Tidal flat shallow stratum deformation prediction device and method

Abstract

The invention discloses a tidal flat shallow stratum deformation prediction device and method, and relates to the field of stratum deformation prediction.The tidal flat shallow stratum deformation prediction device comprises the steps that a plurality of node components are connected in sequence and vertically penetrate into a shallow stratum of a target tidal flat monitoring point; the rainfall sensor and the tide sensor are respectively arranged in a set radius range of a target tidal flat monitoring point; the controller is respectively connected with a node part at the tail part of the sensing array, the rainfall sensor and the tide sensor; the three-axis acceleration sensor is used for collecting three-axis acceleration of the corresponding node part at different moments; the controller is used for acquiring an original data set of a target tidal flat monitoring point; determining N groups of state sequences based on the feature sequences of the N data clusters, the rainfall sequence and the tide height sequence, and inputting each group of state sequences into different trained prediction models to obtain N deformation sequences; and based on the N deformation sequences, determining future deformation of the shallow stratum of the target tidal flat monitoring point. Effective prediction of future deformation of the tidal flat shallow stratum is realized.

Description

Technical Field

[0001] The present application relates to the technical field of stratum deformation prediction, and in particular to a device and method for predicting tidal flat shallow stratum deformation. Background Art

[0002] As one of the most common landforms in coastal areas, tidal flats have important economic, ecological and research values. In recent years, a large number of researchers have conducted in-depth research on the formation, protection and evolution of tidal flats.

[0003] As a natural landform affected by multi-scale factors, the tidal flat's stratigraphic deformation can reflect its complex evolution law. In particular, the deformation of the shallow strata of the tidal flat is often affected by the superposition of multiple environmental factors. The deformation prediction of the shallow strata of the tidal flat is conducive to understanding the generation mechanism and evolution mechanism of the tidal flat environment, and can also provide a scientific basis for predicting natural geological disasters.

[0004] Since the shallow strata of tidal flats have high water content, are easily affected by periodic tides, and change irregularly, general technologies cannot predict the deformation of shallow strata of tidal flats. Currently, commonly used technologies include LiDAR, GNSS, InSAR, etc., but these technologies are only suitable for specific monitoring of deformation of shallow strata of tidal flats, and cannot predict the future deformation state of shallow strata of tidal flats. Therefore, deformation prediction of shallow strata of tidal flats still faces huge challenges. Summary of the invention

[0005] The purpose of the present application is to provide a device and method for predicting deformation of shallow strata on tidal flats, which can effectively predict the future deformation of shallow strata on tidal flats.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a tidal flat shallow stratum deformation prediction device, the tidal flat shallow stratum deformation prediction device comprising: a sensor array, a rainfall sensor, a tide sensor and a controller; the sensor array comprises a plurality of node components; each of the node components is equipped with a triaxial acceleration sensor;

[0008] A plurality of node components are connected in sequence and vertically penetrate into the shallow stratum of the target tidal flat monitoring point; the rainfall sensor and the tidal sensor are respectively arranged within the set radius range of the target tidal flat monitoring point; the controller is respectively connected to the node component at the tail of the sensor array, the rainfall sensor and the tidal sensor;

[0009] The three-axis acceleration sensor is used to collect the three-axis acceleration of the corresponding node component at different times; the rainfall sensor is used to collect the rainfall at the target tidal flat monitoring point at different times; the tidal sensor is used to collect the tidal height at the target tidal flat monitoring point at different times; the controller is used to: obtain the original data set of the target tidal flat monitoring point within a set time period; the original data set includes: the original three-axis acceleration data set, the original rainfall data set and the original tidal height data set; the original three-axis acceleration data set includes the node three-axis acceleration at different times; the node three-axis acceleration includes the three-axis acceleration of different node components; the original rainfall data set includes the rainfall at different times; the original tidal height data set includes the tidal height at different times; based on the original three-axis acceleration data set, the node displacement deformation at different times is calculated to obtain the displacement deformation data set; the node displacement deformation includes the displacement deformation of different node components; the displacement deformation data set is hierarchically clustered to obtain N data clusters, perform principal component analysis on each data cluster respectively, select the principal component components with contribution greater than a set threshold and calculate the average value of the principal component components to obtain the characteristic sequences of the N data clusters; N is an integer greater than 0; perform cumulative calculation on the original rainfall data set, and correct the sequence length after the cumulative calculation to obtain a rainfall sequence; perform sliding average calculation on the original tide height data set, and correct the sequence length after the sliding average calculation to obtain a tide height sequence; determine N groups of state sequences based on the characteristic sequences of the N data clusters, the rainfall sequence and the tide height sequence, and input each group of state sequences into different trained prediction models respectively to obtain N deformation sequences; a group of state sequences includes: a characteristic sequence of a data cluster, the rainfall sequence and the tide height sequence; a group of state sequences corresponds to a trained prediction model; based on the N deformation sequences, determine the future deformation of the shallow stratum at the target tidal flat monitoring point; a deformation sequence corresponds to the deformation value of a stratum.

