus, the wavelet neural network output layer can be obtained as follows:
the Morlet wavelet function has better local characteristics of time domain and frequency domain and higher resolution, so the Morlet wavelet function is adopted as the hidden layer excitation function of the wavelet neural network. In order to make the predicted output value of the wavelet neural network approach to the expected value to the maximum extent, the weight of the wavelet neural network and the parameters of the wavelet basis function are adaptively corrected by using a gradient descent correction method, and the specific correction process is as follows:
(1) Calculating the minimum mean square error function E predicted by the wavelet neural network, and assuming that the expected output value is d k
(2) Correcting weight and wavelet basis function parameters according to prediction error E
When the wavelet neural network method is practically applied to prediction, the specific process is as follows:
(1) Initializing a network structure: randomly initializing network connection weights v ij 、v jk And wavelet scaling function parameter a j 、b j Setting the network learning rate to be eta;
(2) Data normalization and classification: standardizing the original data and dividing the standardized original data into training data and testing data, wherein the training data is used for network learning, and the testing data is used for judging the prediction accuracy of the network;
(3) Training a wavelet neural network: inputting training data into a wavelet neural network structure to obtain a corresponding prediction result, comparing the prediction result with an actual result, and calculating an error between the network output and an expected output;
(4) Self-adaptive correction of weight: various parameters of the wavelet neural network are continuously corrected according to the errors, so that the output value of the network continuously approaches to the expected output value;
(5) Wavelet neural network prediction: and inputting the test data into the trained wavelet neural network to obtain a corresponding prediction result.
S5: and adding the predicted value of the high-frequency component, the predicted value of the low-frequency component and the predicted value of the trend term to obtain a predicted value of the concentration of the toxic and harmful gas.
Fig. 3 is a schematic structural diagram of an intelligent prediction apparatus for disaster gas concentration in a limited space of a mine, which is provided by the invention, and comprises:
the first preprocessing module is used for processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
in an embodiment of the present invention, the first preprocessing module includes:
the judging module is used for judging the abnormal data in the original concentration monitoring time sequence of the toxic and harmful gas by a historical data judging method and a Grabbs criterion judging method and selecting target abnormal data;
and the first preprocessing submodule is used for processing the target abnormal data by adopting a moving average line method so as to obtain an estimation value of each target abnormal data according to the statistical characteristics of the data in the preset range of each target abnormal data and replace the target abnormal data with the corresponding estimation value.
The first decomposition module is used for determining the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence and decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term;
in an embodiment of the present invention, the first decomposition module includes:
the continuation module is used for adding a characteristic wave at each end of the target toxic and harmful gas concentration sequence through a wavelet neural network method to obtain a prolonged toxic and harmful gas concentration sequence;
the envelope line acquisition module is used for determining local extreme points of each signal in the extended toxic and harmful gas concentration sequence, and after cubic spline interpolation processing is carried out on each extreme point, an upper envelope line, a lower envelope line and a mean value envelope line of each signal are obtained;
and the first decomposition submodule is used for selecting a target signal according with the intrinsic mode function according to the upper envelope line, the lower envelope line and the mean envelope line of each signal, and decomposing the target signal into the sum of a limited intrinsic mode function and a residual quantity so as to obtain a high-frequency component, a low-frequency component and a trend item of the target toxic and harmful gas concentration sequence.
The prediction module is used for respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension, and then respectively obtaining a prediction value of the high-frequency component, a prediction value of the low-frequency component and a prediction value of the trend term as input quantities of a wavelet neural network prediction model;
in an embodiment of the present invention, the apparatus further includes:
the second preprocessing module is used for processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
the second decomposition module is used for decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term, and respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence;
the model training module is used for classifying the samples after the phase space reconstruction, determining training samples and testing samples, training a wavelet neural network prediction model by the training samples, and continuously correcting parameters of the wavelet neural network prediction model according to errors of a prediction output result and an expected output result so as to enable the prediction output result of the wavelet neural network prediction model to continuously approach the expected output result, thereby obtaining the trained wavelet neural network prediction model;
and the model verification module is used for inputting the test sample into the trained wavelet neural network prediction model so as to judge the prediction precision of the trained wavelet neural network prediction model.
