VMD-LSTM-ELMAN-based open-air area radon exhalation rate prediction method

Through the VMD-LSTM-ELMAN method, the historical data of radon precipitation rate is decomposed into different characteristic components and combined with environmental factors, the problem of poor prediction accuracy of radon precipitation rate in the existing technology is solved, and more accurate prediction of radon precipitation rate and trend analysis is achieved. , provides effective guidance for radon prevention and control.

CN119004249AActive Publication Date: 2024-11-22SHENZHEN UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410887826.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-03
Publication Date
2024-11-22
Estimated Expiration
2044-07-03

AI Technical Summary

Technical Problem

The existing radon precipitation rate prediction method fails to effectively consider the hysteresis effect of environmental indicators on radon precipitation rate, resulting in poor accuracy of prediction results, difficult to accurately reflect the fluctuation trend of radon precipitation rate, and fail to effectively guide radon prevention and control measures.

Method used

Using the VMD-LSTM-ELMAN method, the historical data of radon precipitation rate is decomposed into three components with different physical significance, and the prediction is performed using the LSTM model, ELMAN neural network model and normal distribution model, and the external environmental factors are considered. The hysteresis effect improves prediction accuracy through modal decomposition and neural network prediction.

Benefits of technology

It improves the accuracy and stability of radon precipitation rate prediction, can better reflect the rules and trends of radon precipitation, and is suitable for radon precipitation prediction in open-air places, providing effective guidance for radon prevention and control and environmental governance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004249A_ABST
    Figure CN119004249A_ABST
Patent Text Reader

Abstract

The invention provides an open-air area radon exhalation rate prediction method based on VMD-LSTM-ELMAN, and the method comprises the steps: decomposing the historical data of the radon exhalation rate into three groups of components with different feature information through VMD, and carrying out the data training and prediction through selecting an LSTM model, an ELMAN neural network model and a normal distribution model, and the environmental factors influencing radon precipitation are used as prediction indexes to be input into the model, so that the influence of the environmental indexes on radon precipitation can be reflected more truly, the model prediction result can accurately reflect the precipitation rule of radon precipitation, and the prediction precision is high. According to the method, the VMD, the LSTM and the ELMAN are jointly adopted to establish the radon exhalation rate prediction model, the hysteresis effect of environmental indexes on the radon exhalation rate influence is particularly considered, the radon exhalation rate numerical curve is decomposed, the change trend of the radon exhalation rate numerical curve is predicted, and guidance is provided for radon prevention and control and environmental governance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of radon exhalation monitoring, and in particular to a method for predicting radon exhalation rate in an open-air area based on VMD-LSTM-ELMAN. Background Art

[0002] Radon is a colorless, odorless, radioactive inert gas, which is listed as the second leading cause of lung cancer by the World Health Organization. It is often precipitated in underground projects, mines, tourist caves, nuclear facilities, tailings ponds, and on the surfaces of building materials, soil, and the ground. The diffusion and migration of radon will cause huge radioactive pollution to the surrounding environment and seriously affect human life and health. Therefore, the prediction and early warning of radon are of great significance.

[0003] Many scholars have conducted research on radon prediction. Khaled F. Al-Shboul et al. used global soil radionuclide levels and Rn-ER datasets to train and test models, and compared the qualitative results of descriptive statistics with artificial neural network (ANN) and gene expression programming (GEP) models to estimate radon exhalation rates based on the natural radionuclide content in the soil. They found that the ANN accuracy and precision were higher than the best performing GEP model. Adil Aslam Mir et al. used different ensemble and individual machine learning methods to predict radon data anomalies caused by seismic activity for real-time radon time series data in different scenarios, and found that boosted trees and radial kernel support vector machines were proven to be better regression models for predicting soil radon concentration anomalies during seismic activity. Adil Aslam Mir et al. compared different methods including mean, median, mode, predicted mean matching (PMM) and hot-deck interpolation when processing soil radon gas concentration (SRGC) time series data, and found that the IBFI estimation data set provided a more suitable model, and the model generation and prediction based on the IBFI estimation time series would produce more accurate predictions. Hao Duong Van et al. conducted a study to optimize the machine learning model, namely artificial neural network (ANN), to accurately predict radon diffusion in Vietnam and concluded that ANN performed very well to predict radon diffusion with low performance index values. Tran Dinh Bao et al. designed a simple single hidden layer artificial neural network (ANN) model to predict the release of radon gas in radioactive mines during ore dressing and mining. Compared with the two hidden layer ANN, random forest and support vector machine models applied to Sin Quyen copper deposit, it has a lower computational cost and can be trained, referenced and obtains the best results.