[0010] In a second aspect, the present application further provides a method for predicting deformation of a tidal flat shallow stratum, the method for predicting deformation of a tidal flat shallow stratum being applied to the device for predicting deformation of a tidal flat shallow stratum described in the first aspect, the method for predicting deformation of a tidal flat shallow stratum comprising:

[0011] Acquire the original data set of the target tidal flat monitoring point within the set time period; the original data set includes: an original triaxial acceleration data set, an original rainfall data set and an original tide height data set; the original triaxial acceleration data set includes the triaxial acceleration of nodes at different times; the triaxial acceleration of nodes includes the triaxial acceleration of different node components; the original rainfall data set includes the rainfall at different times; the original tide height data set includes the tide height at different times;

[0012] Based on the original triaxial acceleration data set, the node displacement deformation at different times is calculated to obtain a displacement deformation data set; the node displacement deformation includes the displacement deformation of different node components;

[0013] Hierarchical clustering of the displacement deformation data set is performed to obtain N data clusters, principal component analysis is performed on each data cluster, principal component components whose contribution is greater than a set threshold are selected, and the average value of the principal component components is calculated to obtain a characteristic sequence of the N data clusters; N is an integer greater than 0;

[0014] Performing cumulative calculation on the original rainfall data set, and correcting the length of the sequence after the cumulative calculation, to obtain a rainfall sequence;

[0015] Performing sliding average calculation on the original tide height data set, and correcting the sequence length after the sliding average calculation, to obtain a tide height sequence;

[0016] Based on the characteristic sequences of N data clusters, the rainfall sequence and the tidal height sequence, N groups of state sequences are determined, and each group of state sequences is input into different trained prediction models to obtain N deformation sequences; a group of state sequences includes: a characteristic sequence of a data cluster, the rainfall sequence and the tidal height sequence; a group of state sequences corresponds to a trained prediction model;

[0017] Based on N deformation sequences, the future deformation of the shallow strata at the target tidal flat monitoring point is determined; one deformation sequence corresponds to the deformation value of one stratum.

[0018] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0019] The present application effectively collects the triaxial acceleration data of different node components by inserting a sensor array into the shallow stratum of the target tidal flat monitoring point, and further calculates the displacement deformation of different node components, thereby obtaining a displacement deformation data set characterizing the deformation of different strata. In order to realize the division of deformation data of different strata, the present application also divides the displacement deformation data set into N data clusters by hierarchical clustering, and analyzes the principal component of each data cluster, and finally obtains the characteristic sequence of N data clusters, i.e., the characteristic sequence corresponding to each stratum. In order to improve the effectiveness of the prediction of deformation of the shallow stratum of the tidal flat, the present application also collects the environmental data of the target tidal flat monitoring point by laying out rainfall sensors and tidal sensors, and obtains the rainfall sequence and the tidal height sequence after further calculation and correction. Finally, it is only necessary to combine the rainfall sequence and the tidal height sequence with the characteristic sequence of each data cluster, and respectively input them into the corresponding trained prediction model, so as to obtain the deformation sequence characterizing the future deformation of each stratum, and all deformation sequences jointly characterize the future deformation of the shallow stratum of the target tidal flat monitoring point. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 A layout diagram of a tidal flat shallow stratum deformation prediction device provided in an embodiment of the present application;

[0022] Figure 2 A structural diagram of a node component provided in an embodiment of the present application;

[0023] Figure 3 A flow chart of a method for predicting deformation of shallow tidal flats provided in an embodiment of the present application.

[0024] Explanation of symbols:

[0025] Sensor array-1, node component-11, cabin-111, hatch cover-112, controller-2. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0027] The purpose of the present application is to provide a device and method for predicting deformation of shallow strata on tidal flats, which can effectively predict the future deformation of shallow strata on tidal flats.

