Underground water monitoring water level data restoration method

Abstract

The invention provides an underground water monitoring water level data recovery method, and belongs to the technical field of underground water monitoring of geological exploration. The problems of low recovery efficiency and low accuracy of existing underground water monitoring water level data are solved. The groundwater monitoring water level data restoration method comprises the steps of data reading, clustering analysis, vacancy detection, data segmentation, data inspection, data interpolation, data restoration and an ending stage to achieve restoration of missing data. According to the method, sudden change data and missing data of underground water monitoring water level data are restored mainly through the six steps of clustering analysis, vacancy detection, data segmentation, data inspection, data interpolation and data restoration, and the data restoration efficiency and the data restoration accuracy are remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field of groundwater monitoring in geological survey, and particularly relates to a method for restoring groundwater monitoring water level data. Background Art

[0002] Groundwater monitoring is the monitoring of groundwater levels, water quality and other data within the jurisdiction by groundwater monitoring management departments in order to timely grasp the dynamic changes and provide long-term protection for groundwater. The preliminary work of groundwater monitoring generally requires the establishment of long-term monitoring points through drilling. During the long-term groundwater monitoring process, due to the complex law of groundwater level changes and the great influence of human intervention on water level changes, groundwater monitoring water level data often changes suddenly or even disappears.

[0003] At present, traditional groundwater monitoring water level data often rely on monitoring personnel to restore groundwater monitoring water level data mutations and groundwater monitoring water level data missing based on previous monitoring operation logs and expert experience. This data restoration method is very inefficient, and the accuracy of the restored data varies greatly among different monitoring personnel, resulting in low accuracy of the restored data. Summary of the invention

[0004] The purpose of the present invention is to address the above-mentioned problems in the existing technology and propose a method for restoring groundwater monitoring water level data. The technical problem to be solved by the present invention is: how to solve the problem of low efficiency and low accuracy of existing groundwater monitoring water level data restoration.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A method for restoring groundwater monitoring water level data, characterized in that it comprises the following steps:

[0007] A. Data reading: Groundwater monitoring personnel read the atmospheric pressure data and groundwater pressure data of the monitoring site from the monitoring wells. After the atmospheric pressure data and groundwater pressure data are converted, the groundwater monitoring water level data of the monitoring well is obtained;

[0008] B. Cluster analysis: Taking each natural day as the time unit, the groundwater monitoring water level data of the monitoring well is clustered using the DBSCAN algorithm to obtain the noise data and one or more clusters of groundwater monitoring water level data determined by the clustering algorithm. First, all the noise data determined by the clustering algorithm are deleted, and then it is determined in turn whether the proportion of the data volume of each cluster of groundwater monitoring water level data to the data volume of the remaining groundwater monitoring water level data is higher than the given proportion. If the data volume of the cluster of groundwater monitoring water level data is lower than the given proportion, the cluster of groundwater monitoring water level data is also deleted;

[0009] C. Gap detection: Perform gap detection on the groundwater monitoring water level data that has completed the cluster analysis, determine the mutation position and missing position of the groundwater monitoring water level data of the monitoring well, and obtain the corresponding date of the groundwater monitoring data mutation and the corresponding date of the data missing of the monitoring well;

[0010] D. Data segmentation: According to the vacancy detection results, the groundwater monitoring water level data of the monitoring well is divided into several time periods of complete groundwater monitoring water level data and several time periods of groundwater monitoring water level data to be restored, thereby obtaining several complete monitoring water level time series data and several monitoring water level time series data to be restored;

[0011] E. Data test: Perform unit root test on the first section of water level monitoring time series data to determine whether the water level monitoring time series data meets the application conditions of the autoregressive moving average model. If so, perform data restoration. If not, perform data interpolation.

[0012] F. Data interpolation: according to the monitoring water level time sequence of the monitoring well, the first section of monitoring water level time series data and the second section of monitoring water level time series data are used as input, and the restoration result of the first section of monitoring water level time series data to be restored is obtained by linear interpolation, and then the restoration results of the first section of monitoring water level time series data, the second section of monitoring water level time series data and the first section of monitoring water level time series data to be restored are combined as the new first section of monitoring water level time series data. If the length of the first section of monitoring water level time series data is equal to the total monitoring time length of the monitoring well, then enter the end stage step, otherwise return to the data verification step;