[0004] However, the prediction of radon exhalation in the above prediction method does not take into account the lag effect of environmental indicators on the radon exhalation rate, ignores external factors, has large errors, and is difficult to effectively reflect the fluctuation trend of the radon exhalation rate. The prediction results are of poor accuracy and it is difficult to guide management units to effectively prevent and control radon.

[0005] In view of this, it is necessary to design an improved VMD-LSTM-ELMAN-based method for predicting radon exhalation rate in open-air areas to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to provide a method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN. VMD is used to decompose historical data of radon exhalation rate into three components with different physical meanings. LSTM model, ELMAN neural network model and normal distribution model are selected for data training and prediction, so as to accurately reflect the law of radon exhalation and ensure the prediction accuracy, so as to make it better applied to the actual radon exhalation prediction work.

[0007] To achieve the above-mentioned purpose of the invention, the present invention provides a method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN, comprising the following steps:

[0008] S1. Obtain the historical radon exhalation rate monitoring data of the area to be predicted to form the original data set of the time series;

[0009] S2, normalizing the original data set in step S1, and performing modal decomposition using VMD to decompose it into three groups of modal components with different characteristic information;

[0010] S3. Divide the three groups of modal components in step S2 according to the characteristic information, and select one of the LSTM model, ELMAN neural network model or normal distribution model for data training and prediction respectively; finally, add the three groups of predicted values ​​obtained in time series, and then perform data denormalization to obtain the predicted value of radon exhalation rate in the area to be predicted.

[0011] As a further improvement of the present invention, in step S3, the model selection principle for predicting the three groups of modal components is as follows: the imf1 component with nonlinear characteristics and non-periodicity is predicted by the LSTM model, the imf2 component with fluctuation characteristics and periodicity is predicted by the ELMAN neural network model, and the imf3 component randomly distributed around 0 and with poor regularity is predicted by the normal distribution model.

[0012] As a further improvement of the present invention, when the imf2 component is predicted using the ELMAN neural network model, external environmental factors that affect the radon exhalation rate are simultaneously input as prediction indicators.

[0013] As a further improvement of the present invention, in step S2, when VMD is used for modal decomposition, a variational model is first constructed, and then an augmented Lagrangian function is introduced to transform the constrained variational problem into an unconstrained variational problem for solution, and an alternating direction multiplier algorithm is used for iterative update until the convergence condition is met and the modal components are output.

[0014] As a further improvement of the present invention, the process of using the LSTM model to predict the imf1 component is as follows:

[0015] Input gate: remember the current state information and calculate the input gate i t The value of the candidate state value of the input cell at time t

[0016] i t =σ·(W i ·[h t-1 ,x t ]+b i ) (1)

[0017]

[0018] Where: W i ,W c represents the corresponding weight, b i With b c represents the corresponding bias;

[0019] Forget gate: controls the forgotten information and calculates the activation value f of the forget gate at time t t :

[0020] f t =σ·(W f ·[h t-1 ,x t ]+b f ) (3)

[0021] Where: W f ,b f Respectively represent the weight and bias of the forget gate, σ represents the relu function;

[0022] Update cell state: According to the calculation results of the input gate and the forget gate, the cell state update value C at time t is obtained t :

[0023]

[0024] Output gate: controls the output information and updates the value C according to the cell state t Get the output gate o t The calculation formula is:

[0025] o t =σ·(W o ·[h t-1 ,x t ]+b o ) (5)

[0026] h t =o t *tanh(C t ) (6)

[0027] Where: W o and b o represents the weight and bias of the output gate, h t It is the final output result at the current moment, and the output results constitute the predicted value y1 sequence.

[0028] As a further improvement of the present invention, when the ELMAN neural network model is used to predict the imf2 component, the mathematical model of the ELMAN neural network is:

[0029]

[0030] Where: k is the current state, Y(k), X(k), u(k-1), X c (k) are the m-dimensional output node vector, the n-dimensional intermediate layer node unit vector, the r-dimensional input vector and the n-dimensional feedback state vector, ω3, ω2, ω1 are the connection weights from the hidden layer to the output layer, from the input layer to the hidden layer, and from the connection layer to the hidden layer, respectively. i is the threshold of the hidden layer and the output layer, G(·) is the activation function of the output neuron, using the tan-sigmoid function, and F(·) is the transfer function of the hidden layer neuron, using the sigmoid function;

[0031] The ELMAN neural network uses the error square sum function as the learning indicator function.

[0032]

[0033] Where: is the target output vector, which constitutes the predicted value y2 sequence.