[0028] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0029] Example 1

[0030] This embodiment provides a device for predicting deformation of shallow tidal flats. Figure 1 As shown, the tidal flat shallow layer deformation prediction device includes: a sensor array 1, a rainfall sensor (not shown in the figure), a tide sensor (not shown in the figure) and a controller 2; the sensor array 1 includes a plurality of node components 11; each node component 11 is equipped with a triaxial acceleration sensor, which is as follows:

[0031] Multiple node components 11 are connected in sequence and vertically penetrate into the shallow layer of the target tidal flat monitoring point. Figure 2As shown, the node component 11 further includes: a cabin 111 and a cabin cover 112; the cabin cover 112 is sealed and fixed to the cabin 111, and is used to protect the triaxial acceleration sensor inside the cabin.

[0032] As a preferred embodiment, a plurality of node components 11 are connected in series in sequence, and every two node components 11 are connected by a watertight cable. The length of the sensor array 1 is greater than 1m, and the number of node components 11 is greater than 3. In order to facilitate the vertical penetration of the sensor array 1 into the shallow stratum of the target tidal flat monitoring point, it is necessary to use a penetrometer and achieve it through manual assistance. Among them, the penetrometer is a hollow barrel-shaped structure, and a conical component is arranged at the head end of the structure. In the actual operation process, the sensor array 1 is first vertically inserted from the end of the structure, and then the structure is inserted into the shallow stratum of the target tidal flat monitoring point as a whole. After the sensor array 1 is stabilized inside the stratum, the structure can be directly pulled out. In addition, the cabin 111 and the hatch 112 should be made of 316 material to prevent the entire node component from rusting after the watertightness fails.

[0033] The rainfall sensor and the tide sensor are respectively arranged within the set radius range of the target tidal flat monitoring point; the controller 2 is respectively connected to the node component 11 at the tail of the sensor array, the rainfall sensor and the tide sensor.

[0034] As a preferred implementation, the rainfall sensor and the tidal sensor need to be set within a radius of 1km from the target tidal flat monitoring point, the rainfall sensor is set at the dock within a radius of 1km from the target tidal flat monitoring point, the tidal sensor is set on the tidal flat within a radius of 1km from the target tidal flat monitoring point, and the controller 2, the rainfall sensor and the tidal sensor are all connected by a watertight cable. In addition, the tidal sensor has the function of monitoring the in-situ tide height, and the in-situ monitoring data can be stored locally in a variety of ways. The operating frequencies of the rainfall sensor, the tidal sensor and the three-axis acceleration sensor in each node component 11 are consistent and all higher than 1Hz, and their single working time is greater than or equal to one week.

[0035] Furthermore, the three-axis acceleration sensor is used to collect the three-axis acceleration of the corresponding node component at different times; the rainfall sensor is used to collect the rainfall at different times of the target tidal flat monitoring point (around); the tide sensor is used to collect the tide height at different times of the target tidal flat monitoring point (around).

[0036] The controller 2 is used to: obtain the original data set of the target tidal flat monitoring point within the set time period; the original data set includes: the original triaxial acceleration data set, the original rainfall data set and the original tide height data set; the original triaxial acceleration data set includes the triaxial acceleration of the node at different times; the triaxial acceleration of the node includes the triaxial acceleration of different node components; the original rainfall data set includes the rainfall at different times; the original tide height data set includes the tide height at different times; based on the original triaxial acceleration data set, calculate the node displacement deformation at different times to obtain the displacement deformation data set; the node displacement deformation includes the displacement deformation of different node components; hierarchically cluster the displacement deformation data set to obtain N data clusters, perform principal component analysis on each data cluster, select the principal component component whose contribution is greater than the set threshold and calculate the principal component component The average value is used to obtain the characteristic sequences of N data clusters; N is an integer greater than 0; the original rainfall data set is accumulated and the sequence length after the accumulation is corrected to obtain the rainfall sequence; the original tide height data set is sliding averaged and the sequence length after the sliding average is corrected to obtain the tide height sequence; N groups of state sequences are determined based on the characteristic sequences, rainfall sequences and tide height sequences of the N data clusters, and each group of state sequences is input into different trained prediction models to obtain N deformation sequences; a group of state sequences includes: a characteristic sequence, rainfall sequence and tide height sequence of a data cluster; a group of state sequences corresponds to a trained prediction model; based on N deformation sequences, the future deformation of the shallow stratum at the target tidal flat monitoring point is determined; a deformation sequence corresponds to the deformation value of a stratum.

[0037] As a preferred implementation, the controller 2 is also used to control the operating frequency and storage frequency of the sensor array 1, the rainfall sensor and the tide sensor, and store the original data set locally.

[0038] Example 2

[0039] This embodiment provides a method for predicting deformation of shallow strata on tidal flats, and the method for predicting deformation of shallow strata on tidal flats is applied to the above-mentioned device for predicting deformation of shallow strata on tidal flats. Figure 3 As shown in FIG. 1 , the tidal flat shallow layer deformation prediction method includes:

[0040] Step S1: Obtain the original data set of the target tidal flat monitoring point within a set time period.

[0041] In this embodiment, the original data set includes: an original triaxial acceleration data set, an original rainfall data set, and an original tidal height data set; the original triaxial acceleration data set A 0 =[a 0 ,a 1 ,...,a t ] includes the node triaxial acceleration a at different times t; The triaxial acceleration of the node includes the triaxial acceleration of different node components Original rainfall dataset P 0 =[p 0 ,p 1 ,...,p t ] including the rainfall at different times p t ; Original tide height dataset H 0 =[h 0 ,h 1 ,...,h t ] including the tidal height h at different times t .

[0042] In one example, in order to predict the deformation of the shallow strata at a target tidal flat monitoring point in the next week, it is necessary to obtain the original data set of the target tidal flat monitoring point this week and implement the prediction based on the data set.

[0043] Step S2: Based on the original triaxial acceleration data set, the node displacement deformation at different times is calculated to obtain a displacement deformation data set; the node displacement deformation includes the displacement deformation of different node components.

[0044] In this embodiment, step S2 specifically includes:

[0045] The first step is to calculate the node yaw angle at different times based on the original three-axis acceleration data set. The node yaw angle includes the yaw angles corresponding to different node components, and the calculation formula for the yaw angle is:

[0046]

[0047] In the formula, is the yaw angle of the jth node component at time t, is the acceleration of the jth node component in the X-axis direction at time t, and g is the gravitational acceleration.

[0048] The second step is to use the linear model to calculate the node displacement deformation at different times based on the node yaw angle at different times, specifically:

[0049]

[0050] In the formula, is the displacement deformation of the i-th node component at time t, is the X-axis deformation of the i-th node component at time t, is the Y-axis deformation of the i-th node component at time t, is the Z-axis deformation of the i-th node component at time t, l is the distance between node components, and n is the number of node components.

[0051] In addition, because the sensor array in this embodiment is vertically penetrated into the shallow stratum of the target tidal flat monitoring point, only the deformation in the X-axis direction and the Y-axis direction of each node component is considered, and the deformation in the Z-axis direction is ignored.

[0052] The third step is to obtain the displacement deformation data set based on the node displacement deformation at different times.

[0053] Step S3: hierarchical clustering of the displacement deformation data set is performed to obtain N data clusters, and principal component analysis is performed on each data cluster respectively, and the principal component components whose contribution is greater than the set threshold are selected and the average value of the principal component components is calculated to obtain the characteristic sequence of the N data clusters; N is an integer greater than 0.

[0054] As a preferred implementation, the number of data clusters obtained by hierarchical clustering is set to 3, that is, the number of data clusters corresponds to the number of layers in conventional shallow strata, the clustering distance is set to the Euclidean distance, and the number of leaf nodes of the hierarchical tree is set according to the size of the data set.

[0055] In this embodiment, step S3 specifically includes:

[0056] The first step is to divide the node displacement deformation at different times into the X-axis and Y-axis directions based on the node displacement data set to obtain the X-axis displacement data set and the Y-axis displacement data set. Since the displacement deformation in the Z-axis direction is not considered, it is sufficient to divide it only in the X-axis and Y-axis directions.

[0057] In the second step, hierarchical clustering of the X-axis displacement data set is performed to obtain N data clusters. Principal component analysis is performed on each data cluster, and the principal component components with a contribution greater than 90% are selected and the average of the principal component components is calculated to obtain the X-axis direction sequence of the N data clusters. Although the node components in the sensor matrix are interconnected, the displacement deformation of adjacent node components may be clustered into different data clusters due to different deformation laws. In addition, the contribution in each data cluster is calculated as follows:

[0058]

[0059] In the formula, CCR is the contribution value, λ k is the kth principal component, and n is the number of principal components (the same as the number of node components in the sensor matrix).

[0060] In the third step, the Y-axis displacement data set is hierarchically clustered to obtain N data clusters, and principal component analysis is performed on each data cluster. The principal component components with a contribution greater than 90% are selected and the average of the principal component components is calculated to obtain the Y-axis direction sequence of the N data clusters. The basic principle is the same as the processing method for the X-axis displacement data set in the second step, and will not be repeated here.

[0061] The fourth step is to obtain the feature sequences of N data clusters based on the X-axis sequence of N data clusters and the Y-axis sequence of N data clusters. The feature sequences of N data clusters finally obtained actually correspond to the deformation characteristics of each layer in the shallow layer, and this feature sequence plays a vital role in the deformation prediction of the shallow layer and the subsequent model training process.

[0062] Step S4: Perform cumulative calculation on the original rainfall data set and correct the length of the sequence after the cumulative calculation to obtain a rainfall sequence.

[0063] In this embodiment, the cumulative sequence obtained after cumulative calculation is recorded as:

[0064]

[0065] It should be noted that the length of the accumulated sequence is inconsistent with the original sequence, so it needs to be corrected to In order to ensure consistency with the original sequence length, the correction method is to expand the cumulative sequence by continuously copying the last value until it reaches the same length as the original sequence.

[0066] As a preferred implementation, when accumulating the data in the original rainfall data set, the selection of the accumulated data points is set between 24 hours and 36 hours.

[0067] Step S5: Perform sliding average calculation on the original tide height data set, and correct the sequence length after the sliding average calculation to obtain a tide height sequence.

[0068] As a preferred implementation, the sliding window size for sliding average calculation of the original tidal height data set is generally set to half of the local tidal cycle (6 hours for diurnal tide, 3 hours for semi-diurnal tide, and 3 hours for irregular tide). The correction method is also to expand the accumulated sequence by continuously copying the last value until it reaches the same length as the original sequence.

[0069] Step S6: Determine N groups of state sequences based on the feature sequences, rainfall sequences and tide height sequences of the N data clusters, and input each group of state sequences into different trained prediction models to obtain N deformation sequences; a group of state sequences includes: a feature sequence, rainfall sequence and tide height sequence of a data cluster; a group of state sequences corresponds to a trained prediction model.

[0070] As a preferred implementation, before executing step S6, it is necessary to obtain N trained prediction models, the number of trained prediction models, the number of data clusters obtained by hierarchical clustering, and the number of strata in the shallow strata of the target tidal flat monitoring point are the same, and the training process of each prediction model is also the same, as follows:

[0071] The first step is to obtain the original data sets of the target tidal flat monitoring points in different historical time periods. Generally, multiple sets of original data sets of adjacent historical time periods are obtained, for example, the original data sets of T-1 week, T-2 week, and T-3 week are obtained respectively.

[0072] In the second step, the same process as steps S2 to S5 above is performed on all original data sets, and the processed data is divided into a training set and a test set in a ratio of 8:2. The prediction model is trained on the training set, and the accuracy of the prediction model is verified on the test set.

[0073] The third step is to use the random forest regression model as the prediction model and set the parameters of the random forest regression model. The number of trees in the random forest regression model should be determined by the amount of training data. The default value is 100. The training stops when the number of training times reaches 1000 or the Gini coefficient is less than 1e -4 .

[0074] The fourth step is to train the random forest regression model. For example, the rainfall sequence, tide height sequence and feature sequence of the first data cluster obtained from the original data set of T-2 weeks are used as input, and the original sequence of the first data cluster obtained from the original data set of T-1 weeks is used as output to train the prediction model corresponding to the first data cluster (i.e., the first type layer in the shallow stratum). Similarly, the rainfall sequence, tide height sequence and feature sequence of the first data cluster obtained from the original data set of T-3 weeks are used as input, and the original sequence of the first data cluster obtained from the original data set of T-2 weeks is used as output to train the prediction model corresponding to the first data cluster (i.e., the first type layer in the shallow stratum). The above basic training principles are to use the rainfall sequence, tide height sequence and feature sequence of a single data cluster in the historical time period as input, and the original sequence of a single data cluster in the time period after the historical time period as output, and the prediction model corresponding to each data cluster (i.e., each type layer in the shallow stratum) is trained in this way.

[0075] The fifth step is to verify the training accuracy of the random forest regression model. The calculation indicators of accuracy verification mainly include: root mean square error (RMSE), absolute error (MAE) and determination coefficient (R2). With the goal of RMSE and MAE close to 0 and R2 close to 1, the accuracy of the trained random forest regression model is tuned.

[0076] Step S7: Based on the N deformation sequences, determine the future deformation of the shallow strata at the target tidal flat monitoring point.

[0077] In this embodiment, a deformation sequence corresponds to a deformation value of a stratum. By observing the deformation values ​​corresponding to different deformation sequences, the deformation state of each type of stratum in the shallow stratum of the target tidal flat monitoring point can be determined, and the deformation state is the future deformation of the shallow stratum of the tidal flat.

[0078] In summary, the effective prediction of the future deformation of the shallow strata of the tidal flat in this application facilitates scientific researchers to understand the generation mechanism and evolution mechanism of the tidal flat environment, and provides a strong scientific basis for the prediction of natural geological disasters.

[0079] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0080] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A tidal flat shallow layer deformation prediction device, characterized in that: The tidal flat shallow layer deformation prediction device comprises: a sensor array, a rainfall sensor, a tide sensor and a controller; the sensor array comprises a plurality of node components; each of the node components is equipped with a triaxial acceleration sensor; A plurality of node components are connected in sequence and vertically penetrate into the shallow stratum of the target tidal flat monitoring point; the rainfall sensor and the tidal sensor are respectively arranged within the set radius range of the target tidal flat monitoring point; the controller is respectively connected to the node component at the tail of the sensor array, the rainfall sensor and the tidal sensor; The triaxial acceleration sensor is used to collect triaxial acceleration of the corresponding node component at different times; the rainfall sensor is used to collect rainfall at the target tidal flat monitoring point at different times; the tidal sensor is used to collect tidal height at the target tidal flat monitoring point at different times; the controller is used to: Acquire the original data set of the target tidal flat monitoring point within the set time period; the original data set includes: an original triaxial acceleration data set, an original rainfall data set and an original tide height data set; the original triaxial acceleration data set includes the triaxial acceleration of nodes at different times; the triaxial acceleration of nodes includes the triaxial acceleration of different node components; the original rainfall data set includes the rainfall at different times; the original tide height data set includes the tide height at different times; Based on the original triaxial acceleration data set, the node displacement deformation at different times is calculated to obtain a displacement deformation data set; the node displacement deformation includes the displacement deformation of different node components; Hierarchical clustering of the displacement deformation data set is performed to obtain N data clusters, principal component analysis is performed on each data cluster, principal component components whose contribution is greater than a set threshold are selected, and the average value of the principal component components is calculated to obtain a characteristic sequence of the N data clusters; N is an integer greater than 0; Performing cumulative calculation on the original rainfall data set, and correcting the length of the sequence after the cumulative calculation, to obtain a rainfall sequence; Performing sliding average calculation on the original tide height data set, and correcting the sequence length after the sliding average calculation, to obtain a tide height sequence; Based on the characteristic sequences of N data clusters, the rainfall sequence and the tidal height sequence, N groups of state sequences are determined, and each group of state sequences is input into different trained prediction models to obtain N deformation sequences; a group of state sequences includes: a characteristic sequence of a data cluster, the rainfall sequence and the tidal height sequence; a group of state sequences corresponds to a trained prediction model; Based on N deformation sequences, the future deformation of the shallow strata at the target tidal flat monitoring point is determined; one deformation sequence corresponds to the deformation value of one stratum.

2. The device for predicting deformation of shallow tidal flats according to claim 1, characterized in that: The controller is also used to control the operating frequency and storage frequency of the sensor array, the rainfall sensor and the tide sensor.

3. The device for predicting deformation of shallow tidal flats according to claim 1, characterized in that: The node component further includes: a cabin and a cabin cover; the cabin cover is sealed and fixed to the cabin and is used to protect the triaxial acceleration sensor inside the cabin.

4. The device for predicting deformation of shallow tidal flats according to claim 1, characterized in that: The length of the sensor array is greater than 1 m; the number of the node components is greater than 3.

5. The device for predicting deformation of shallow tidal flats according to claim 1, characterized in that: The node components are connected by watertight cables.

6. A method for predicting deformation of shallow tidal flats, characterized in that: The method for predicting deformation of shallow strata on tidal flats is applied to the device for predicting deformation of shallow strata on tidal flats according to any one of claims 1 to 5, and the method for predicting deformation of shallow strata on tidal flats comprises: Acquire the original data set of the target tidal flat monitoring point within the set time period; the original data set includes: an original triaxial acceleration data set, an original rainfall data set and an original tide height data set; the original triaxial acceleration data set includes the triaxial acceleration of nodes at different times; the triaxial acceleration of nodes includes the triaxial acceleration of different node components; the original rainfall data set includes the rainfall at different times; the original tide height data set includes the tide height at different times; Based on the original triaxial acceleration data set, the node displacement deformation at different times is calculated to obtain a displacement deformation data set; the node displacement deformation includes the displacement deformation of different node components; Hierarchical clustering of the displacement deformation data set is performed to obtain N data clusters, principal component analysis is performed on each data cluster, principal component components whose contribution is greater than a set threshold are selected, and the average value of the principal component components is calculated to obtain a characteristic sequence of the N data clusters; N is an integer greater than 0; Performing cumulative calculation on the original rainfall data set, and correcting the length of the sequence after the cumulative calculation, to obtain a rainfall sequence; Performing sliding average calculation on the original tide height data set, and correcting the sequence length after the sliding average calculation, to obtain a tide height sequence; Based on the characteristic sequences of N data clusters, the rainfall sequence and the tidal height sequence, N groups of state sequences are determined, and each group of state sequences is input into different trained prediction models to obtain N deformation sequences; a group of state sequences includes: a characteristic sequence of a data cluster, the rainfall sequence and the tidal height sequence; a group of state sequences corresponds to a trained prediction model; Based on N deformation sequences, the future deformation of the shallow strata at the target tidal flat monitoring point is determined; one deformation sequence corresponds to the deformation value of one stratum.

7. The method for predicting deformation of shallow tidal flats according to claim 6, characterized in that: Based on the original triaxial acceleration data set, the node displacement deformation at different times is calculated to obtain a displacement deformation data set, which specifically includes: Based on the original three-axis acceleration data set, calculating the node yaw angle at different times; the node yaw angle includes the yaw angles corresponding to different node components; Based on the node yaw angles at different times, the node displacement deformation at different times is calculated using a linear model; Based on the node displacement deformation at different times, a displacement deformation data set is obtained.

8. The method for predicting deformation of shallow tidal flats according to claim 7, characterized in that: The straight line model is: In the formula, is the displacement deformation of the i-th node component at time t, is the X-axis deformation of the i-th node component at time t, is the Y-axis deformation of the i-th node component at time t, is the Z-axis deformation of the i-th node component at time t, is the yaw angle of the jth node component at time t, l is the distance between node components, and n is the number of node components.

9. The method for predicting deformation of shallow tidal flats according to claim 6, characterized in that: The displacement deformation data set is hierarchically clustered to obtain N data clusters, and principal component analysis is performed on each data cluster. The principal component components whose contribution is greater than a set threshold are selected and the average value of the principal component components is calculated to obtain the characteristic sequence of the N data clusters, which specifically includes: Based on the node displacement data set, the node displacement deformation at different times is divided in the X-axis and Y-axis directions to obtain an X-axis displacement data set and a Y-axis displacement data set; Hierarchical clustering of the X-axis displacement data set is performed to obtain N data clusters, principal component analysis is performed on each data cluster, principal component components with a contribution greater than 90% are selected, and the average value of the principal component components is calculated to obtain the X-axis direction sequence of the N data clusters; Hierarchical clustering of the Y-axis displacement data set is performed to obtain N data clusters, principal component analysis is performed on each data cluster, principal component components with a contribution greater than 90% are selected, and the average value of the principal component components is calculated to obtain the Y-axis direction sequence of the N data clusters; Based on the X-axis direction sequences of the N data clusters and the Y-axis direction sequences of the N data clusters, the feature sequences of the N data clusters are obtained.

10. The method for predicting deformation of shallow tidal flats according to claim 6, characterized in that: The prediction model is a random forest regression model.

Images