Infiltration line prediction method and system based on LSTM time sequence neural network

Abstract

The invention discloses an infiltration line prediction method and system based on an LSTM time sequence neural network, and relates to the technical field of tailing pond safety monitoring, and the key points of the technical scheme are that the method comprises the steps: laying a sensor network to collect monitoring data; carrying out abnormal value restoration, missing value filling and normalization processing on the monitoring data; screening input features and carrying out dimension reduction processing by adopting principal component analysis; establishing an LSTM neural network model, and configuring an activation function and an optimization algorithm; training and verifying the model by using the training set and the test set; and preprocessing to-be-predicted monitoring data, inputting the preprocessed to-be-predicted monitoring data into the model, and carrying out anti-normalization processing to obtain an actual infiltration line position. According to the method, modeling is carried out on the time sequence characteristics of the infiltration line data through the LSTM, the prediction precision of the infiltration line is improved by utilizing historical monitoring data, the method does not depend on tailings pond geology and complex physical models and is adaptive to various tailings ponds, meanwhile, the model training efficiency is improved by adopting a principal component analysis method, and real-time decision support can be provided for safe operation of the tailings ponds.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tailings pond safety monitoring, more specifically, it relates to a seepage line prediction method and system based on an LSTM time series neural network. BACKGROUND

[0002] As a key supporting facility in the production chain of mines, tailings ponds are mainly used to store tailings produced during the mining and beneficiation processes. The safe and stable operation of tailings ponds is directly related to the surrounding ecological environment, the safety of residents' lives and property, and the continuity of mine production. In the safety monitoring system of tailings ponds, the seepage line is one of the core monitoring indicators. It specifically refers to the free water surface line of groundwater within the tailings dam. The dynamic changes in its position directly reflect the seepage state inside the dam, and are strongly related to the stability of the dam. When the seepage line is too high, it will significantly increase the pore water pressure of the dam, reduce the shear strength of the dam soil, and easily cause piping, soil flow, and even dam failure and other major safety accidents. Therefore, accurate and efficient prediction of the position of the seepage line is a key technical support for identifying safety risks in tailings ponds, developing prevention and control measures, and ensuring the safe operation of tailings ponds. It has important practical significance and application value in the field of mine safety engineering.

[0003] Currently, the technical means for predicting the seepage line of tailings ponds in the industry mainly includes two categories: traditional empirical formula method and physical model method. The empirical formula method is based on a large amount of field measurement data and establishes a fitting formula for the position of the seepage line and a single or small number of influencing factors such as reservoir water level and rainfall, such as a linear correlation formula based on reservoir water level and seepage line depth, an empirical expression corrected with rainfall, etc. This method has a simple calculation process, low requirements for calculation devices, and has been widely used in early tailings pond monitoring. The physical model method is based on the principles of seepage mechanics and constructs a physical and mathematical model of the seepage field of the tailings dam. By inputting geological and engineering parameters such as the permeability coefficient, porosity, and boundary seepage conditions of the dam rock-soil mass, the seepage process of groundwater inside the dam is simulated, and the position of the seepage line is predicted. Common methods include finite element seepage simulation models, analytical seepage field models, etc. This method is more in line with the nature of seepage at the theoretical level and has certain application value in small and medium-sized tailings ponds with clear geological conditions.

[0004] However, the traditional seepage line prediction method still has the following problems: first, it requires high geological conditions and operating conditions of the tailings pond, and needs a large number of geological parameters and accurate boundary conditions; second, the prediction accuracy is limited and it is difficult to adapt to complex tailings pond conditions; third, it cannot fully utilize the large amount of historical data accumulated by the tailings pond monitoring system.

[0005] In recent years, with the rapid development of artificial intelligence technology in the field of time series prediction, long short-term memory network (LSTM) as a special recurrent neural network has shown good prediction performance in meteorological prediction, water quality change prediction and other fields due to its unique advantage of capturing long and short-term dependencies in time series data, but its application research in the specific scenario of tailing pond infiltration line prediction is still in the initial stage, and the related technical scheme has not formed a mature and popularized system.

[0006] Therefore, the present application aims to provide a kind of based on LSTM time series neural network's infiltration line prediction method and system, by the LSTM time series neural network is applied to the field of infiltration line prediction, to solve the defects of traditional infiltration line prediction method. SUMMARY

[0007] The purpose of the present application is to provide a kind of based on LSTM time series neural network's infiltration line prediction method and system, the present application is modeled by LSTM to the time series characteristics of infiltration line data, improves the precision of infiltration line prediction using historical monitoring data, this method does not depend on the tailing pond geology and complex physical model, adapts to various tailing ponds, while using principal component analysis method to improve model training efficiency, can provide real-time decision support for safe operation of tailing pond.

