Radon Stability Zone Division and Prediction and Early Warning Methods in Uranium Tailings Ponds

Abstract

The present invention discloses a method for dividing and predicting radon stable zones in uranium tailings ponds. The method uses a combination of the range method and the variance method to divide the radon concentration data of the tailings pond into a radon concentration stable zone, an asymptotically stable zone, and an unstable zone. The radon concentration in the stable zone is then predicted based on the WOA-BP neural network algorithm, and the warning threshold is determined by the maximum value sequence and the minimum value sequence to establish an alarm relationship model. Through the above method, the present invention can effectively determine the radon concentration stable zone by combining the range method and the variance method. Compared with the direct prediction method, the prediction accuracy and warning scientificity are effectively improved. At the same time, the present invention adopts the WOA-BP algorithm that combines the advantages of the Whale Optimization Algorithm algorithm and the Back Propagation algorithm to ensure the accuracy and efficiency of the prediction results.

Description

Technical Field

[0001] The invention relates to the technical field of uranium tailings pond early warning, and in particular to a radon stability zone division and prediction and early warning method for a uranium tailings pond. Background Art

[0002] In the nuclear power sector, large quantities of uranium tailings generated during nuclear power development are centrally stored, forming uranium tailings ponds. The safety of uranium tailings ponds has always been a crucial aspect of nuclear power development. However, a uranium tailings pond failure could result in the release of large amounts of radioactive contaminants, significantly impacting the surrounding ecological environment, particularly water and soil.

[0003] Radon is a radioactive gas produced by the decay of the radioactive element Ra (radium) deep underground. It is chemically stable and highly mobile, capable of migrating upward through voids in the soil. Radon anomalies can be used to indicate poor underground containment, the development of fissures, and rock fragmentation. Due to its unique geophysical and chemical properties, scholars both domestically and internationally have begun applying radon detection technology to engineering projects to address practical technical challenges. Radon has proven effective in predicting critical landslide points in landslide areas and parameters of underground rock fracture zones.

[0004] Currently, most methods for predicting radon concentrations in tailings ponds rely on direct prediction, which leaves much to be desired in terms of accuracy and scientific validity. Radon concentrations in tailings ponds exhibit a consistent downward trend over time, with significant fluctuations early on, which gradually decrease and stabilize over time. Radon prediction and early warning research should be conducted within a stable range to ensure the accuracy of the final prediction results, but this research has yet to be conducted.

[0005] Therefore, it is necessary to design a uranium tailings pond radon stability zone division and prediction warning method that can divide radon concentration into intervals, has good warning accuracy, high warning efficiency and is scientific. Summary of the Invention

[0006] In order to overcome the above problems, the present invention provides a uranium tailings pond radon stable zone division and prediction and early warning method, which adopts a combination of range method and variance method to divide the radon concentration data of the tailings pond into a radon concentration stable zone, an asymptotically stable zone and an unstable zone. Then, the radon concentration in the stable zone is predicted according to the WOA-BP neural network algorithm, and the early warning threshold is determined by the maximum value sequence and the minimum value sequence to establish an alarm relationship model.

[0007] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is:

[0008] A method for dividing and predicting radon stability zones in a uranium tailings pond comprises the following steps:

[0009] S1. Based on the definite downward trend of radon concentration in time series, the radon concentration is divided into an unstable zone, a gradually stable zone, and a stable zone according to different fluctuations of radon concentration;

[0010] S2. Use the range method and variance method to divide the unstable area, gradually stable area, and stable area into intervals, and integrate the intervals divided by the range method and variance method to obtain the regional division model;

[0011] S3. Collect monitoring data on temperature and rainfall in previous years at the location of the uranium tailings pond, and calculate the average temperature and rainfall for each quarter of previous years;

[0012] S4. Establish a prediction model for radon concentration in stable areas using the WOA-BP neural network algorithm;

[0013] S5. Calculate the alarm threshold interval of radon concentration based on extreme value theory and construct an alarm relationship model.

[0014] Furthermore, in step S2, using the range method to divide the radon concentration into intervals includes the following steps:

[0015] S211. Count and calculate the range of radon concentration data R = maxf(x) - minf(x), where f(x) is the radon concentration at time x, x∈[T0,T max ];

[0016] S212. Set the critical range of two adjacent intervals to: k1=m1R, k2=m2R, where k1 is the critical range between the unstable region and the gradually stable region, k2 is the critical range between the gradually stable region and the stable region, and m1 and m2 are threshold parameters.