And the prediction result determining module is used for adding the prediction value of the high-frequency component, the prediction value of the low-frequency component and the prediction value of the trend item to obtain the concentration prediction value of the toxic and harmful gas.
The specific implementation of each module may refer to the description in the method embodiment, and the embodiment of the present invention will not be described again.
In another embodiment of the present invention, a computer readable storage medium is further provided, on which program instructions are stored, and when the program instructions are executed by a processor, the method for intelligently predicting the concentration of the disaster gas in the mine-limited space is implemented as described in any one of the above.
It should be noted that, according to the implementation requirement, each step/component described in the present application can be divided into more steps/components, and two or more steps/components or partial operations of the steps/components can be combined into new steps/components to achieve the purpose of the present invention.
The above-described method according to the present invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium such as a CD ROM, a RAM, a floppy disk, a hard disk, or a magneto-optical disk, or as computer code originally stored in a remote recording medium or a non-transitory machine readable medium downloaded through a network and to be stored in a local recording medium, so that the method described herein can be stored as such software processing on a recording medium using a general purpose computer, a dedicated processor, or programmable or dedicated hardware such as an ASIC or FPGA. It will be appreciated that the computer, processor, microprocessor controller or programmable hardware includes memory components (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code that, when accessed and executed by the computer, processor or hardware, implements the processing methods described herein. Further, when a general-purpose computer accesses code for implementing the processes shown herein, execution of the code transforms the general-purpose computer into a special-purpose computer for performing the processes shown herein.
Taking the monitoring research of the concentration of toxic and harmful gas in a certain mine in Guizhou as an example, the intelligent prediction application effect of the gas concentration of the mine limited space disaster is analyzed.
(1) In the embodiment, certain phosphorite mountain carbon monoxide gas is selected as an analysis object, 1440 groups of monitoring data of 1130-middle-section return airway 1# monitoring points from 2015 at 1 month and 1 day to 2015 at 3 months and 15 days are collected, the sampling interval is 15min, the data of the first 10 days are selected as training samples of a wavelet neural network prediction model, the data of the last 5 days are selected as test samples, and the original toxic and harmful gas concentration time sequence is shown in fig. 4.
(2) The time sequence after filtering and preprocessing the abnormal data in this embodiment is shown in fig. 5.
(3) The results of applying the modified empirical mode decomposition method to the processed CO gas time series for decomposition are shown in fig. 6.
(4) In this embodiment, the instantaneous frequency of each IMF component obtained by subjecting the IMF component obtained by decomposition to hilbert transform is shown in table 1. As can be seen from table 1, the instantaneous frequency of IMF7 fluctuates greatly with respect to the aforementioned instantaneous frequency, and accordingly, the IMF1 to IMF6 components are defined as higher frequency components, IMF7 and IMF8 are defined as lower frequency components, IMF9 is defined as a trend term, and then the higher frequency components are summed to form high frequency components, and the lower frequency components are summed to form low frequency components, as shown in fig. 7, 8, and 9.
TABLE 1 instantaneous frequency of each IMF component after IEMD decomposition
| Component(s) of | IMF1 | IMF2 | IMF3 | IMF4 | IMF5 | IMF6 | IMF7 | IMF8 | IMF9 |
| Instantaneous frequency | 1.373 | 0.709 | 0.536 | 0.707 | 0.491 | 0.732 | 0.423 | 0.284 | 0.006 |
(5) As can be seen from fig. 7, 8, and 9, the high-frequency component, the low-frequency component, and the trend term of the decomposed CO gas concentration are obvious in characteristic, and the resolution is high, which is beneficial to improving the prediction accuracy of the wavelet neural network model.
(6) In this embodiment, the phase space reconstruction is performed on the high-frequency components, the low-frequency components and the trend terms in the training samples and the test samples according to the calculated spatial reconstruction dimension of the samples, so as to form input samples of the wavelet neural network prediction model. In the embodiment, the wavelet neural network structure is 4-8-1, the input layer has 4 nodes and represents the concentration values of adjacent CO gases at the previous 4 time points of the predicted nodes; the output layer has 1 node, which is the CO gas concentration to be predicted.