[0008] The above technical purpose of the present application is realized by the following technical scheme: a kind of based on LSTM time series neural network's infiltration line prediction method, comprising the following steps:

[0009] S1, data acquisition: tailing pond sensor network is laid out to collect the corresponding monitoring data of tailing pond in real time;

[0010] S2, data preprocessing: the monitoring data collected is repaired and filled with missing values, the cleaning of monitoring data is completed, then the monitoring data after cleaning is mapped to the [0,1] interval, and the normalization processing of monitoring data is completed;

[0011] S3, feature selection: analyze the correlation between the preprocessed monitoring data and the position of the infiltration line, select the features with higher correlation with the position of the infiltration line as input features, and then use principal component analysis to reduce the dimension of the input features;

[0012] S4, LSTM neural network construction: an LSTM neural network model containing an input layer, an LSTM layer, a fully connected layer and an output layer is established, the number of hidden units and the number of layers in the LSTM layer are set to capture the time series characteristics of the infiltration line data, and the LSTM neural network model is configured with an activation function and an optimization algorithm;

[0013] S5, model training and verification: the preprocessed monitoring data is divided into a training set and a test set, the training set is used to train the constructed LSTM neural network model, the LSTM neural network model is adjusted to reach a preset fitting effect on the training set, and then the test set is used to verify the trained LSTM neural network model to evaluate the prediction accuracy and generalization ability of the LSTM neural network model;

[0014] S6, infiltration line prediction: after the tailing pond monitoring data to be predicted is preprocessed according to step S2, the preprocessed data is input into the trained and verified LSTM neural network model to obtain a predicted value of the infiltration line position, and the actual infiltration line position is obtained by inverse normalization processing of the predicted value.

[0015] The application further provides that: in step S1, the monitoring data includes infiltration line position data, reservoir water level data, rainfall data and dam displacement data.

[0016] The application further provides that: in step S1, the sampling frequency of the monitoring data is once a day.

[0017] The application further provides that: in step S2, the abnormal value repair is corrected according to the monitoring data of adjacent time points of the abnormal value, and the missing value filling is filled with the corresponding monitoring data of the previous sampling period.

[0018] The application further provides that: in step S3, the selected input features include the infiltration line position, reservoir water level, rainfall and dam longitudinal displacement of the previous 3 days, and the dimension of the input features is reduced from 12 to 6 through principal component analysis.

[0019] The application further provides that: in step S4, the number of hidden units of the LSTM layer is set to 50, and the number of layers is set to 2.

[0020] The application further provides that: in step S4, the LSTM neural network model is configured with a ReLU activation function and an Adam optimization algorithm.

[0021] The application further provides that: in step S5, the division ratio of the training set and the test set is 7:3.

[0022] The application further provides that: in step S5, the prediction accuracy of the LSTM neural network model is evaluated by the mean absolute error and the root mean square error, the mean absolute error of the LSTM neural network model is not more than 0.012, and the root mean square error is not more than 0.015.

[0023] The application further provides a kind of infiltration line prediction system based on LSTM time series neural network, including data layer, processing layer and application layer;

[0024] The data layer includes a data acquisition module and a historical database module; the data acquisition module is used to acquire monitoring data of the tailings dam in real time through a sensor network, and the historical database module is used to store historical monitoring data acquired by the sensor network.

[0025] The processing layer includes a data preprocessing module, a feature engineering module, and an LSTM prediction engine module. The data preprocessing module is used to clean and normalize the monitoring data collected and stored in the data layer. The feature engineering module is used to analyze the correlation between the monitoring data and the location of the infiltration line to filter input features and to use principal component analysis to reduce the dimensionality of the input features. The LSTM prediction engine module is used to build an LSTM neural network model, train and validate the model, and predict the infiltration line based on the trained and validated model.

[0026] The application layer includes a real-time early warning system module, a visualization monitoring platform module, and a decision support interface module. The real-time early warning system module is used to provide safety warnings based on the infiltration line prediction results output by the LSTM prediction engine. The visualization monitoring platform module is used to visualize the monitoring data and prediction results. The decision support interface module is used to push the prediction results to the tailings dam safety management department in the form of API to provide decision data support.

[0027] The present invention also provides an immersion line prediction device based on an LSTM temporal neural network, comprising at least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to implement an immersion line prediction method based on an LSTM temporal neural network.