[0013] G. Data restoration: according to the monitoring time sequence of the monitoring well, the first section of the monitoring water level time series data is used as input, and the restoration result of the first section of the monitoring water level time series data to be restored is obtained by the ARIMA algorithm, and then the restoration results of the first section of the monitoring water level time series data, the second section of the monitoring water level time series data and the first section of the monitoring water level time series data to be restored are combined as the new first section of the monitoring water level time series data. If the length of the first section of the monitoring water level time series data is equal to the total monitoring time length of the monitoring well, then enter the end stage step, otherwise return to the data verification step;

[0014] H. Ending stage: Through visualization technology, the real groundwater monitoring water level data of the monitoring well and the groundwater monitoring water level restoration data of the monitoring well obtained by this method are displayed within the same map range, and the groundwater monitoring water level restoration data file and data comparison map of the monitoring well are output.

[0015] This method mainly restores the mutation data and missing data of groundwater monitoring water level data through six steps: cluster analysis, vacancy detection, data segmentation, data verification, data interpolation and data restoration. In the cluster analysis, all groundwater monitoring water level data are clustered by the DBSCAN algorithm, and the noise data and data that do not meet the given ratio determined by the DBSCAN algorithm are screened out. The data that meets the given ratio in the DBSCAN algorithm is retained, and these partial interference data are excluded, which is conducive to the next step of vacancy detection to accurately detect the mutation position and data missing position of groundwater monitoring data. The data segmentation step is to segment the data according to the data mutation position and data missing position to obtain several complete monitoring water level time series data and the corresponding several monitoring water levels to be restored. Time series data, that is, each complete monitoring water level time series data corresponds to a monitoring water level time series data to be restored; in the data interpolation step, a restoration result is obtained through two sections of monitoring water level time series data, and then the restoration result is merged with the two sections of monitoring water level time series data to obtain one section of monitoring water level time series data. This method ensures that the data restoration has a high accuracy. The restoration result of the first section of monitoring water level time series data to be restored is obtained by the ARIMA algorithm, and then the restoration result is merged with the two sections of monitoring water level time series data. The restoration result is merged multiple times in step F and step G to obtain the final restoration result, thereby improving the accuracy of the restoration result. Therefore, this method significantly improves the data restoration efficiency and the accuracy of data restoration.

[0016] In the above-mentioned method for restoring groundwater monitoring water level data, in the above-mentioned step B, the groundwater monitoring water level data of the monitoring well is clustered using the DBSCAN algorithm with each natural day as the time unit, specifically:

[0017] a. Input all groundwater monitoring water level data of the monitoring well;

[0018] b. Input the size of the ε neighborhood defined by the DBSCAN algorithm and the minimum number of data points MinPts contained in the ε neighborhood;

[0019] c. Mark all groundwater monitoring water level data as not yet accessed by the DBSCAN algorithm;

[0020] d. Traverse each groundwater monitoring water level data. If the groundwater monitoring data is marked as unaccessed, perform the following steps:

[0021] (1) Mark the groundwater monitoring water level data as being in access status;

[0022] (2) Find all water level data within the ε neighborhood of the groundwater monitoring water level data;

[0023] (3) If the number of all groundwater monitoring water level data points within the ε neighborhood is greater than or equal to MinPts, then the groundwater monitoring water level data point is the core point;

[0024] (4) If the number of all groundwater monitoring water level data points within the ε neighborhood is less than MinPts, but the point is within the ε neighborhood of other core points, then the groundwater monitoring water level data point is a boundary point;

[0025] (5) If the groundwater monitoring water level data does not belong to the core point or the boundary point, the groundwater monitoring water level data point is a noise point;

[0026] (6) Repeat the above steps (1) to (5) until all groundwater monitoring water level data are marked as accessed;

[0027] e. Traverse each groundwater monitoring water level data. If the groundwater monitoring water level data is marked as a core point, perform the following steps:

[0028] (1) If point p and point q are both within each other's ε neighborhood, then p and q are said to be density-reachable to each other;

[0029] (2) Starting from any core point, recursively add all points that are mutually density-reachable to the same cluster;

[0030] (3) If all the groundwater monitoring water level data belonging to the cluster have been classified, start by looking for another unclassified core point;

[0031] (4) Repeat the above steps (2) and (3) until all core points are classified into different clusters;

[0032] f. Add all boundary points to the cluster where the core point that makes it a boundary point is located;

[0033] g. Output the different clusters of groundwater monitoring water level data and the identified noise data.