(7) In this embodiment, the weights of the wavelet neural network prediction model and the parameters of the wavelet basis functions are randomly obtained during initialization, the network is trained 100 times by using training samples of high-frequency components, low-frequency components and trend items, the trained wavelet neural network model is used to predict the high-frequency components, the low-frequency components and the trend items in the test sample, and the prediction results are shown in fig. 10, fig. 11 and fig. 12.
(9) As can be seen from fig. 10, 11, and 12, the predicted values and actual values of the high-frequency component, the low-frequency component, and the trend term have smaller errors, and the prediction accuracy is higher, wherein the prediction accuracy of the low-frequency component and the trend term is higher, mainly because the low-frequency component and the trend term obtained after the time series of the CO gas concentration is subjected to improved empirical mode decomposition have simple features, and simultaneously the local features of the high-frequency component are easy to identify, which indicates that the IEMD method can simplify the prediction problem of the time series of the CO gas concentration and reduce the prediction complexity.
(10) In this embodiment, the final predicted value of the CO gas concentration may be obtained by adding the prediction results of the high-frequency component, the low-frequency component, and the trend term, and the ratio of the predicted value to the actual value of the CO gas concentration is as shown in fig. 13.
(11) As can be seen from fig. 13, through the learning of the CO gas concentration monitoring data of the first 10 days, the prediction effect of the CO gas concentration of the last 5 days is good, and the prediction effect is good due to the high degree of fitting with the actual monitoring value of the CO gas. According to the final prediction result error analysis, the maximum absolute error between the predicted value and the actual value of the CO gas is 0.09, and the error rate is 6.5% at most. The method has the advantages that the prediction complexity of the CO gas concentration is reduced and the accuracy of the prediction result is improved by combining the improved empirical mode decomposition method with the wavelet neural network prediction model, and the method is a feasible prediction method of the concentration of the toxic and harmful gas at the single measuring point of the metal mine and has high practical application value.
By the method, the gas concentration of the mine limited space disaster can be predicted, the error is smaller, and the precision is higher.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (8)
1. The intelligent prediction method for the gas concentration of the mine limited space disaster is characterized by comprising the following steps:
s1: processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
the step S1 comprises the following steps:
judging abnormal data in the original concentration monitoring time sequence of the toxic and harmful gas by a historical data judging method and a Grabbs criterion judging method to select target abnormal data, wherein the historical data judging method comprises the following steps: according to a mine limited space disaster gas database, based on rules and characteristics of various monitoring data, firstly determining that no emergency occurs at a monitoring point, and then judging abnormal data in a concentration monitoring time sequence of original toxic and harmful gas according to characteristics of historical data;
the Grabbs criterion discrimination method is as follows: setting the concentration monitoring time sequence of original toxic and harmful gas at a monitoring point as X = { X (t), t =1,2, …, n }, and arranging X (t) in ascending order to obtain a sequence statistic X (1) is more than or equal to X (2) is more than or equal to … is more than or equal to X (n); calculating the lower side Grabbs number
And an upper Grabbs number>
The arithmetic mean value of the whole time series of concentration monitoring of the original poisonous and harmful gas is sigma of the original poisonous and harmful gasThe integral variance of the concentration monitoring time sequence, wherein n is the number of data in the concentration monitoring time sequence of the original toxic and harmful gas; calculating the number G (n, alpha) of Grabbs criteria according to the significance level alpha; if G (1) is more than or equal to G (n, alpha), x (1) is an abnormal value; if G (n) is more than or equal to G (n, alpha), x (n) is an abnormal value; repeating the step of the Grabbs criterion judging method until all abnormal values in the concentration monitoring time sequence of the original toxic and harmful gas are judged;
processing the target abnormal data by adopting a moving average line method, so as to obtain an estimation value of each target abnormal data according to the statistical characteristics of the data in the preset range of each target abnormal data, and replacing the target abnormal data with the corresponding estimation value, wherein the moving average line method is as follows: the time sequence of monitoring the concentration of the original poisonous and harmful gas is X = { X (t), t =1,2, …, n }, and if an abnormal value occurs at a certain time t = i, a moving average line value X (i) at the time i is calculated instead of the abnormal value, and X (i) is calculated by n times before the time t = i x The mean of the data points is calculated,
s2: determining the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence, and decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term;
s3: respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension, and then respectively obtaining a predicted value of the high-frequency component, a predicted value of the low-frequency component and a predicted value of the trend term as input quantities of a wavelet neural network prediction model;
s4: and adding the predicted value of the high-frequency component, the predicted value of the low-frequency component and the predicted value of the trend term to obtain the predicted value of the concentration of the toxic and harmful gas.