[0034] As a further improvement of the present invention, when the normal distribution model is used to predict the imf3 component, the normal distribution type test is first performed on the imf3 component, and the mean μ and standard deviation σ of the imf3 component are calculated. Then, the probability density function formula of the normal distribution is used to calculate the probability density function value f(x) of the given imf3 component data point x, and these probability density function values ​​are used to generate a set of prediction value y3 sequences:

[0035]

[0036] Where: f(x) is the probability density at x, μ is the mean, and σ is the standard deviation.

[0037] As a further improvement of the present invention, the process of modal decomposition using VMD is as follows:

[0038]

[0039] Where: x is the original data set after normalization; K is the number of intrinsic mode decompositions3; δ(t) is the impulse function; * is the convolution symbol; {ω k}={ω1,…,ω K} is the center frequency of the modal function; {u k}={u1,…,u K} is the decomposed modal component;

[0040] The augmented Lagrangian function is introduced to transform the above constrained variational problem into an unconstrained variational problem for solution, and then equation (11) is transformed as follows:

[0041]

[0042] Where: α is the quadratic penalty factor; λ is the Lagrange multiplier;

[0043] Use the multiplier alternating direction algorithm to iteratively update u k ,ω k and λ:

[0044]

[0045]

[0046]

[0047] Where: Γ is the update factor;

[0048] Until the convergence condition (15) is met, the final modal component is output;

[0049]

[0050] Where: e>0 is the discrimination accuracy.

[0051] As a further improvement of the present invention, the prediction indicators include cumulative rainfall, daily average temperature and daily temperature difference. Considering that the impact of the prediction indicator values ​​on radon precipitation is lagging, the environmental data of the tth day is input into the model as the prediction indicator of the t+1th day.

[0052] The beneficial effects of the present invention are:

[0053] 1. The radon exhalation rate prediction method for open-air areas based on VMD-LSTM-ELMAN of the present invention first uses VMD to decompose the radon exhalation rate historical data into three components with different physical meanings, which not only suppresses the noise component in the signal and avoids the modal aliasing phenomenon, but also separates the main features in the radon exhalation process well, which is helpful for subsequent targeted prediction and can effectively improve the accuracy of the prediction; then, by selecting LSTM model, ELMAN neural network model and normal distribution model for data training and prediction, the exhalation law of radon exhalation is accurately reflected, and better prediction accuracy is guaranteed. The present invention establishes a radon exhalation rate prediction model by jointly adopting VMD, LSTM and ELMAN, especially considering the hysteresis effect of the influence of environmental indicators on the radon exhalation rate, decomposing the radon exhalation rate numerical curve and predicting its changing trend, which is suitable for predicting radon exhalation in open-air places, and provides guidance for radon prevention and control and environmental governance.

[0054] 2. When the present invention predicts the three components with different characteristic information, the trend and non-periodic components are predicted by the long short-term memory neural network model (LSTM), which can accurately capture the long-term and short-term dependencies in the sequence, which is conducive to predicting the long-term radon exhalation trend; the components with fluctuation characteristics and periodicity are predicted by the ELMAN neural network model, while considering the hysteresis effect of the environmental indicators on the radon exhalation rate. The ELMAN neural network can better capture the complex nonlinear relationship between the data due to its multi-layer structure, nonlinear activation function and feedback connection, and can more closely reflect the impact of environmental indicators on radon exhalation; the random fluctuation component without obvious trend is used as the noise part, and the frequency conforms to the normal distribution type. The normal distribution model is used for prediction, which reduces the error of the prediction model. Compared with a single model, the prediction result of the VMD-LSTM-ELMAN composite model proposed by the present invention is closer to the actual radon exhalation situation, which not only reflects the overall trend and nonlinear characteristics of the radon exhalation rate, but also fully considers randomness and uncertainty, and improves the stability of the prediction model by reducing the discreteness and fluctuation range of the model error. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN of the present invention.

[0056] Figure 2 This is a graph showing the radon exhalation rate monitoring data in Example 1.

[0057] Figure 3 These are three groups of modal component data diagrams after VMD modal decomposition in Example 1.

[0058] Figure 4This is a comparison chart of the predicted value y1 sequence in Example 1 and the radon exhalation rate data imf1 component.

[0059] Figure 5 This is a comparison chart of the predicted value y2 sequence in Example 1 and the radon exhalation rate data imf2 component.

[0060] Figure 6 This is a graph showing the normal distribution type, mean value and standard deviation data of the imf3 component in Example 1.

[0061] Figure 7 This is a comparison chart of the predicted value y3 sequence in Example 1 and the radon exhalation rate data imf3 component.

[0062] Figure 8 This is a comparison chart of the predicted values ​​of radon exhalation rate using different models for Example 1 and Comparative Examples 1-2 and the actual 35-day monitoring values. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0064] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the scheme of the present invention are shown in the drawings, while other details that are not closely related to the present invention are omitted.