[0034] In the above-mentioned method for restoring groundwater monitoring water level data, in the above-mentioned step G, the specific steps of the ARIMA algorithm for restoring the monitoring water level time series data are:

[0035] a. Input the first monitoring water level time series data of the monitoring well;

[0036] b. Perform ADF test on the monitoring water level time series data or its different order d difference results to determine which order difference is a stationary time series;

[0037] c. Calculate the autocorrelation coefficient ACF and partial autocorrelation coefficient PACF for the stationary difference results, and analyze to obtain the optimal p and q;

[0038] d. Use nonlinear least squares (NLS) or other optimization algorithms to estimate the ARIMA (p, d, q) model;

[0039] e. The fitting results of the two test models were tested according to the Akaike information criterion and the Bayesian information criterion;

[0040] f. Use methods such as the Ljung-Box test to check whether the residual sequence has white noise characteristics. If the residual does not belong to white noise, the ARIMA model needs to be readjusted;

[0041] g. Predict the first section of the monitored water level time series data to be restored based on the adjusted model;

[0042] h. Update model parameters to improve prediction accuracy after new data points appear;

[0043] i. Output the model and the restoration results of the first section of the monitored water level time series data to be restored.

[0044] In step G, the ARIMA model is continuously adjusted and updated to improve the accuracy of data restoration.

[0045] Compared with the prior art, the groundwater monitoring water level data restoration method of the present invention has the following advantages: this method is based on the groundwater level time series data obtained by analyzing the atmospheric pressure of the groundwater monitoring area and the groundwater pressure of the monitoring well, and studies the principle of the DBSCAN density clustering algorithm to optimize the hyperparameters for the groundwater level time series data, eliminate the spatiotemporal interference data in the data that are not related to the monitoring target, and iteratively select complete groundwater level time series data and time-missing data according to the principle of time continuity. By analyzing the characteristics of the ARIMA data restoration algorithm and the data interpolation algorithm, a spatiotemporal continuous groundwater monitoring data restoration autoregressive moving average model is constructed. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of the method of the present invention.

[0047] Figure 2 It is a flow chart of the subdivided steps of step B of the present invention.

[0048] Figure 3 It is a flow chart of the subdivided steps of step G of the present invention. DETAILED DESCRIPTION

[0049] The following are specific embodiments of the present invention and the accompanying drawings to further describe the technical solution of the present invention, but the present invention is not limited to these embodiments.

[0050] A method for restoring groundwater monitoring water level data comprises the following steps:

[0051] A. Data reading: Groundwater monitoring personnel read the atmospheric pressure data and groundwater pressure data of the monitoring site from the monitoring wells. After the atmospheric pressure data and groundwater pressure data are converted, the groundwater monitoring water level data of the monitoring well is obtained;

[0052] B. Cluster analysis: Taking each natural day as the time unit, the groundwater monitoring water level data of the monitoring well is clustered using the DBSCAN algorithm to obtain noise data and one or more clusters of groundwater monitoring water level data determined by the clustering algorithm. The DBSCAN algorithm is a density-based spatial clustering algorithm. The core idea of ​​the algorithm is: for a given data set, first find the core object, that is, the point containing at least the minimum number of points MinPts within a given radius ε. Starting from these core objects, the closely connected core objects are classified into the same cluster through the density accessibility relationship. For those points that are not core objects, if they are in the neighborhood of the core object, they are also assigned to the corresponding cluster, and the points that do not belong to any cluster are regarded as noise. First, delete all noise data determined by the clustering algorithm, and then determine whether the proportion of the data volume of each cluster of groundwater monitoring water level data to the data volume of the remaining groundwater monitoring water level data is higher than the given proportion. If the data volume of the cluster of groundwater monitoring water level data is lower than the given proportion, the cluster of groundwater monitoring water level data is also deleted; take each natural day as the time unit, and use the DBSCAN algorithm to cluster the groundwater monitoring water level data of the monitoring well, specifically:

[0053] a. Input all groundwater monitoring water level data of the monitoring well;

[0054] b. Input the size of the ε neighborhood defined by the DBSCAN algorithm and the minimum number of data points MinPts contained in the ε neighborhood;

[0055] c. Mark all groundwater monitoring water level data as not yet accessed by the DBSCAN algorithm;

[0056] d. Traverse each groundwater monitoring water level data. If the groundwater monitoring data is marked as unaccessed, perform the following steps:

[0057] (1) Mark the groundwater monitoring water level data as being in access status;

[0058] (2) Find all water level data within the ε neighborhood of the groundwater monitoring water level data;

[0059] (3) If the number of all groundwater monitoring water level data points within the ε neighborhood is greater than or equal to MinPts, then the groundwater monitoring water level data point is the core point;

[0060] (4) If the number of all groundwater monitoring water level data points within the ε neighborhood is less than MinPts, but the point is within the ε neighborhood of other core points, then the groundwater monitoring water level data point is a boundary point;

[0061] (5) If the groundwater monitoring water level data does not belong to the core point or the boundary point, the groundwater monitoring water level data point is a noise point;

[0062] (6) Repeat the above steps (1) to (5) until all groundwater monitoring water level data are marked as accessed;

[0063] e. Traverse each groundwater monitoring water level data. If the groundwater monitoring water level data is marked as a core point, perform the following steps:

[0064] (1) If point p and point q are both within each other's ε neighborhood, then p and q are said to be density-reachable to each other;

[0065] (2) Starting from any core point, recursively add all points that are mutually density-reachable to the same cluster;

[0066] (3) If all the groundwater monitoring water level data belonging to the cluster have been classified, start by looking for another unclassified core point;

[0067] (4) Repeat the above steps (2) and (3) until all core points are classified into different clusters;

[0068] f. Add all boundary points to the cluster where the core point that makes it a boundary point is located;

[0069] g. Output the different clusters of groundwater monitoring water level data and the identified noise data.

[0070] C. Gap detection: Perform gap detection on the groundwater monitoring water level data that has completed the cluster analysis, determine the mutation position and missing position of the groundwater monitoring water level data of the monitoring well, and obtain the corresponding date of the groundwater monitoring data mutation and the corresponding date of the data missing of the monitoring well;

[0071] D. Data segmentation: According to the vacancy detection results, the groundwater monitoring water level data of the monitoring well is divided into several time periods of complete groundwater monitoring water level data and several time periods of groundwater monitoring water level data to be restored, thereby obtaining several complete monitoring water level time series data and several monitoring water level time series data to be restored;

[0072] E. Data test: Perform unit root test on the first section of water level monitoring time series data to determine whether the water level monitoring time series data meets the application conditions of the autoregressive moving average model. If so, perform data restoration. If not, perform data interpolation.

[0073] F. Data interpolation: according to the monitoring water level time sequence of the monitoring well, the first section of monitoring water level time series data and the second section of monitoring water level time series data are used as input, and the restoration result of the first section of monitoring water level time series data to be restored is restored by linear interpolation. Linear interpolation refers to the interpolation method in which the interpolation function is a first-order polynomial, which can be used to approximately replace the original function or calculate the value not in the table during the table lookup process. Then, the restoration results of the first section of monitoring water level time series data, the second section of monitoring water level time series data and the first section of monitoring water level time series data to be restored are combined as the new first section of monitoring water level time series data. If the length of the first section of monitoring water level time series data is equal to the total monitoring time length of the monitoring well, the end stage step is entered, otherwise the data verification step is returned;

[0074] G. Data restoration: According to the monitoring time sequence of the monitoring well, the first monitoring water level time series data is taken as input, and the restoration result of the first monitoring water level time series data to be restored is obtained through the ARIMA model. The ARIMA model, which is called the autoregressive integrated moving average model, refers to the model established by regressing the dependent variable only on its lag value and the present value and lag value of the random error term in the process of converting the non-stationary time series into a stationary time series. Then, the restoration results of the first monitoring water level time series data, the second monitoring water level time series data and the first monitoring water level time series data to be restored are combined as the new first monitoring water level time series data. If the length of the first monitoring water level time series data is equal to the total monitoring time length of the monitoring well, the end stage step is entered, otherwise it returns to the data verification step; the specific steps of restoring the ARIMA model of the monitoring water level time series data are as follows:

[0075] a. Input the first monitoring water level time series data of the monitoring well;

[0076] b. Perform ADF test on the monitoring water level time series data or its different d-order difference calculation results to determine which order difference is a stable time series. ADF test is a method specifically used for testing the stability of time series data. In simple terms, it is used to determine whether the data fluctuates around a mean value, and this mean value has nothing to do with time. If the P value is greater than 0.05, it means that the series is not stable. If the P value is less than 0.05, it means that the series is stable.

[0077] c. Calculate the autocorrelation coefficient ACF and partial autocorrelation coefficient PACF for the stationary difference results, and analyze to obtain the optimal class p and order q;

[0078] d. Use nonlinear least squares (NLS) or other optimization algorithms to estimate the ARIMA (p, d, q) model. Nonlinear least squares (NLS) is used to build regression models for data sets containing nonlinear features. In NLS, our goal is to find the model parameter vector β to minimize the sum of squares of the residuals.

[0079] e. The fitting results of the two test models were tested according to the Akaike information criterion and the Bayesian information criterion;

[0080] f. Use methods such as the Ljung-Box test to check whether the residual sequence has white noise characteristics. If the residual does not belong to white noise, the ARIMA model needs to be readjusted. The Ljung-Box test is used to test whether a set of data has time series correlation. This method is based on the autocorrelation function ACF and partial autocorrelation function PACF of time series data. By testing whether the autocorrelation of the sequence is significant, it is determined whether the sequence has significant correlation;

[0081] g. Predict the first section of the monitored water level time series data to be restored based on the adjusted model;

[0082] h. Update model parameters to improve prediction accuracy after new data points appear;

[0083] i. Output the model and the restoration results of the first section of the monitored water level time series data to be restored.

[0084] H. Ending stage: Through visualization technology, the real groundwater monitoring water level data of the monitoring well and the groundwater monitoring water level restoration data of the monitoring well obtained by this method are displayed within the same map range, and the groundwater monitoring water level restoration data file and data comparison map of the monitoring well are output.

[0085] This method mainly restores the mutation data and missing data of groundwater monitoring water level data through six steps: cluster analysis, vacancy detection, data segmentation, data verification, data interpolation and data restoration. In the cluster analysis, all groundwater monitoring water level data are clustered by the DBSCAN algorithm, and the noise data and data that do not meet the given ratio determined by the DBSCAN algorithm are screened out. The data that meets the given ratio in the DBSCAN algorithm is retained, and these partial interference data are excluded, which is conducive to the next step of vacancy detection to accurately detect the mutation position and data missing position of groundwater monitoring data. The data segmentation step is to segment the data according to the data mutation position and data missing position to obtain a number of complete monitoring water level time series data and a number of corresponding monitoring water level time series data to be restored. That is, each complete monitoring water level time series data corresponds to a monitoring water level time series data to be restored; in the data interpolation step, a restoration result is obtained through two sections of monitoring water level time series data, and then the restoration result is merged with the two sections of monitoring water level time series data to obtain one section of monitoring water level time series data. This method ensures that the data restoration has a high accuracy. The restoration result of the first section of monitoring water level time series data to be restored is obtained by the ARIMA algorithm, and then the restoration result is merged with the two sections of monitoring water level time series data. The restoration result is merged multiple times in step F and step G to obtain the final restoration result, and the ARIMA model is continuously adjusted and updated to improve the accuracy of the restoration result. Therefore, this method significantly improves the data restoration efficiency and the accuracy of data restoration.

[0086] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A method for restoring groundwater monitoring water level data, characterized in that: The following steps are involved: A. Data reading: Groundwater monitoring personnel read the atmospheric pressure data and groundwater pressure data of the monitoring site from the monitoring wells. After the atmospheric pressure data and groundwater pressure data are converted, the groundwater monitoring water level data of the monitoring well is obtained; B. Cluster analysis: Taking each natural day as the time unit, the groundwater monitoring water level data of the monitoring well is clustered using the DBSCAN algorithm to obtain the noise data and one or more clusters of groundwater monitoring water level data determined by the clustering algorithm. First, all the noise data determined by the clustering algorithm are deleted, and then it is determined in turn whether the proportion of the data volume of each cluster of groundwater monitoring water level data to the data volume of the remaining groundwater monitoring water level data is higher than the given proportion. If the data volume of the cluster of groundwater monitoring water level data is lower than the given proportion, the cluster of groundwater monitoring water level data is also deleted; C. Gap detection: Perform gap detection on the groundwater monitoring water level data that has completed the cluster analysis, determine the mutation position and missing position of the groundwater monitoring water level data of the monitoring well, and obtain the corresponding date of the groundwater monitoring data mutation and the corresponding date of the data missing of the monitoring well; D. Data segmentation: According to the vacancy detection results, the groundwater monitoring water level data of the monitoring well is divided into several time periods of complete groundwater monitoring water level data and several time periods of groundwater monitoring water level data to be restored, thereby obtaining several complete monitoring water level time series data and several monitoring water level time series data to be restored; E. Data test: Perform unit root test on the first section of water level monitoring time series data to determine whether the water level monitoring time series data meets the application conditions of the autoregressive moving average model. If so, perform data restoration. If not, perform data interpolation. F. Data interpolation: according to the monitoring water level time sequence of the monitoring well, the first section of monitoring water level time series data and the second section of monitoring water level time series data are used as input, and the restoration result of the first section of monitoring water level time series data to be restored is obtained by linear interpolation, and then the restoration results of the first section of monitoring water level time series data, the second section of monitoring water level time series data and the first section of monitoring water level time series data to be restored are combined as the new first section of monitoring water level time series data. If the length of the first section of monitoring water level time series data is equal to the total monitoring time length of the monitoring well, then enter the end stage step, otherwise return to the data verification step; G. Data restoration: according to the monitoring time sequence of the monitoring well, the first section of the monitoring water level time series data is used as input, and the restoration result of the first section of the monitoring water level time series data to be restored is obtained through the ARIMA model, and then the restoration results of the first section of the monitoring water level time series data, the second section of the monitoring water level time series data and the first section of the monitoring water level time series data to be restored are combined as the new first section of the monitoring water level time series data. If the length of the first section of the monitoring water level time series data is equal to the total monitoring time length of the monitoring well, then enter the end stage step, otherwise return to the data verification step; H. Ending stage: Through visualization technology, the real groundwater monitoring water level data of the monitoring well and the groundwater monitoring water level restoration data of the monitoring well obtained by this method are displayed within the same map range, and the groundwater monitoring water level restoration data file and data comparison map of the monitoring well are output.

2. A method for restoring groundwater monitoring water level data according to claim 1, characterized in that: In the above step B, the groundwater monitoring water level data of the monitoring well is clustered using the DBSCAN algorithm, taking each natural day as the time unit, specifically: a. Input all groundwater monitoring water level data of the monitoring well; b. Input the size of the ε neighborhood defined by the DBSCAN algorithm and the minimum number of data points MinPts contained in the ε neighborhood; c. Mark all groundwater monitoring water level data as not yet accessed by the DBSCAN algorithm; d. Traverse each groundwater monitoring water level data. If the groundwater monitoring data is marked as unaccessed, perform the following steps: (1) Mark the groundwater monitoring water level data as being in access status; (2) Find all water level data within the ε neighborhood of the groundwater monitoring water level data; (3) If the number of all groundwater monitoring water level data points within the ε neighborhood is greater than or equal to MinPts, then the groundwater monitoring water level data point is the core point; (4) If the number of all groundwater monitoring water level data points within the ε neighborhood is less than MinPts, but the point is within the ε neighborhood of other core points, then the groundwater monitoring water level data point is a boundary point; (5) If the groundwater monitoring water level data does not belong to the core point or the boundary point, the groundwater monitoring water level data point is a noise point; (6) Repeat the above steps (1) to (5) until all groundwater monitoring water level data are marked as accessed; e. Traverse each groundwater monitoring water level data. If the groundwater monitoring water level data is marked as a core point, perform the following steps: (1) If point p and point q are both within each other's ε neighborhood, then p and q are said to be density-reachable to each other; (2) Starting from any core point, recursively add all points that are mutually density-reachable to the same cluster; (3) If all the groundwater monitoring water level data belonging to the cluster have been classified, start by looking for another unclassified core point; (4) Repeat the above steps (2) and (3) until all core points are classified into different clusters; f. Add all boundary points to the cluster where the core point that makes it a boundary point is located; g. Output the different clusters of groundwater monitoring water level data and the identified noise data.

3. A method for restoring groundwater monitoring water level data according to claim 1, characterized in that: In the above step G, the specific steps of restoring the ARIMA model of the monitored water level time series data are as follows: a. Input the first monitoring water level time series data of the monitoring well; b. Perform ADF test on the monitored water level time series data or its different d-order difference results to determine which order difference is a stationary time series; c. Calculate the autocorrelation coefficient ACF and partial autocorrelation coefficient PACF for the stationary difference results, and analyze to obtain the optimal class p and order q; d. Use nonlinear least squares to estimate the ARIMA (p, d, q) model; e. The fitting results of the two test models were tested according to the Akaike information criterion and the Bayesian information criterion; f. Use the Ljung-Box test method to check whether the residual sequence has white noise characteristics. If the residual sequence does not belong to white noise, the ARIMA model needs to be readjusted; g. Predict the first section of the monitored water level time series data to be restored based on the adjusted model; h. Update model parameters to improve prediction accuracy after new data points appear; i. Output the model and the restoration results of the first section of the monitored water level time series data to be restored.

Images