2. The method of claim 1, wherein in step S2, decomposing the target toxic and harmful gas concentration sequence into a high frequency component, a low frequency component, and a trend term comprises:
adding a characteristic wave at each end of the target toxic and harmful gas concentration sequence by a wavelet neural network method to obtain a prolonged toxic and harmful gas concentration sequence;
determining local extreme points of each signal in the extended toxic and harmful gas concentration sequence, and performing cubic spline interpolation processing on each extreme point to obtain an upper envelope line, a lower envelope line and a mean envelope line of each signal;
and selecting a target signal according with the intrinsic mode function according to the upper envelope line, the lower envelope line and the mean envelope line of each signal, and decomposing the target signal into the sum of a limited intrinsic mode function and a residual quantity to obtain a high-frequency component, a low-frequency component and a trend item of the target toxic and harmful gas concentration sequence.
3. The method of claim 1, wherein prior to step S3, the method further comprises: the training step of the wavelet neural network prediction model comprises the following steps:
processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term, and respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence;
classifying samples after phase space reconstruction, determining training samples and test samples, training a wavelet neural network prediction model by the training samples, and continuously correcting parameters of the wavelet neural network prediction model according to errors of a prediction output result and an expected output result so as to continuously approximate the prediction output result of the wavelet neural network prediction model to the expected output result to obtain a trained wavelet neural network prediction model;
and inputting the test sample into the trained wavelet neural network prediction model to judge the prediction precision of the trained wavelet neural network prediction model.
4. The method of claim 3, wherein l =2 -1 log 2 And m determines the number of hidden layer neurons in the wavelet neural network prediction model, wherein l is the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence, and m is the number of hidden layer neurons.
5. The utility model provides a mine limited space calamity gas concentration intelligent prediction device which characterized in that includes:
the first preprocessing module is used for processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
the first pre-processing module comprises:
the judging module is used for judging the abnormal data in the original toxic and harmful gas concentration monitoring time sequence by a historical data judging method and a Grabbs criterion judging method to select target abnormal data, wherein the historical data judging method comprises the following steps: according to a mine limited space disaster gas database, based on rules and characteristics of various monitoring data, firstly determining that no emergency occurs at a monitoring point, and then judging abnormal data in a concentration monitoring time sequence of original toxic and harmful gas according to characteristics of historical data;
the Grabbs criterion discrimination method is as follows: setting the concentration monitoring time sequence of original toxic and harmful gas at a monitoring point as X = { X (t), t =1,2, …, n }, and arranging X (t) in ascending order to obtain a sequence statistic X (1) is more than or equal to X (2) is more than or equal to … is more than or equal to X (n); calculating the lower side Grabbs number
And an upper Grubbs number>
The arithmetic mean value of the whole concentration monitoring time sequence of the original toxic and harmful gas is provided, sigma is the variance of the whole concentration monitoring time sequence of the original toxic and harmful gas, and n is the number of data in the concentration monitoring time sequence of the original toxic and harmful gas; calculating the number G (n, alpha) of Grabbs criteria according to the significance level alpha; if G (1) is more than or equal to G (n, alpha), x (1) is an abnormal value; if G (n) is more than or equal to G (n, alpha), x (n) is an abnormal value; repeating the step of the Grabbs criterion judging method until all abnormal values in the concentration monitoring time sequence of the original toxic and harmful gas are judged; />
The first preprocessing submodule is used for processing the target abnormal data by adopting a moving average line method, so as to obtain an estimation value of each target abnormal data according to the statistical characteristics of the data in a preset range of each target abnormal data, and replace the target abnormal data with the corresponding estimation value, wherein the moving average line method is as follows: the time sequence of monitoring the concentration of the original poisonous and harmful gas is X = { X (t), t =1,2, …, n }, and if an abnormal value occurs at a certain time t = i, a moving average line value X (i) at the time i is calculated instead of the abnormal value, and X (i) is calculated by n times before the time t = i x The mean of the data points is calculated,
the first decomposition module is used for determining the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence and decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend term;
the prediction module is used for respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend term according to the reconstruction embedding dimension, and then respectively obtaining a prediction value of the high-frequency component, a prediction value of the low-frequency component and a prediction value of the trend term as input quantities of a wavelet neural network prediction model;
and the prediction result determining module is used for adding the prediction value of the high-frequency component, the prediction value of the low-frequency component and the prediction value of the trend item to obtain the concentration prediction value of the toxic and harmful gas.