[0065] In addition, it should be noted that the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus.

[0066] See also Figure 1 As shown, a method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN includes the following steps:

[0067] S1. Obtain the historical radon exhalation rate monitoring data of the area to be predicted to form the original data set of the time series;

[0068] S2, normalizing the original data set in step S1, and performing modal decomposition using VMD to decompose it into three groups of modal components with different characteristic information;

[0069] S3. Divide the three groups of modal components in step S2 according to the characteristic information, and select one of the LSTM model, ELMAN neural network model or normal distribution model for data training and prediction respectively; finally, add the three groups of predicted values ​​obtained in time series, and then perform data denormalization to obtain the predicted value of radon exhalation rate in the area to be predicted.

[0070] Specifically, in step S3, the model selection principle for the prediction of the three groups of modal components is: the imf1 component with trend, nonlinear characteristics and non-periodicity is predicted by the LSTM model, the imf2 component with fluctuation characteristics and periodicity is predicted by the ELMAN neural network model, and the imf3 component randomly distributed around 0 with poor regularity is predicted by the normal distribution model. Those skilled in the art should know that imf1, imf2 and imf3 are only component names, and when actually selecting a prediction model, it should be selected according to the characteristic information of the component.

[0071] The process of using the LSTM model to predict the imf1 component is as follows:

[0072] Input gate: remember the current state information and calculate the input gate i t The value of the candidate state value of the input cell at time t

[0073] i t =σ·(W i ·[h t-1 ,x t ]+b i ) (1)

[0074]

[0075] Where: W i ,W c represents the corresponding weight, b i With b c represents the corresponding bias;

[0076] Forget gate: controls the forgotten information and calculates the activation value f of the forget gate at time t t :

[0077] f t =σ·(W f ·[h t-1 ,x t ]+b f ) (3)

[0078] Where: W f ,b f Respectively represent the weight and bias of the forget gate, σ represents the relu function;

[0079] Update cell state: According to the calculation results of the input gate and the forget gate, the cell state update value C at time t is obtained t :

[0080]

[0081] Output gate: controls the output information and updates the value C according to the cell state t Get the output gate o t The calculation formula is:

[0082] o t =σ·(W o ·[h t-1 ,x t ]+b o ) (5)

[0083] h t =o t *tanh(C t ) (6)

[0084] Where: W o and b o represents the weight and bias of the output gate, h t It is the final output result at the current moment, and the output results constitute the predicted value y1 sequence.

[0085] The long short-term memory neural network model LSTM is a special type of recurrent neural network that solves the gradient vanishing and exploding problems in long-term time series analysis and is suitable for processing and predicting important events with very long intervals and delays in time series. The present invention uses the LSTM model to predict trend components, which can accurately capture the long-term and short-term dependencies in the sequence, and is conducive to predicting the long-term radon exudation trend.

[0086] When the imf2 component is predicted using the ELMAN neural network model, external environmental factors that affect the radon exhalation rate are input as prediction indicators. The mathematical model of the ELMAN neural network is:

[0087]

[0088] Where: k is the current state, Y(k), X(k), u(k-1), X c (k) are the m-dimensional output node vector, the n-dimensional intermediate layer node unit vector, the r-dimensional input vector and the n-dimensional feedback state vector, ω3, ω2, ω1 are the connection weights from the hidden layer to the output layer, from the input layer to the hidden layer, and from the connection layer to the hidden layer, respectively. iis the threshold of the hidden layer and the output layer, G(·) is the activation function of the output neuron, using the tan-sigmoid function, and F(·) is the transfer function of the hidden layer neuron, using the sigmoid function;

[0089] The ELMAN neural network uses the error square sum function as the learning indicator function.

[0090]

[0091] Where: is the target output vector, which constitutes the predicted value y2 sequence.

[0092] ELMAN neural network is a typical global feedforward and local recursive network model, which has stronger computing power than feedforward neural network and can better solve the problem of rapid optimization. The present invention uses ELMAN neural network model to predict components with fluctuation characteristics and periodicity, and considers the hysteresis effect of environmental indicators on radon exhalation rate. ELMAN neural network can better capture the complex nonlinear relationship between data due to its multi-layer structure, nonlinear activation function and feedback connection, and can more realistically reflect the influence of environmental indicators on radon exhalation.

[0093] When using the normal distribution model to predict the IMF3 component, first perform a normal distribution type test on the IMF3 component, calculate the mean μ and standard deviation σ of the IMF3 component, and then use the probability density function formula of the normal distribution to calculate the probability density function value f(x) of the given IMF3 component data point x. Use these probability density function values ​​to generate a set of predicted value y3 sequences:

[0094]

[0095] Where: f(x) is the probability density at x, μ is the mean, and σ is the standard deviation.