[0017] S213, let T2 = T max -l, and enter the interval range [T2,T max ] radon concentration data, where l = Δt;

[0018] S214, calculation interval [T2,T max ] The inner sample range k = maxf(x) - minf(x), if k < k2, then execute step S215, if k ≥ k2, then output T2;

[0019] S215, set T2 = T2-1, then go to step S214;

[0020] S216, let T1 = T2-1, and input the interval range [T1, T max ]’s radon concentration data;

[0021] S217, calculation interval [T1,Tmax ] The inner sample range k = maxf(x) - minf(x), if k < k1, then execute step S218, if k ≥ k1, then output T1;

[0022] S218, set T1 = T1-1, then go to step S217;

[0023] S219, after integrating the above calculation results, it is concluded that the unstable region is [T0, T1], the gradually stable region is [T1, T2], and the stable region is [T2, T max ].

[0024] Furthermore, in step S2, using the variance method to divide the radon concentration into intervals includes the following steps:

[0025] S221. Count and calculate the variance of radon concentration data

[0026]

[0027] Where, represents the overall mean of radon, f i represents the radon concentration at each moment, and n represents the number of all radon data;

[0028] S222. Set the critical variance values ​​of two adjacent intervals to: k1=m1D(x), k2=m2D(x), where k1 is the critical range difference between the unstable region and the gradually stable region, k2 is the critical range difference between the gradually stable region and the stable region, and m1 and m2 are threshold parameters;

[0029] S223, let T2 = T max -l, and enter the interval range [T2,T max ] radon concentration data, where l = Δt;

[0030] S224, calculation interval [T2,T max ]Intra-sample variance If D < k2, execute step S225; if D ≥ k2, output T2;

[0031] S225, set T2 = T2-1, then go to step S224;

[0032] S226, let T1 = T2-1, and input the interval range [T1, T max ]’s radon concentration data;

[0033] S227, calculation interval [T1,T max ]Internal sample range If D < k1, execute step S228; if D ≥ k1, output T1;

[0034] S228, set T1 = T1-1, then go to step S227;

[0035] S229, after integrating the above calculation results, it is concluded that the unstable region is [T0, T1], the gradually stable region is [T1, T2], and the stable region is [T2, T max ].

[0036] Furthermore, the unstable interval, gradually stable interval, and stable interval divided by the range method and the variance method are respectively averaged to obtain the division model of each area.

[0037] Furthermore, in step S4, the whale optimization algorithm is used to optimize the initial weights and thresholds of the BP neural network, thereby obtaining a relatively stable WOA-BP neural network model, which includes the following steps:

[0038] S41, setting initialization parameters;

[0039] S42, determining a fitness function, and calculating fitness values ​​of individual whales at all positions;

[0040] S43, after comparing the calculated fitness values ​​at all positions, find the position with the best fitness value and record it as the current best individual position X best (t);

[0041] S44. Based on the whale's three behaviors of randomly searching for prey, surrounding selected prey, and preying on selected prey, different position update models are used to update the position. According to the iteration rule, if the current optimal value is better than the previous iteration result, the update is continued; otherwise, no update is performed and the iteration is continued.

[0042] S45. When the fitness value F is less than the initial precision or the number of iterations is exhausted, the iteration is terminated and the best fitness value and its corresponding global optimal position X are output. best ;

[0043] S46, the global optimal position X when the algorithm iteration ends best Optimize the BP neural network, that is, obtain the optimal weight and threshold parameters and assign them to the BP neural network model, thereby realizing network training and simulation prediction.

[0044] Furthermore, in step S41, the initialization parameters include the setting of BP neural network initialization parameters and WOA parameter initialization; wherein, BP neural network initialization is to determine the input and output structure of the BP neural network, as well as the initial connection weights and thresholds; WOA parameter initialization is to convert the initial weights and thresholds in the BP neural network into the position vector of the WOA; in addition, the setting of initialization parameters also includes other basic parameters of the initialization algorithm, such as: population size N, maximum number of iterations T max , and the initial convergence factor a.

[0045] Furthermore, in step S42, the fitness function of the WOA is defined as the mean square error of the training set and the test set of radon concentration, specifically:

[0046]

[0047] Where N is the total number of samples, N1 is the number of training sets, N2 is the number of test sets, N=N1+N2; train Output value for the training set, y test is the test set output value, and y is the actual radon concentration value.

[0048] Furthermore, in step S5, before adopting the extreme value theory, a maximum value sequence and a minimum value sequence are first obtained; wherein, the maximum value sequence is obtained by selecting a maximum value of radon concentration in each year and then sampling and selecting it year by year; the minimum value sequence is obtained by selecting a minimum value of radon concentration in each year and then sampling and selecting it year by year.

[0049] Furthermore, in step S5, the alarm threshold interval of the radon concentration is established according to the extreme value sequence rule of the obtained stable zone radon concentration, which specifically includes the following steps:

[0050] S51, take the average maximum value of the maximum value sequence as the upper threshold u1 of the stable region, and take the average minimum value of the minimum value sequence as the lower threshold u2 of the stable region;

[0051] S52, set the tolerance interval, that is, at time t, when or The tolerance interval is defined as time, where y is the actual monitoring data at time t, y' is the predicted data at time t, u1 and u2 are the upper and lower bounds of the threshold respectively, and the tolerance parameter a = 0.1.

[0052] Furthermore, in step S5, the alarm conditions of the alarm relationship model include the following two situations: when the actual monitored radon concentration exceeds the tolerance interval, an alarm is issued; when three consecutive monitoring values ​​are outside the threshold interval but do not exceed the tolerance interval (indicating that the area is very likely to have internal damage), an alarm is issued.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention provides a method for dividing and predicting the radon stability zone of a uranium tailings reservoir by combining the range method and the variance method to divide the radon concentration data of the tailings reservoir into a radon concentration stability zone, an asymptotic stability zone, and an unstable zone. The radon concentration in the stable zone is then predicted based on the WOA-BP neural network algorithm, and the warning threshold is determined by the maximum value sequence and the minimum value sequence to establish an alarm relationship model, thereby effectively determining the radon concentration stability zone. Compared with the direct prediction method, the prediction accuracy and warning scientificity are effectively improved. At the same time, the present invention adopts the WOA-BP algorithm that combines the advantages of the Whale Optimization Algorithm algorithm and the Back Propagation algorithm to ensure the accuracy and efficiency of the prediction results, thereby efficiently and accurately predicting and warning the data in the stable zone. Compared with the existing uranium tailings reservoir radon concentration prediction and warning model, the present invention establishes a warning model based on the determined radon concentration stability zone using the WOA-BP algorithm that is most suitable for the stability zone. The prediction and warning effect of the combination of the two can greatly improve the credibility of the entire model. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of the method for dividing radon stability zones and predicting and warning uranium tailings reservoirs of the present invention;

[0056] Figure 2 This is a flow chart of a criticality determination algorithm using a range method for the uranium tailings reservoir radon stability zone division and prediction and early warning method of the present invention;

[0057] Figure 3 This is a schematic flow chart of a variance method critical determination algorithm for the uranium tailings reservoir radon stability zone division and prediction and early warning method of the present invention;

[0058] Figure 4 It is a flow chart of the WOA-BP neural network prediction model of the uranium tailings pond radon stability zone division and prediction and early warning method of the present invention;

[0059] Figure 5 This is a pre-processed data diagram of the uranium tailings reservoir radon stability zone division and prediction and early warning method of the present invention;

[0060] Figure 6 This is a schematic diagram of the division of unstable areas, gradually stable areas, and stable areas of the uranium tailings pond radon stability area division and prediction and early warning method of the present invention;

[0061] Figure 7 It is a schematic diagram comparing the prediction results before and after the radon stability zone division and prediction and early warning method of the uranium tailings pond of the present invention;

[0062] Figure 8 1. It is a schematic diagram comparing the WOA-BP and BP prediction results of the uranium tailings pond radon stability zone division and prediction and early warning method of the present invention; DETAILED DESCRIPTION

[0063] In order to make the objectives, 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 here that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions 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 includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus.

[0066] Example

[0067] like Figure 1 As shown, a method 100 for dividing and predicting radon stable zones in uranium tailings ponds is provided. The radon concentration data of the tailings pond are divided into a radon concentration stable zone, an asymptotically stable zone, and an unstable zone by combining the range method and the variance method. The radon concentration in the stable zone is then predicted based on the WOA-BP neural network algorithm. The warning threshold is determined by the maximum value sequence and the minimum value sequence to establish an alarm relationship model, thereby effectively determining the radon concentration stable zone. Compared with the direct prediction method, the prediction accuracy and the scientific nature of the warning are effectively improved.

[0068] At the same time, the present invention adopts the WOA-BP algorithm that combines the advantages of the Whale Optimization Algorithm algorithm and the BackPropagation algorithm, which can ensure the accuracy and efficiency of the prediction results, thereby efficiently and accurately predicting and warning the data in the stable zone. Compared with the existing uranium tailings pond radon concentration prediction and warning model, the present invention establishes a warning model based on the determined radon concentration stable zone using the WOA-BP algorithm that is most suitable for the stable zone. The prediction and warning effect after combining the two can greatly improve the credibility of the entire model.

[0069] Specifically, the radon stability zone division and prediction and early warning method of uranium tailings ponds includes the following steps:

[0070] S1. Based on the definite downward trend of radon concentration in time series, the radon concentration is divided into an unstable zone, a gradually stable zone, and a stable zone according to different fluctuations of radon concentration.

[0071] In this step, after the radon concentration is divided into intervals, subsequent prediction and early warning can be performed based on the radon concentration in the stable area to prevent the radon concentration value from fluctuating too much and causing large errors in the final prediction results.

[0072] S2. The range method and variance method are used to divide the unstable area, gradually stable area, and stable area respectively, and the intervals divided by the range method and variance method are integrated to obtain the regional division model.

[0073] In this step, both the range method and the variance method have good regional critical division effects. In view of the differences in their applicability, in order to ensure the division effect of the three regions of radon concentration, the intervals divided by the two methods are averaged to obtain the division model of each region.

[0074] Among them, the range method can be used to evaluate the dispersion of a set of data. Since the radon concentration shows a trend of continuous decrease until it reaches a flat trend in time series, the range method can be used to achieve a rough regional division of radon concentration. Specifically, the following steps are included:

[0075] S211. Count and calculate the range of radon concentration data R = maxf(x) - minf(x), where f(x) is the radon concentration at time x, x∈[T0,T max ];

[0076] S212. Set the critical range of two adjacent intervals to: k1=m1R, k2=m2R, where k1 is the critical range between the unstable region and the gradually stable region, k2 is the critical range between the gradually stable region and the stable region, and m1 and m2 are threshold parameters.

[0077] S213, let T2 = T max -l, and enter the interval range [T2,T max ] radon concentration data, where l = Δt;

[0078] S214, calculation interval [T2,T max ] The inner sample range k = maxf(x) - minf(x), if k < k2, then execute step S215, if k ≥ k2, then output T2;

[0079] S215, set T2 = T2-1, then go to step S214;

[0080] S216, let T1 = T2-1, and input the interval range [T1, T max ]’s radon concentration data;

[0081] S217, calculation interval [T1,T max ] The inner sample range k = maxf(x) - minf(x), if k < k1, then execute step S218, if k ≥ k1, then output T1;

[0082] S218, set T1 = T1-1, then go to step S217;

[0083] S219, after integrating the above calculation results, it is concluded that the unstable region is [T0, T1], the gradually stable region is [T1, T2], and the stable region is [T2, T max ].

[0084] The variance method can represent the degree of dispersion of the monitoring sequence. The larger the variance, the greater the fluctuation of the radon concentration. The variance method can be used to roughly divide the radon concentration into regions. Specifically, the following steps are included:

[0085] S221. Count and calculate the variance of radon concentration data

[0086]

[0087] Where, represents the overall mean of radon, f i represents the radon concentration at each moment, and n represents the number of all radon data;

[0088] S222. Set the critical variance values ​​of two adjacent intervals to: k1=m1D(x), k2=m2D(x), where k1 is the critical range difference between the unstable region and the gradually stable region, k2 is the critical range difference between the gradually stable region and the stable region, and m1 and m2 are threshold parameters;

[0089] S223, let T2 = T max -l, and enter the interval range [T2,T max ] radon concentration data, where l = Δt;

[0090] S224, calculation interval [T2,T max ]Intra-sample variance If D < k2, execute step S225; if D ≥ k2, output T2;

[0091] S225, set T2 = T2-1, then go to step S224;

[0092] S226, let T1 = T2-1, and input the interval range [T1, T max ]’s radon concentration data;

[0093] S227, calculation interval [T1,T max]Internal sample range If D < k1, execute step S228; if D ≥ k1, output T1;

[0094] S228, set T1 = T1-1, then go to step S227;

[0095] S229, after integrating the above calculation results, it is concluded that the unstable region is [T0, T1], the gradually stable region is [T1, T2], and the stable region is [T2, T max ].

[0096] S3. Collect the temperature and rainfall monitoring data of the uranium tailings pond in previous years, and calculate the average temperature and rainfall in each quarter of previous years.

[0097] In this step, radon is produced by the decay of radium in uranium tailings and migrates into the atmosphere through the pores of the overburden. It is significantly affected by temperature and soil moisture. Therefore, using ambient temperature and rainfall as interference factors can ensure the reliability of the prediction results.

[0098] S4. Establish a prediction model for radon concentration in stable areas using the WOA-BP neural network algorithm.

[0099] In this step, the standard BP neural network algorithm not only has slow convergence speed and is very sensitive to the selection of initial weights, but also has a blind approach to selecting the number of hidden layer nodes. This often leads to problems such as low accuracy and a tendency to fall into local minima. The WOA-BP neural network algorithm, however, combines the advantages of the Whale Optimization Algorithm and the BackPropagation algorithm, achieving excellent results in terms of function solving and convergence speed.

[0100] The Whale Optimization Algorithm (WOA) is a meta-heuristic swarm optimization algorithm that imitates the "spiral bubble net" strategy of the humpback whale. It mainly includes three behavioral modes: randomly searching for prey, surrounding selected prey, and preying on selected prey. Different position update models are established based on these three behavioral modes.

[0101] Specifically, the three location update models are all known models and can be directly applied when used. The establishment of each model is as follows:

[0102] (1) Position update model based on random prey-seeking behavior

[0103] First, the coefficients A and C are introduced and calculated as follows:

[0104] A=2a·r1-a

[0105] C=2r2

[0106] a=2-2t / T max

[0107] Where r1 and r2 are random numbers between the interval [0,1]; a decreases linearly during the iteration process; T max represents the maximum number of iterations, and t represents the current number of iterations.

[0108] Finding a solution to a problem can be understood as the process of a whale group searching for prey. During the random search for prey, the random value of |A| is set to be greater than or equal to 1 or less than 1, forcing the search agent to move away from the reference whale individual. The position change during the random search for prey can be converted into the following mathematical model:

[0109] D k =|X(t) k -C·X rand (t) k |

[0110] X(t+1) k =X rand (t) k -A·D k

[0111] Where D represents the distance between individual whales and prey; X rand is a position vector randomly selected from the current whale group, which contains a feasible solution; X(t) is the position vector of the individual, and X(t+1) is the next position vector of the individual in search of prey; the subscript k represents the kth component of the spatial coordinate.

[0112] (2) Position update model based on the behavior of surrounding target prey

[0113] In this model, the computational process of approaching a feasible solution can be simulated as a humpback whale approaching a selected prey. If the prey is determined to be the best prey at the moment, the position is updated. The mathematical model is as follows:

[0114] D k =|X(t) k -CX best (t) k |

[0115] X(t+1) k =X best (t) k -A·D k

[0116] Where t is the current iteration number; A and C are coefficients; X bestis the current best position; X(t) is the current position, and X(t+1) is the next position; the subscript k represents the kth component of the spatial coordinate.

[0117] (3) Position update model based on predation behavior of selected prey

[0118] In this model, humpback whales update their positions by spiraling upwards to catch their chosen prey. This behavior can be simulated using a logarithmic spiral equation, as shown in the following mathematical model:

[0119] D′ k =|X * (t) k -X(t) k |

[0120] X(t+1) k =D′ k ·e bl ·cos(2πl)+X * (t) k

[0121] Where, X * is the position of the best whale individual; b is a logarithmic spiral constant, l is a random number in the interval [-1,1]; D′ represents the distance between the current best position of the first whale group individual and the prey; the subscript k represents the kth component of the spatial coordinate.

[0122] In this model, the shrinking and circling mechanism is implemented by reducing the value of a in the position update model based on random prey search behavior. The fluctuation range of A will continue to decrease as a decreases. A is a random value in the interval [-a, a]. Setting A to a random value in the interval [-1, 1] will cause the search agent's new position to randomly appear somewhere between the previous position and the current optimal solution position. The humpback whale shrinks and circling while moving along a spiral path toward the selected prey. Both behaviors occur simultaneously. At the same time, assuming that the update probability of circling predation and bubble net attack is 0.5 during the development phase, the mathematical model is as follows:

[0123]

[0124] Where p is a random number in the interval [0,1].

[0125] It is worth noting that in order to obtain a more stable WOA-BP neural network model, the present invention uses the whale optimization algorithm to optimize the initial weights and thresholds of the BP neural network, which includes the following steps:

[0126] S41, setting initialization parameters;

[0127] In this step, the initialization parameters include the BP neural network initialization parameters and the WOA parameter initialization. BP neural network initialization is to determine the input and output structure of the BP neural network, as well as the initial connection weights and thresholds. WOA parameter initialization is to convert the initial weights and thresholds in the BP neural network into the position vectors of the WOA. In addition, the initialization parameter settings also include other basic parameters of the initialization algorithm, such as: population size N, maximum number of iterations T max , and the initial convergence factor a.

[0128] S42, determining a fitness function, and calculating fitness values ​​of individual whales at all positions;

[0129] In this step, the fitness function of WOA is defined as the mean square error of the training set and the test set of radon concentration, specifically:

[0130]

[0131] Where N is the total number of samples, N1 is the number of training sets, N2 is the number of test sets, N=N1+N2; train Output value for the training set, y test is the test set output value, and y is the actual radon concentration value.

[0132] S43, after comparing the calculated fitness values ​​at all positions, find the position with the best fitness value and record it as the current best individual position X best (t);

[0133] S44. Based on the whale's three behaviors of randomly searching for prey, surrounding selected prey, and preying on selected prey, different position update models are used to update the position. According to the iteration rule, if the current optimal value is better than the previous iteration result, the update is continued; otherwise, no update is performed and the iteration is continued.

[0134] S45. When the fitness value F is less than the initial precision or the number of iterations is exhausted, the iteration is terminated and the best fitness value and its corresponding global optimal position X are output. best ;

[0135] S46, the global optimal position X when the algorithm iteration ends best Optimize the BP neural network, that is, obtain the optimal weight and threshold parameters and assign them to the BP neural network model, thereby realizing network training and simulation prediction.

[0136] S5. Calculate the alarm threshold interval of radon concentration based on extreme value theory and construct an alarm relationship model.

[0137] In this step, before applying extreme value theory, we first obtain a maximum value sequence and a minimum value sequence. The maximum value sequence is obtained by selecting a maximum radon concentration each year and then sampling it year by year. The minimum value sequence is obtained by selecting a minimum radon concentration each year and then sampling it year by year.

[0138] At the same time, the alarm threshold interval of radon concentration is established according to the extreme value sequence law of radon concentration in the obtained stable zone, which specifically includes the following steps:

[0139] S51, take the average maximum value of the maximum value sequence as the upper threshold u1 of the stable region, and take the average minimum value of the minimum value sequence as the lower threshold u2 of the stable region;

[0140] S52, set the tolerance interval, that is, at time t, when or The tolerance interval is defined as time, where y is the actual monitoring data at time t, y' is the predicted data at time t, u1 and u2 are the upper and lower bounds of the threshold respectively, and the tolerance parameter a = 0.1.

[0141] The alarm conditions of the alarm relationship model include the following two situations: when the actual monitored radon concentration exceeds the tolerance interval, an alarm is issued; when three consecutive monitoring values ​​are outside the threshold interval but do not exceed the tolerance interval (indicating that the area is very likely to have internal damage), an alarm is issued.

[0142] The feasibility of this application is verified as follows:

[0143] A tailings pond in southern China, located in a densely populated area, was used as a validation target. The pond consists of 10 dam sections, totaling 4,600 meters in length, representing approximately 77% of the pond's total perimeter. Quarterly radon concentration monitoring data from 2001 to 2020 was used to validate the proposed method. The experimental group included the predicted and early warning results after radon stability zones were defined, while the control group included the warning results from all data without radon stability zones.

[0144] First, this application uses the radon concentration monitoring data of a certain dam from 2001 to 2020 for verification. The radon concentration data is shown in the figure below. Figure 5 As shown, it can be seen that the fluctuation of radon concentration has a definite downward trend in time series.

[0145] Then, using the expert scoring method, we took m1 = 0.5 and m2 = 0.25, and performed the range method stable interval calculation. At the same time, we took m1 = 0.5 and m2 = 0.2, and performed the variance method stable interval calculation. Then, based on the calculation results of the range method and the variance method, we took the average value and calculated that the unstable area was [2001, 2003], the gradually stable area was [2003, 2007], and the stable interval was [2007, 2021]. The intervals after division are as follows Figure 6 As shown, the red dashed line represents the critical line between the three intervals.

[0146] Next, we perform WOA-BP prediction on the data before and after the stable zone is divided. It can be seen that the prediction results after the stable interval is divided are closer to the true value, such as Figure 7 As shown in the figure, this figure is a comparison of the prediction results before and after the stable zone division. It can be seen from the figure that the stable zone division method is reliable.

[0147] In terms of prediction model, the prediction results before and after the optimization of the whale algorithm are compared during the experiment. It can be seen that the prediction results of WOA-BP are closer to the true value. Figure 8 As shown in the figure, this figure is a comparison of the prediction results of WOA-BP and BP. It can be directly seen from the figure that the WOA-BP prediction method is more efficient and reliable.

[0148] Finally, in order to verify the practicality of the model warning algorithm, the radon concentration monitoring data from 2018 to 2020 were used for forecasting and warning, and the upper and lower thresholds were calculated according to the method of this application, and then the warning range from 2018 to 2020 was obtained. The specific contents are shown in Table 1.

[0149]

[0150] Table 1 2018-2020 graded warning intervals

[0151] At the same time, combined with Table 1, the warning range of radon concentration from 2018 to 2020 is obtained, as shown in Table 3.

[0152]

[0153]

[0154] Table 3 Radon warning range from 2018 to 2020

[0155] From Table 1 and Table 2, we can see the specific ranges of the safe area, tolerance area, and alarm area of ​​the radon concentration safety status of the dam body from 2018 to 2020. Figure 5As can be seen, between 2001 and 2020, the radon concentration data within the dam showed an overall downward trend. This is due to the following reasons: First, with the advancement of scientific and technological means, more advanced radon control technologies for uranium tailings dams have been gradually implemented in engineering applications, effectively controlling radon release from the dam. Second, with rainfall and evaporation, the tailings dam undergoes continuous wet-dry cycles, compacting the soil and reducing radon release pathways. Furthermore, the data in Table 2 exhibit significant errors between the predicted and actual values ​​for some quarters. This issue arises because the quarterly temperature averages do not fully represent the actual temperature conditions on the day the tailings dam management unit conducted radon measurements. Therefore, subsequent engineering applications should directly use the same day's meteorological data for prediction. However, the historical data indicates that the predicted and actual warning levels are within the same range, indicating that the error is within the tolerance range and the prediction results remain reasonable. This application has determined a "safe" warning level for the dam, which is consistent with the safety assessment report provided by the tailings dam management unit. This demonstrates that the stable interval partitioning method combined with the WOA-BP algorithm employed in this application to establish a warning model has excellent reference value and guiding significance.

[0156] The above description is 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 aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some or all of the technical features therein can be replaced by equivalents. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied to other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for dividing radon stability zones and predicting and warning of uranium tailings ponds, characterized in that: The following steps are involved: S1. Based on the definite downward trend of radon concentration in time series, the radon concentration is divided into an unstable zone, a gradually stable zone, and a stable zone according to different fluctuations of radon concentration; S2. Use the range method and variance method to divide the unstable area, gradually stable area, and stable area into intervals, and integrate the intervals divided by the range method and variance method to obtain the regional division model; S3. Collect monitoring data on temperature and rainfall in previous years at the location of the uranium tailings pond, and calculate the average temperature and rainfall for each quarter of previous years; S4. Establish a prediction model for radon concentration in stable areas using the WOA-BP neural network algorithm; S5. Calculate the alarm threshold interval of radon concentration based on extreme value theory and construct an alarm relationship model; In step S2, the interval division of radon concentration using the range method includes the following steps: S211. Count and calculate the range of radon concentration data R = maxf(x) - minf(x), where f(x) is the radon concentration at time x, x∈[T0,T max ]; S212. Set the critical range of two adjacent intervals to: k1=m1R, k2=m2R, where k1 is the critical range between the unstable region and the gradually stable region, k2 is the critical range between the gradually stable region and the stable region, and m1 and m2 are threshold parameters. S213, let T2 = T max -l, and enter the interval range [T2,T max ] radon concentration data, where l = Δt; S214, calculation interval [T2,T max ] The inner sample range k = maxf(x) - minf(x), if k < k2, then execute step S215, if k ≥ k2, then output T2; S215, set T2 = T2-1, then go to step S214; S216, let T1 = T2-1, and input the interval range [T1, T max ]’s radon concentration data; S217, calculation interval [T1,T max ] The inner sample range k = maxf(x) - minf(x), if k < k1, then execute step S218, if k ≥ k1, then output T1; S218, set T1 = T1-1, then go to step S217; S219, after integrating the above calculation results, it is concluded that the unstable region is [T0, T1], the gradually stable region is [T1, T2], and the stable region is [T2, T max ].

2. The method for dividing radon stability zones and predicting and warning uranium tailings according to claim 1, characterized in that: In step S2, the interval division of radon concentration using the variance method includes the following steps: S221. Count and calculate the variance of radon concentration data Where, represents the overall mean of radon, f i represents the radon concentration at each moment, and n represents the number of all radon data; S222. Set the critical variance values ​​of two adjacent intervals to: k3 = m1D(x), k4 = m2D(x), where k3 is the critical variance value between the unstable region and the gradually stable region, k4 is the critical variance value between the gradually stable region and the stable region, and m1 and m2 are threshold parameters; S223, let T2 = T max -l, and enter the interval range [T2,T max ] radon concentration data, where l = Δt; S224, calculation interval [T2,T max ]Intra-sample variance If D < k4, execute step S225; if D ≥ k4, output T2; S225, set T2 = T2-1, then go to step S224; S226, let T1 = T2-1, and input the interval range [T1, T max ]’s radon concentration data; S227, calculation interval [T1,T max ]Intra-sample variance If D < k3, execute step S228; if D ≥ k3, output T1; S228, set T1 = T1-1, then go to step S227; S229, after integrating the above calculation results, it is concluded that the unstable region is [T0, T1], the gradually stable region is [T1, T2], and the stable region is [T2, T max ].

3. The method for dividing radon stability zones and predicting and warning uranium tailings according to claim 2, characterized in that: The unstable interval, gradually stable interval, and stable interval divided by the range method and variance method are respectively averaged to obtain the division model of each area.

4. The method for dividing radon stability zones and predicting and warning uranium tailings according to claim 3, characterized in that: In step S4, the whale optimization algorithm is used to optimize the initial weights and thresholds of the BP neural network, thereby obtaining a relatively stable WOA-BP neural network model, which includes the following steps: S41, setting initialization parameters; S42, determining a fitness function, and calculating fitness values ​​of individual whales at all positions; S43, after comparing the calculated fitness values ​​at all positions, find the position with the best fitness value and record it as the current best individual position X best (t); S44. Based on the whale's three behaviors of randomly searching for prey, surrounding selected prey, and preying on selected prey, different position update models are used to update the position. According to the iteration rule, if the current optimal value is better than the previous iteration result, the update is continued; otherwise, no update is performed and the iteration is continued. S45. When the fitness value F is less than the initial precision or the number of iterations is exhausted, the iteration is terminated and the best fitness value and its corresponding global optimal position X are output. best ; S46, the global optimal position X when the algorithm iteration ends best Optimize the BP neural network, that is, obtain the optimal weight and threshold parameters and assign them to the BP neural network model, thereby realizing network training and simulation prediction.

5. The method for dividing and predicting radon stability zones in uranium tailings reservoirs according to claim 4, wherein: In step S41, the initialization parameters include the setting of BP neural network initialization parameters and WOA parameter initialization; wherein, BP neural network initialization is to determine the input and output structure of BP neural network, as well as the initial connection weights and thresholds; WOA parameter initialization is to convert the initial weights and thresholds in BP neural network into the position vector of WOA; in addition, the setting of initialization parameters also includes other basic parameters of initialization algorithm, such as population size N, maximum number of iterations T' max , and the initial convergence factor a.

6. The method for dividing and predicting radon stability zones in uranium tailings reservoirs according to claim 5, characterized in that: In step S42, the fitness function of the WOA is defined as the mean square error of the training set and the test set of radon concentration, specifically: Where N is the total number of samples, N1 is the number of training sets, N2 is the number of test sets, N=N1+N2; train Output value for the training set, y test is the test set output value, and y is the actual radon concentration value.

7. The method for dividing and predicting radon stability zones in uranium tailings reservoirs according to claim 6, characterized in that: In step S5, before adopting the extreme value theory, the maximum value sequence and the minimum value sequence are first obtained; wherein, the maximum value sequence is obtained by selecting a maximum value of the radon concentration in each year and then sampling and selecting it year by year; the minimum value sequence is obtained by selecting a minimum value of the radon concentration in each year and then sampling and selecting it year by year.

8. The method for dividing and predicting radon stability zones in uranium tailings reservoirs according to claim 7, characterized in that: In step S5, the alarm threshold interval of the radon concentration is established according to the extreme value sequence rule of the radon concentration in the obtained stable area.

9. The method for dividing and predicting radon stability zones in uranium tailings reservoirs according to claim 8, characterized in that: In step S5, the alarm conditions of the alarm relationship model include the following two situations: an alarm is issued when the actual monitored radon concentration exceeds the tolerance interval; an alarm is issued when three consecutive monitoring values ​​are outside the threshold interval but do not exceed the tolerance interval.

Images