6. The apparatus of claim 5, wherein the first decomposition module comprises:
the continuation module is used for adding a characteristic wave at each of two ends of the target toxic and harmful gas concentration sequence through a wavelet neural network method to obtain a prolonged toxic and harmful gas concentration sequence;
the envelope acquisition module is used for determining local extreme points of each signal in the extended toxic and harmful gas concentration sequence, and after cubic spline interpolation processing is carried out on each extreme point, an upper envelope, a lower envelope and a mean envelope of each signal are obtained;
and the first decomposition submodule is used for selecting a target signal according with the intrinsic mode function according to the upper envelope line, the lower envelope line and the mean envelope line of each signal, and decomposing the target signal into the sum of a limited number of intrinsic mode functions and a residual amount so as to obtain a high-frequency component, a low-frequency component and a trend item of the target toxic and harmful gas concentration sequence.
7. The apparatus of claim 5, further comprising:
the second preprocessing module is used for processing abnormal data in the concentration monitoring time sequence of the original toxic and harmful gas to obtain a processed target toxic and harmful gas concentration sequence;
the second decomposition module is used for decomposing the target toxic and harmful gas concentration sequence into a high-frequency component, a low-frequency component and a trend item, and respectively carrying out phase space reconstruction on the high-frequency component, the low-frequency component and the trend item according to the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence;
the model training module is used for classifying the samples after the phase space reconstruction, determining training samples and test samples, training a wavelet neural network prediction model by the training samples, and continuously correcting parameters of the wavelet neural network prediction model according to errors of a prediction output result and an expected output result and the errors so as to continuously approximate the prediction output result of the wavelet neural network prediction model to the expected output result to obtain the trained wavelet neural network prediction model;
and the model verification module is used for inputting the test sample into the trained wavelet neural network prediction model so as to judge the prediction precision of the trained wavelet neural network prediction model.