[0096] The imf3 component that randomly fluctuates around 0 without an obvious trend is taken as the noise part, and its frequency conforms to the normal distribution type. The normal distribution model is used for prediction, which reduces the error of the prediction model.

[0097] When the present invention adopts VMD for modal decomposition, a variational model is first constructed, and then an augmented Lagrangian function is introduced to transform the constrained variational problem into an unconstrained variational problem for solution, and an alternating direction multiplier algorithm is used for iterative update until the convergence condition is satisfied, and the modal components are output. Variational mode decomposition (VMD) is an adaptive, completely non-recursive modal variation and signal processing method, which overcomes the problems of endpoint effect and modal component aliasing in empirical mode decomposition (EMD), and can reduce the non-stationarity of time series with high complexity and strong nonlinearity, and decompose to obtain subsequences containing multiple different frequency scales and relatively stable. The present invention uses VMD to decompose the historical data of radon exhalation rate into three components with different physical meanings, which not only suppresses the noise component in the signal and avoids the modal aliasing phenomenon, but also separates the main features in the radon exhalation process well, which is helpful for subsequent targeted prediction and can effectively improve the accuracy of prediction.

[0098] The process of modal decomposition using VMD is as follows:

[0099]

[0100] Where: x is the original data set after normalization; K is the number of intrinsic mode decompositions3; δ(t) is the impulse function; * is the convolution symbol; {ω k}={ω1,…,ω K} is the center frequency of the modal function; {u k}={u1,…,u K} is the decomposed modal component;

[0101] The augmented Lagrangian function is introduced to transform the above constrained variational problem into an unconstrained variational problem for solution, and then equation (11) is transformed as follows:

[0102]

[0103] Where: α is the quadratic penalty factor; λ is the Lagrange multiplier;

[0104] Use the multiplier alternating direction algorithm to iteratively update u k ,ω k and λ:

[0105]

[0106]

[0107]

[0108] Where: Γ is the update factor;

[0109] Until the convergence condition (15) is met, the final modal component is output;

[0110]

[0111] Where: e>0 is the discrimination accuracy.

[0112] More specifically, when the ELMAN neural network model is used to predict the mf2 component, the prediction indicators include cumulative rainfall, daily average temperature and daily temperature difference. Considering that the impact of the prediction indicator values ​​on radon exudation is lagging, the environmental data on the tth day is input into the model as the prediction indicator for the t+1th day. The prediction indicators are specifically obtained by calculating the mutual information value to analyze the correlation between variables and obtain the importance ranking of each characteristic variable. The mutual information value calculation formula is:

[0113]

[0114] Where: p(a) and p(b) are the probability distributions of variables a and b respectively, and p(a,b) is the joint probability distribution of variables a and b.

[0115] It should be noted that the cumulative rainfall, daily average temperature and daily temperature difference are all affected by the weather, and the weather forecast time in most areas is up to 40 days. Therefore, in order to ensure the accuracy of the prediction, the best prediction time using this method does not exceed 40 days, and the radon exhalation rate in the prediction area within the next 1 to 40 days can be predicted.

[0116] The radon exhalation rate prediction method for open-air areas based on VMD-LSTM-ELMAN of the present invention establishes a radon exhalation rate prediction model by jointly adopting VMD, LSTM, and ELMAN, and particularly considers the hysteresis effect of the environmental indicators on the radon exhalation rate, decomposes the radon exhalation rate numerical curve and predicts its change trend, which is suitable for predicting radon exhalation in open-air places, and provides guidance for radon prevention and control and environmental management. In practical applications, the prediction of radon exhalation by this method can not only guide radon prevention and control, such as sprinkling water to reduce radon in the upcoming hot weather, but also can judge whether structural damage occurs in the monitoring area according to the deviation between the predicted value and the actual radon exhalation rate, and can also be used for early warning of structural instability accidents.

[0117] Example 1

[0118] This embodiment provides a VMD-LSTM-ELMAN method for predicting radon exhalation rate in open-air areas; specifically, the following steps are included:

[0119] S1. The radon exhalation rate monitoring results of a certain open-pit uranium tailings pond beach model for 179 consecutive days were obtained from the Hunan Uranium Tailings Pond Decommissioning Technology Engineering Research Center. The obtained radon exhalation rate monitoring results are shown in Table 1. Figure 2 shown.

[0120] Table 1 Radon exhalation rate results of a beach model of an open-pit uranium tailings pond for 179 consecutive days

[0121]

[0122]

[0123] S2, the radon exhalation rate result of step S1 is used as the original data and normalized, wherein the first 144 days of data are used as training data, and the last 35 days of data are used as test data; first, VMD is used to perform modal decomposition and decompose it into three groups of modal components with different characteristic information, such as Figure 3 As shown;

[0124] The process of modal decomposition using VMD is as follows: set the model parameters: quadratic penalty factor α = 1500, discrimination accuracy e = 10 -8 , the number of modes K = 3;

[0125]

[0126] Where: x is the original data set after normalization; K is the number of intrinsic mode decompositions3; δ(t) is the impulse function; * is the convolution symbol; {ω k}={ω1,…,ω K} is the center frequency of the modal function; {u k}={u1,…,u K} is the decomposed modal component;

[0127] By introducing the augmented Lagrangian function, the above constrained variational problem is transformed into an unconstrained variational problem for solution, and then equation (2) is transformed as follows:

[0128]

[0129] Where: α is the quadratic penalty factor; λ is the Lagrange multiplier;

[0130] Use the multiplier alternating direction algorithm to iteratively update u k ,ω k and λ:

[0131]

[0132]

[0133]

[0134] Where: Γ is the update factor;

[0135] Until the convergence condition (6) is met, the final modal component is output;

[0136]

[0137] Where: e>0 is the discrimination accuracy.

[0138] S3. Divide the three groups of modal components in step S2 according to the characteristic information, and select one of the LSTM model, ELMAN neural network model or normal distribution model for data training (the first 144 days of data are used as training) and prediction (the last 35 days of data are used as prediction); finally, add the three groups of predicted values ​​in time series, and then perform data denormalization to obtain the predicted value of radon exhalation rate in the area to be predicted.

[0139] Among them, the process of using the LSTM model to predict the imf1 component is as follows: the parameters are set as follows: the initial learning rate is 0.005, and the maximum number of iterations is 500;

[0140] Input gate: remember the current state information and calculate the input gate i t The value of the candidate state value of the input cell at time t

[0141] i t =σ·(W i ·[h t-1 ,x t ]+b i ) (7)

[0142]

[0143] Where: W i ,W c represents the corresponding weight, b i With b c represents the corresponding bias;

[0144] Forget gate: controls the forgotten information and calculates the activation value f of the forget gate at time t t :

[0145] f t =σ·(W f ·[h t-1 ,x t ]+b f ) (9)

[0146] Where: W f ,b f Respectively represent the weight and bias of the forget gate, σ represents the relu function;

[0147] Update cell state: According to the calculation results of the input gate and the forget gate, the cell state update value C at time t is obtained t:

[0148]

[0149] Output gate: controls the output information and updates the value C according to the cell state t Get the output gate o t The calculation formula is:

[0150] o t =σ·(W o ·[h t-1 ,x t ]+b o ) (11)

[0151] h t =o t *tanh(C t ) (12)

[0152] Where: W o and b o represents the weight and bias of the output gate, h t is the final output result at the current moment, and the output results constitute the predicted value y1 sequence, such as Figure 4 As shown, it is a comparison chart of the predicted value y1 series and the imf1 component of the radon exhalation rate data.

[0153] When the ELMAN neural network model is used to predict the imf2 component, the external environmental factors that affect the radon exhalation rate are also input as prediction indicators.

[0154] The mutual information value is calculated to analyze the correlation between variables and obtain the importance ranking of each feature variable. The mutual information value calculation formula is:

[0155]

[0156] Where: p(a) and p(b) are the probability distributions of variables a and b respectively, and p(a,b) is the joint probability distribution of variables a and b.

[0157] It can be seen from Table 2 that the characteristic importance of cumulative rainfall, daily average temperature, and daily temperature difference is greater than 0.5. Finally, the characteristic variables of cumulative rainfall, daily average temperature, and daily temperature difference are input as prediction indicators.

[0158] Table 2 Importance of characteristic variables affecting radon exhalation

[0159]

[0160]

[0161] The mathematical model of the ELMAN neural network is as described in formula (14). The model parameter settings are: the maximum number of training times is 1000, the learning rate is 0.01, and the minimum error is 1e -6 ,

[0162]

[0163] Where: k is the current state, Y(k), X(k), u(k-1), X c (k) are the m-dimensional output node vector, the n-dimensional intermediate layer node unit vector, the r-dimensional input vector and the n-dimensional feedback state vector, ω3, ω2, ω1 are the connection weights from the hidden layer to the output layer, from the input layer to the hidden layer, and from the connection layer to the hidden layer, respectively. i is the threshold of the hidden layer and the output layer, G(·) is the activation function of the output neuron, using the tan-sigmoid function, and F(·) is the transfer function of the hidden layer neuron, using the sigmoid function;

[0164] The ELMAN neural network uses the error square sum function as the learning indicator function.

[0165]

[0166] Where: is the target output vector, forming the predicted value y2 sequence, such as Figure 5 As shown, it is a comparison chart of the predicted value y2 series and the imf2 component of the radon exhalation rate data.

[0167] When using the normal distribution model to predict the imf3 component, first perform a normal distribution type test on the imf3 component, and calculate the mean μ and standard deviation σ of the imf3 component to obtain Figure 6 ; Then, the probability density function formula (16) of the normal distribution is used to calculate the probability density function value f(x) of the given imf3 component data point x, and these probability density function values ​​are used to generate a set of prediction value y3 sequences,

[0168]

[0169] Where: f(x) is the probability density at x, μ is the mean, and σ is the standard deviation.

[0170] like Figure 7 As shown, it is a comparison chart of the predicted value y3 series and the imf3 component of the radon exhalation rate data.

[0171] Finally, all component prediction values ​​are added together to obtain the sequence y=y1+y2+y3 and denormalized to obtain the final radon exhalation rate prediction value.

[0172] Comparative Example 1

[0173] Comparative Example 1 provides a method for predicting radon exhalation rate. Compared with Example 1, the difference is that step S2 is not performed, and in step S3, only the LSTM model is used for prediction. The rest is roughly the same as Example 1 and will not be repeated here.

[0174] Comparative Example 2

[0175] Comparative Example 2 provides a method for predicting radon exhalation rate. Compared with Example 1, the difference is that step S2 is not performed, and in step S3, only the ELMAN model is used for prediction. The rest is roughly the same as Example 1 and will not be repeated here.

[0176] The data of radon exhalation rate prediction using different models in Example 1 and Comparative Examples 1-2 were compared with the actual data for the last 35 days. The results are as follows: Figure 8 As shown in Table 3, the MSE (mean square error), MAPE (mean absolute percentage error) and R2 (the proportion of variance explained by the model to the total variance) of each model are calculated to analyze the modeling performance of different models.

[0177] Table 3 Prediction accuracy of different models of Example 1 and Comparative Examples 1-2

[0178] MSE MAPE R2 LSTM 0.1120 44.49% -0.3755 ELMAN 0.0996 38.73% -0.2238 VMD-LSTM-ELMAN 0.0044 7.09% 0.9456

[0179] From Table 3 and Figure 8 It can be seen that although the LSTM model has a good prediction effect on the precipitation trend of radon exhalation rate, it cannot predict the fluctuation position, ignores the influence of external factors, and has a large error; although the ELMAN model takes into account the influence of external factors, it does not take into account the precipitation trend of radon and random errors, so it only predicts that the fluctuation amplitude is consistent with the actual situation, but the numerical value is quite different; the VMD-LSTM-ELMAN model adopted in Example 1 is generally the best, not only can it make a better prediction of the radon exhalation trend, but also takes into account the fluctuation caused by the influence of external factors, while ensuring that the prediction error is relatively small.

[0180] In summary, the present invention provides a method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN. First, VMD is used to decompose the historical data of radon exhalation rate into three components with different physical meanings, which not only suppresses the noise components in the signal, but also separates the main features in the radon exhalation process, and can effectively improve the accuracy of the prediction; then, the LSTM model, ELMAN neural network model and normal distribution model are selected for data training and prediction, which accurately reflects the exhalation law of radon exhalation and ensures better prediction accuracy. The present invention establishes a radon exhalation rate prediction model by jointly adopting VMD, LSTM and ELMAN, especially considering the hysteresis effect of the influence of environmental indicators on the radon exhalation rate, decomposing the radon exhalation rate numerical curve and predicting its changing trend, which is suitable for predicting radon exhalation in open-air places, and provides guidance for radon prevention and control and environmental governance.

[0181] The above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN, characterized in that: The following steps are involved: S1. Obtain the historical radon exhalation rate monitoring data of the area to be predicted to form the original data set of the time series; S2, normalizing the original data set in step S1, and performing modal decomposition using VMD to decompose it into three groups of modal components with different characteristic information; S3. Divide the three groups of modal components in step S2 according to the characteristic information, and select one of the LSTM model, ELMAN neural network model or normal distribution model for data training and prediction respectively; finally, add the three groups of predicted values ​​obtained in time series, and then perform data denormalization to obtain the predicted value of radon exhalation rate in the area to be predicted.

2. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 1, characterized in that: In step S3, the model selection principle for predicting the three groups of modal components is as follows: the imf1 component with nonlinear characteristics and non-periodicity is predicted by the LSTM model, the imf2 component with fluctuation characteristics and periodicity is predicted by the ELMAN neural network model, and the imf3 component randomly distributed around 0 with poor regularity is predicted by the normal distribution model.

3. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 2, characterized in that: When the ELMAN neural network model is used to predict the imf2 component, the external environmental factors that affect the radon exhalation rate are also input as prediction indicators.

4. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 1, characterized in that: In step S2, when VMD is used for modal decomposition, a variational model is first constructed, and then the augmented Lagrangian function is introduced to transform the constrained variational problem into an unconstrained variational problem for solution, and the multiplier alternating direction algorithm is used for iterative update until the convergence condition is met and the modal components are output.

5. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 2, characterized in that: The process of using the LSTM model to predict the imf1 component is as follows: Input gate: remember the current state information and calculate the input gate i t The value of the candidate state value of the input cell at time t i t =σ·(W i ·[h t-1 ,x t ]+b i ) (1) Where: W i ,W c represents the corresponding weight, b i With b c represents the corresponding bias; Forget gate: controls the forgotten information and calculates the activation value f of the forget gate at time t t : f t =σ·(W f ·[h t-1 ,x t ]+b f ) (3) Where: W f ,b f Respectively represent the weight and bias of the forget gate, σ represents the relu function; Update cell state: According to the calculation results of the input gate and the forget gate, the cell state update value C at time t is obtained t : Output gate: controls the output information and updates the value C according to the cell state t Get the output gate o t The calculation formula is: the t =σ·(W o ·[h t-1 ,x t ]+b o ) (5) h t =o t *tanh(C t ) (6) Where: W o and b o represents the weight and bias of the output gate, h t It is the final output result at the current moment, and the output results constitute the predicted value y1 sequence.

6. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 3, characterized in that: When the ELMAN neural network model is used to predict the imf2 component, the mathematical model of the ELMAN neural network is: Where: k is the current state, Y(k), X(k), u(k-1), X c (k) are the m-dimensional output node vector, the n-dimensional intermediate layer node unit vector, the r-dimensional input vector and the n-dimensional feedback state vector, ω3, ω2, ω1 are the connection weights from the hidden layer to the output layer, from the input layer to the hidden layer, and from the connection layer to the hidden layer, respectively. i is the threshold of the hidden layer and the output layer, G(·) is the activation function of the output neuron, using the tan-sigmoid function, and F(·) is the transfer function of the hidden layer neuron, using the sigmoid function; The ELMAN neural network uses the error square sum function as the learning indicator function. Where: is the target output vector, which constitutes the predicted value y2 sequence.

7. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 2, characterized in that: When using the normal distribution model to predict the IMF3 component, first perform a normal distribution type test on the IMF3 component, calculate the mean μ and standard deviation σ of the IMF3 component, and then use the probability density function formula of the normal distribution to calculate the probability density function value f(x) of the given IMF3 component data point x. Use these probability density function values ​​to generate a set of predicted value y3 sequences: Where: f(x) is the probability density at x, μ is the mean, and σ is the standard deviation.

8. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 4, characterized in that: The process of modal decomposition using VMD is as follows: Where: x is the original data set after normalization; K is the number of intrinsic mode decompositions3; δ(t) is the impulse function; * is the convolution symbol; {ω k }={ω1,…,ω K } is the center frequency of the modal function; {u k }={u1,…,u K } is the modal component after decomposition; The augmented Lagrangian function is introduced to transform the above constrained variational problem into an unconstrained variational problem for solution, and then equation (11) is transformed as follows: Where: α is the quadratic penalty factor; λ is the Lagrange multiplier; Use the multiplier alternating direction algorithm to iteratively update u k ,ω k and λ: Where: Γ is the update factor; Until the convergence condition (15) is met, the final modal component is output; Where: e>0 is the discrimination accuracy.

9. The method for predicting radon exhalation rate in open-air areas based on VMD-LSTM-ELMAN according to claim 3, characterized in that: The prediction indicators include cumulative rainfall, daily average temperature and daily temperature difference. Considering that the impact of the prediction indicator values ​​on radon release is lagging, the environmental data on the tth day is input into the model as the prediction indicator for the t+1th day.

Citation Information

Patent Citations

  • Nonlinear weighted combination wind power prediction method based on deep belief network

    CN112434848A

  • Roof photovoltaic power prediction method based on VMD-BILSTM neural network fusion attention mechanism

    CN115630726A

  • Single-ended fault positioning method and system for high-voltage direct-current transmission line of hybrid network

    US20220196720A1

  • LSTM-SVR subway station temperature prediction method based on characteristic of multiple periods

    WO2024077969A1