[0028] The present invention also provides a computer-readable storage medium storing computer instructions for execution by a computer to implement an infiltration line prediction method based on an LSTM temporal neural network.

[0029] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements an infiltration line prediction method based on an LSTM temporal neural network.

[0030] In summary, the present invention has the following beneficial effects:

[0031] 1. The prediction accuracy of this invention is significantly improved. By using an LSTM temporal neural network to construct the prediction model, it can accurately capture the temporal dependence characteristics of the seepage line data, fully explore and utilize the historical monitoring data such as seepage line location, reservoir water level, rainfall, and dam displacement accumulated by the tailings dam monitoring system over a long period of time, effectively avoid the errors caused by the reliance on simplified assumptions in traditional empirical formulas or physical models, and greatly improve the accuracy of seepage line location prediction. In a specific tailings dam embodiment, the average absolute error of this invention during the model verification stage is as low as 0.012, and the root mean square error is only 0.015, indicating that this invention can accurately reflect the trend of seepage line changes and provide reliable data support for the prediction of tailings dam safety risks.

[0032] 2. This invention has a wide range of applications and strong adaptability. It can be distinguished from traditional methods that impose strict restrictions on the geological conditions and operating conditions of tailings ponds (requiring a large number of accurate geological parameters and boundary conditions). It does not rely on complex physical models and specific geological attributes of tailings ponds. It achieves prediction only through monitoring data and time series modeling of LSTM networks. This allows tailings ponds with different mineral types, different dam structures, and different geographical environments to quickly adapt to the model by accessing monitoring data, significantly reducing the dependence on specific scenarios. It is suitable for the infiltration line prediction needs of various tailings ponds and has high application value.

[0033] 3. The model training and prediction efficiency of this invention is high. In the feature processing stage, principal component analysis (PCA) is used to reduce the dimensionality of the input features. Specifically, the initial 12-dimensional input features (including the position of the seepage line and the water level of the reservoir in the first 3 days) are reduced to 6 dimensions, which effectively reduces the amount of data calculation during model training, reduces network complexity, and shortens the model training time. At the same time, the Adam optimization algorithm and ReLU activation function are used to further improve the model convergence speed, ensuring that high-precision prediction results are obtained while taking into account training and prediction efficiency. It can quickly respond to the real-time monitoring and prediction needs of tailings dams.

[0034] 4. This invention provides real-time decision support for the safe operation of tailings dams. The overall prediction method of this invention is simple and efficient. From the preprocessing of the monitoring data to be predicted and the input model to the output of the actual phreatic line position, the overall process is responsive and can generate phreatic line prediction results in real time. Combined with the real-time early warning system module and the visualization monitoring platform module in the system, the prediction results can be transformed into safety early warning signals and intuitive data displays in a timely manner, helping managers to quickly grasp the stability status of the tailings dam body, identify the risk of dam failure in advance, and provide real-time and accurate technical support for tailings dam safety scheduling, emergency response and other decisions, effectively ensuring the safe operation of tailings dams. Attached Figure Description

[0035] Figure 1 This is a data transmission diagram of an immersion line prediction method based on an LSTM temporal neural network in Embodiment 1 of the present invention;

[0036] Figure 2 This is a flowchart illustrating an infiltration line prediction method based on an LSTM temporal neural network in Embodiment 1 of the present invention.

[0037] Figure 3 This is a flowchart illustrating the infiltration line prediction system based on an LSTM temporal neural network in Embodiment 2 of the present invention.

[0038] Figure 4 This is a schematic diagram of the module structure of an immersion line prediction system based on an LSTM temporal neural network in Embodiment 2 of the present invention;

[0039] Figure 5 This is a schematic diagram of the structural framework of an immersion line prediction device based on an LSTM temporal neural network in Embodiment 3 of the present invention. Detailed Implementation

[0040] The following is in conjunction with the appendix Figures 1-5 The present invention will be described in further detail below.

[0041] Example 1: An Immersion Line Prediction Method Based on LSTM Temporal Neural Network

[0042] This embodiment uses the tailings dam of Gaowanqiu Mine in Hunan Province as the test object, and specifically includes the following steps:

[0043] S1. Data Acquisition: Deploy a sensor network in the tailings dam to collect real-time monitoring data corresponding to the tailings dam.

[0044] The monitoring data of the tailings dam were collected, including the location of the seepage line, the water level, the rainfall, and the displacement of the dam. The data sampling frequency was once a day.

[0045] S2. Data preprocessing: The collected monitoring data is cleaned by repairing outliers and filling in missing values. Then, the cleaned monitoring data is mapped to the [0,1] interval to complete the normalization process of the monitoring data.

[0046] In this embodiment, missing rainfall data is filled using the rainfall value from the previous day; abnormal infiltration line location data is corrected based on data from adjacent time points. The cleaned data is then normalized to a range between [0,1].

[0047] S3. Feature Selection: Analyze the correlation between the preprocessed monitoring data and the location of the immersion line, select features with high correlation to the location of the immersion line as input features, and then use principal component analysis to reduce the dimensionality of the input features.

[0048] By analyzing the correlation between monitoring data and the location of the seepage line, features with a high correlation to the location of the seepage line were selected as input features, including the location of the seepage line, reservoir water level, rainfall, and longitudinal displacement of the dam body in the previous 3 days. Principal component analysis (PCA) was used to reduce the dimensionality of the input features from 12 to 6.

[0049] S4. LSTM Neural Network Construction: Build an LSTM neural network model that includes an input layer, an LSTM layer, a fully connected layer, and an output layer. Set the number of hidden units and layers in the LSTM layer to capture the temporal characteristics of the immersion line data, and configure activation functions and optimization algorithms for the LSTM neural network model.

[0050] An LSTM neural network model is constructed, consisting of an input layer, an LSTM layer, a fully connected layer, and an output layer. In the LSTM layer, the number of hidden units is set to 50 and the number of layers is 2. The network is trained using the ReLU activation function and the Adam optimization algorithm.

[0051] S5. Model Training and Validation: The preprocessed monitoring data is divided into training and test sets. The constructed LSTM neural network model is trained using the training set. The network parameters are adjusted to make the LSTM neural network model achieve the preset fitting effect on the training set. Then, the trained LSTM neural network model is validated using the test set to evaluate the prediction accuracy and generalization ability of the LSTM neural network model.

[0052] The normalized data was divided into a training set and a test set, with the training set accounting for 70% and the test set accounting for 30%. The LSTM neural network was trained using the training set data. By adjusting the network parameters, the model achieved a good fit on the training set. The trained model was validated using the test set data to evaluate its prediction accuracy and generalization ability. The results showed that the model had a mean absolute error of 0.012 and a root mean square error of 0.015, exhibiting high prediction accuracy and good generalization ability.

[0053] S6. Prediction of the leaching line: After preprocessing the tailings dam monitoring data to be predicted according to step S2, the data is input into the trained and validated LSTM neural network model to obtain the predicted value of the leaching line position. The actual leaching line position is obtained by inverse normalization of the predicted value.

[0054] The monitoring data to be predicted is input into the trained LSTM neural network model to obtain the predicted value of the phreatic line position. The predicted value is then denormalized to convert it into the actual phreatic line position. As can be seen from the prediction results of this embodiment, the method of the present invention can accurately predict the phreatic line position of the tailings dam, providing strong technical support for the safe operation of the tailings dam.

[0055] Example 2: An Immersion Line Prediction System Based on LSTM Temporal Neural Network

[0056] This embodiment specifically includes a data layer, a processing layer, and an application layer; the data layer further includes a data acquisition module and a historical database module; the data acquisition module is used to collect monitoring data of the tailings dam in real time through a sensor network, and the historical database module is used to store historical monitoring data collected by the sensor network.

[0057] The processing layer includes a data preprocessing module, a feature engineering module, and an LSTM prediction engine module. The data preprocessing module is used to clean and normalize the monitoring data collected and stored in the data layer. The feature engineering module is used to analyze the correlation between the monitoring data and the location of the infiltration line to filter input features and to use principal component analysis to reduce the dimensionality of the input features. The LSTM prediction engine module is used to build an LSTM neural network model, train and validate the model, and predict the infiltration line based on the trained and validated model.

[0058] The application layer includes a real-time early warning system module, a visualization monitoring platform module, and a decision support interface module. The real-time early warning system module is used to provide safety warnings based on the infiltration line prediction results output by the LSTM prediction engine. The visualization monitoring platform module is used to visualize the monitoring data and prediction results. The decision support interface module is used to push the prediction results to the tailings dam safety management department in the form of API to provide decision data support.

[0059] Example 3: A seepage line prediction device based on LSTM temporal neural network, comprising at least one processor; and a memory communicatively connected to at least one processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to implement the seepage line prediction method based on LSTM temporal neural network in Example 1, the method comprising: data acquisition: deploying a sensor network in the tailings dam to collect monitoring data corresponding to the tailings dam in real time; data preprocessing: performing outlier repair and missing value filling on the collected monitoring data to clean the monitoring data, and then mapping the cleaned monitoring data to the [0,1] interval to complete the normalization processing of the monitoring data; feature selection: analyzing the correlation between the preprocessed monitoring data and the seepage line position, selecting features with high correlation to the seepage line position as input features, and then using principal component analysis to reduce the dimensionality of the input features; LSTM neural network construction. The process involves: 1) Establishing an LSTM neural network model comprising an input layer, an LSTM layer, a fully connected layer, and an output layer. The number of hidden units and layers in the LSTM layer is set to capture the temporal characteristics of the leaching line data. Activation functions and optimization algorithms are configured for the LSTM neural network model. 2) Model training and validation: The preprocessed monitoring data is divided into training and testing sets. The constructed LSTM neural network model is trained using the training set. Network parameters are adjusted to achieve a preset fit on the training set. The trained LSTM neural network model is then validated using the testing set to evaluate its prediction accuracy and generalization ability. 3) Illumination line prediction: The preprocessed tailings dam monitoring data is input into the trained and validated LSTM neural network model to obtain predicted leaching line locations. These predicted values ​​are then inversely normalized to obtain the actual leaching line locations.

[0060] Example 4: A computer-readable storage medium storing computer instructions for execution by a computer to implement the leaching line prediction method based on an LSTM temporal neural network in Example 1. The method includes: data acquisition: deploying a sensor network in the tailings dam to collect real-time monitoring data corresponding to the tailings dam; data preprocessing: cleaning the collected monitoring data by repairing outliers and filling in missing values, then mapping the cleaned monitoring data to the [0,1] interval to normalize the monitoring data; feature selection: analyzing the correlation between the preprocessed monitoring data and the leaching line location, selecting features with high correlation to the leaching line location as input features, and then using principal component analysis to reduce the dimensionality of the input features; LSTM neural network construction: establishing an LSTM neural network including an input layer, an LSTM layer, and a fully connected layer. The LSTM neural network model for the output layer is constructed by setting the number of hidden units and layers in the LSTM layer to capture the temporal characteristics of the leaching line data, and configuring activation functions and optimization algorithms for the LSTM neural network model. Model training and validation: The preprocessed monitoring data is divided into training and test sets. The constructed LSTM neural network model is trained using the training set. By adjusting the network parameters, the LSTM neural network model achieves the preset fitting effect on the training set. Then, the trained LSTM neural network model is validated using the test set to evaluate the prediction accuracy and generalization ability of the LSTM neural network model. Leakage line prediction: The monitoring data of the tailings dam to be predicted is preprocessed and input into the trained and validated LSTM neural network model to obtain the predicted value of the leaching line position. By inverse normalizing the predicted value, the actual leaching line position is obtained.

[0061] Example 5: A computer program product, including a computer program, which, when executed by a processor, implements the leaching line prediction method based on an LSTM temporal neural network described in Example 1. The method includes: data acquisition: deploying a sensor network in the tailings dam to collect real-time monitoring data corresponding to the tailings dam; data preprocessing: performing outlier repair and missing value filling on the collected monitoring data to clean the data, then mapping the cleaned monitoring data to the [0,1] interval to normalize the monitoring data; feature selection: analyzing the correlation between the preprocessed monitoring data and the leaching line location, selecting features with high correlation to the leaching line location as input features, and then using principal component analysis to reduce the dimensionality of the input features; LSTM neural network construction: constructing an LTM neural network containing an input layer, an LSTM layer, a fully connected layer, and an output layer. The STM neural network model is developed by setting the number of hidden units and layers in the LSTM layer to capture the temporal characteristics of the leaching line data, and configuring activation functions and optimization algorithms for the LSTM neural network model. Model training and validation: The preprocessed monitoring data is divided into training and testing sets. The constructed LSTM neural network model is trained using the training set, and the network parameters are adjusted to achieve a preset fit on the training set. Then, the trained LSTM neural network model is validated using the testing set to evaluate its prediction accuracy and generalization ability. Leakage line prediction: The preprocessed tailings dam monitoring data is input into the trained and validated LSTM neural network model to obtain the predicted leaching line location. The predicted value is then inversely normalized to obtain the actual leaching line location.

[0062] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A method for predicting infiltration lines based on LSTM temporal neural networks, characterized in that: Includes the following steps: S1. Data Acquisition: Deploy a sensor network in the tailings dam to collect the corresponding monitoring data of the tailings dam in real time; S2. Data preprocessing: The collected monitoring data is cleaned by repairing outliers and filling in missing values. Then, the cleaned monitoring data is mapped to the [0,1] interval to complete the normalization process of the monitoring data. S3. Feature Selection: Analyze the correlation between the preprocessed monitoring data and the location of the immersion line, select features with high correlation to the location of the immersion line as input features, and then use principal component analysis to reduce the dimensionality of the input features. S4. LSTM Neural Network Construction: Establish an LSTM neural network model containing an input layer, an LSTM layer, a fully connected layer, and an output layer. Set the number of hidden units and the number of layers in the LSTM layer to capture the temporal characteristics of the infiltration line data, and configure activation functions and optimization algorithms for the LSTM neural network model. S5. Model Training and Validation: The preprocessed monitoring data is divided into a training set and a test set. The constructed LSTM neural network model is trained using the training set. The network parameters are adjusted to make the LSTM neural network model achieve the preset fitting effect on the training set. Then, the trained LSTM neural network model is validated using the test set to evaluate the prediction accuracy and generalization ability of the LSTM neural network model. S6. Prediction of the leaching line: After preprocessing the tailings dam monitoring data to be predicted according to step S2, the data is input into the trained and validated LSTM neural network model to obtain the predicted value of the leaching line position. The actual leaching line position is obtained by inverse normalization of the predicted value.

2. The infiltration line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S1, the monitoring data includes infiltration line location data, reservoir water level data, rainfall data, and dam displacement data.

3. The infiltration line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S1, the sampling frequency of the monitoring data is once a day.

4. The infiltration line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S2, outlier repair is performed by correcting the outlier based on monitoring data at time points adjacent to the outlier, and missing value filling is performed by filling the missing value with the corresponding monitoring data from the previous sampling period.

5. The infiltration line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S3, the selected input features include the location of the seepage line, reservoir water level, rainfall, and longitudinal displacement of the dam body for the previous 3 days. Principal component analysis is used to reduce the dimension of the input features from 12 to 6.

6. The infiltration line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S4, the number of hidden units in the LSTM layer is set to 50, and the number of layers is set to 2.

7. The infiltration line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S4, the LSTM neural network model is configured with the ReLU activation function and the Adam optimization algorithm.

8. The infiltration line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S5, the ratio of the training set to the test set is 7:

3.

9. The immersion line prediction method based on LSTM temporal neural network according to claim 1, characterized in that: In step S5, the prediction accuracy of the LSTM neural network model is evaluated by the mean absolute error and root mean square error. The mean absolute error of the LSTM neural network model is no greater than 0.012, and the root mean square error is no greater than 0.

015.

10. The immersion line prediction system based on LSTM temporal neural network according to claim 1, applied to the immersion line prediction method based on LSTM temporal neural network according to any one of claims 1-9, characterized in that: It includes a data layer, a processing layer, and an application layer; The data layer includes a data acquisition module and a historical database module; the data acquisition module is used to acquire monitoring data of the tailings dam in real time through a sensor network, and the historical database module is used to store historical monitoring data acquired by the sensor network. The processing layer includes a data preprocessing module, a feature engineering module, and an LSTM prediction engine module. The data preprocessing module is used to clean and normalize the monitoring data collected and stored in the data layer. The feature engineering module is used to analyze the correlation between the monitoring data and the location of the infiltration line to filter input features and to use principal component analysis to reduce the dimensionality of the input features. The LSTM prediction engine module is used to build an LSTM neural network model, train and validate the model, and predict the infiltration line based on the trained and validated model. The application layer includes a real-time early warning system module, a visualization monitoring platform module, and a decision support interface module. The real-time early warning system module is used to provide safety warnings based on the infiltration line prediction results output by the LSTM prediction engine. The visualization monitoring platform module is used to visualize the monitoring data and prediction results. The decision support interface module is used to push the prediction results to the tailings dam safety management department in the form of API to provide decision data support.

11. An infiltration line prediction device based on an LSTM temporal neural network, characterized in that: The method includes at least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by the processor to implement an immersion line prediction method based on an LSTM temporal neural network as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions for execution by the computer to implement an immersion line prediction method based on an LSTM temporal neural network as described in any one of claims 1-9.

13. A computer program product, characterized in that: Includes a computer program, which, when executed by a processor, implements an immersion line prediction method based on an LSTM temporal neural network as described in any one of claims 1-9.

Images