8. The device of claim 7, wherein l =2 -1 log 2 And m determines the number of hidden layer neurons in the wavelet neural network prediction model, wherein l is the reconstruction embedding dimension of the target toxic and harmful gas concentration sequence, and m is the number of hidden layer neurons.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910871668.6A CN110555570B (en) | 2019-09-16 | 2019-09-16 | Intelligent prediction method and device for gas concentration of mine limited space disaster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910871668.6A CN110555570B (en) | 2019-09-16 | 2019-09-16 | Intelligent prediction method and device for gas concentration of mine limited space disaster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110555570A CN110555570A (en) | 2019-12-10 |
| CN110555570B true CN110555570B (en) | 2023-04-18 |
Family
ID=68740345
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910871668.6A Active CN110555570B (en) | 2019-09-16 | 2019-09-16 | Intelligent prediction method and device for gas concentration of mine limited space disaster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110555570B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114358244B (en) * | 2021-12-20 | 2023-02-07 | 淮阴工学院 | Big data intelligent detection system of pressure based on thing networking |
| CN114399024B (en) * | 2021-12-20 | 2023-02-03 | 淮阴工学院 | Oil gas concentration big data intelligent detection system |
| CN116070163B (en) * | 2023-03-07 | 2023-07-11 | 深圳市特安电子有限公司 | Indoor harmful gas concentration anomaly monitoring data processing method |
| CN116109014B (en) * | 2023-04-11 | 2023-08-01 | 广东广宇科技发展有限公司 | Simulation fire-fighting evacuation method for urban rail transit large transfer station |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5155690B2 (en) * | 2008-02-25 | 2013-03-06 | 株式会社日立製作所 | Coal-fired boiler gas concentration estimation device and gas concentration estimation method |
| JP5403575B2 (en) * | 2008-01-25 | 2014-01-29 | 国立大学法人長岡技術科学大学 | Environmental load concentration change judgment device |
| CN104156422A (en) * | 2014-08-06 | 2014-11-19 | 辽宁工程技术大学 | Gas concentration real-time prediction method based on dynamic neural network |
| CN106503792B (en) * | 2016-10-25 | 2018-12-18 | 西安科技大学 | A kind of gas density prediction technique based on adaptation module neural network |
| CN107506859A (en) * | 2017-08-17 | 2017-12-22 | 国网重庆市电力公司电力科学研究院 | A kind of SF6The Forecasting Methodology and system of decomposition gas concentration-time sequence |
| CN109583566A (en) * | 2018-11-20 | 2019-04-05 | 东北农业大学 | Ammonia concentration prediction technique in a kind of cold ground pig house based on empirical mode decomposition and Elman neural network |
-
2019
- 2019-09-16 CN CN201910871668.6A patent/CN110555570B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110555570A (en) | 2019-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110555570B (en) | Intelligent prediction method and device for gas concentration of mine limited space disaster | |
| CN111562358B (en) | Transformer oil gas content prediction method and system based on combined model | |
| CN110674604B (en) | Transformer DGA data prediction method based on multi-dimensional time sequence frame convolution LSTM | |
| CN110779746B (en) | Diagnosis method for improving composite fault of deep sparse self-encoder network rotating machinery | |
| CN110766192B (en) | Drilling well leakage prediction system and method based on deep learning | |
| CN110361778B (en) | Seismic data reconstruction method based on generation countermeasure network | |
| CN110163433B (en) | Ship flow prediction method | |
| CN107292446B (en) | Hybrid wind speed prediction method based on component relevance wavelet decomposition | |
| CN113923104B (en) | Network fault diagnosis method, equipment and storage medium based on wavelet neural network | |
| WO2022121312A1 (en) | Food safety risk level prediction method and apparatus, and electronic device | |
| US11954115B2 (en) | Methods and systems for wavelet based representation | |
| CN108491953B (en) | PM2.5 prediction and early warning method and system based on nonlinear theory | |
| CN112819217A (en) | Method, system and storage medium for predicting main influence factors of mobile source pollution emission | |
| CN110879927A (en) | Sea clutter amplitude statistical distribution field modeling method for sea target detection | |
| CN113011660A (en) | Air quality prediction method, system and storage medium | |
| CN115221793A (en) | Tunnel surrounding rock deformation prediction method and device | |
| CN111310122A (en) | Model data processing method, electronic device and storage medium | |
| CN117216522A (en) | Bearing fault type diagnosis method, device and equipment based on machine learning | |
| CN111832748A (en) | Electronic nose width learning method for performing regression prediction on concentration of mixed gas | |
| CN111582132A (en) | Improved EEMD and PCNN-based gas leakage signal noise reduction method | |
| CN117035155A (en) | Water quality prediction method | |
| CN105911016A (en) | Non-linear modeling method for spectral properties of crude oil | |
| CN115883424A (en) | Method and system for predicting traffic data between high-speed backbone networks | |
| CN115829157A (en) | Chemical water quality index prediction method based on variational modal decomposition and auto former model | |
| CN115389743A (en) | Method, medium and system for predicting content interval of dissolved gas in transformer oil |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |