Intelligent Control Device for Dynamic Balance of Groundwater Recharge

Abstract

The present invention relates to the technical field of groundwater recharge, and particularly to an intelligent control device for the dynamic balance of groundwater recharge. It includes a data acquisition module that collects characteristic data of groundwater in the target area through i observation wells in the target area, analyzes the characteristic data of groundwater in the target area, predicts the groundwater level in the target area at the future P moment, obtains the groundwater loss rate in the target area at the future P moment based on the groundwater level in the target area at the future P moment, and then calculates the groundwater storage capacity coefficient of the target area by combining the thickness, area, porosity of the aquifer in the target area with the permeability coefficient. Then, the obtained groundwater loss rate, groundwater storage capacity coefficient, and the real-time collected groundwater level in the target area are input into the trained machine learning model to predict the groundwater recharge volume. The present invention ensures the control accuracy of the groundwater recharge volume and realizes the dynamic balance of groundwater.

Description

Technical Field

[0001] The present invention relates to the technical field of groundwater recharge, and specifically to an intelligent control device for the dynamic balance of groundwater recharge. Background Art

[0002] Groundwater is one of the important fresh water resources globally and is widely used for agricultural irrigation, industrial water use, and domestic water use. However, due to overexploitation and unreasonable water use, the groundwater resources in many regions are facing the risk of depletion. To address the problem of declining groundwater levels, groundwater recharge technology has gradually been applied, aiming to artificially replenish water to the underground aquifer to restore or maintain the dynamic balance of groundwater.

[0003] A Chinese patent with the authorization announcement number CN110457831B discloses a method for evaluating the recharge suitability of a groundwater source area, including: evaluating the recharge potential of the target groundwater source area to obtain a potential index; evaluating the recharge risk of the target groundwater source area to obtain a risk index; evaluating the recharge benefit of the target groundwater source area to obtain a benefit index; and generating a recharge suitability evaluation result by weighting the potential index, risk index, and benefit index according to the constructed comprehensive index model, and obtaining the corresponding evaluation level. The advantages of this invention are that it realizes the quantitative evaluation of the recharge suitability of groundwater source areas, has advantages such as clear indicators and simple calculations, and can accurately screen the target areas of groundwater source areas suitable for recharge.

[0004] As in the above application, existing groundwater recharge control devices generally analyze the feasibility of groundwater recharge in some areas from the aspect of whether there is a water storage space in the underground aquifer, but the evaluation indicators are relatively single, lacking the prediction of future groundwater loss and the analysis of groundwater storage capacity, that is, formulating the groundwater recharge volume based on single groundwater loss, resulting in insufficient control accuracy and difficulty in effectively achieving the dynamic balance of groundwater. Summary of the Invention

[0005] To solve the above problems, the present invention provides an intelligent control device for the dynamic balance of groundwater recharge.

[0006] The present invention adopts the following technical solutions. The intelligent control device for the dynamic balance of groundwater recharge includes:

[0007] A data acquisition module that collects characteristic data of groundwater in the target area through i observation wells in the target area, where the characteristic data of the groundwater includes the thickness Hh of the aquifer, the groundwater level h, as well as the rainfall S in the target area at the future time P and the evaporation amount Z of the target area;

[0008] The loss prediction module analyzes the characteristic data of groundwater in the target area, predicts the groundwater level in the target area at the future time P, and obtains the groundwater loss rate in the target area at the future time P based on the groundwater level in the target area at the future time P;

[0009] The water storage capacity calculation module calculates the groundwater storage capacity coefficient of the target area by obtaining the thickness, area, porosity of the aquifer in the target area and combining with the permeability coefficient;

[0010] The data collection module collects historical groundwater recharge training data of the target area. The historical groundwater recharge training data is collected under the condition that the groundwater reaches dynamic equilibrium under recharge. The historical groundwater recharge training data includes the groundwater level in the target area, the groundwater loss rate, the groundwater storage capacity coefficient and the groundwater recharge volume;

[0011] The model training module trains a machine learning model to predict the groundwater recharge volume at the future time P based on the historical groundwater recharge training data, and collects the groundwater level, groundwater loss rate, and groundwater storage capacity coefficient in the target area in real time, and predicts the groundwater recharge volume based on the trained machine learning model.

[0012] As a further description of the above technical solution: The method for analyzing the characteristic data of groundwater in the target area and predicting the groundwater level in the target area at the future time P includes:

[0013] The characteristic data of historical groundwater is transformed into a training set by using a sliding window method to train a machine learning model for predicting the groundwater level in the target area at the future time P;

[0014] The training set generated by using the sliding window method is used as the input of the machine learning model. The machine learning model takes the predicted value of the groundwater level in the next P hours as the output, takes the real-time groundwater level as the prediction target, and takes minimizing the sum of the prediction accuracies of all training data as the training target; among them, the calculation formula of the prediction accuracy is: zk = (ak - wk)2, where k is the number of training data, zk is the prediction accuracy, ak is the predicted groundwater level corresponding to the kth group of training data, and wk is the actual groundwater level corresponding to the kth group of training data; the machine learning model is trained until the sum of the prediction accuracies reaches convergence and then the training stops; the machine learning model is a recurrent neural network model.

[0015] As a further description of the above technical solution: The groundwater loss rate in the target area at the future time P obtained based on the groundwater level in the target area at the future time P is:

[0016]

[0017] Wherein, Qp is the groundwater loss rate, h is the groundwater level, hp is the groundwater level at the future time P predicted, and p is the time interval.

[0018] As a further description of the above technical solution: The expression of the groundwater storage capacity coefficient is:

[0019]

[0020] Wherein, DXScs is the groundwater storage capacity coefficient, φ is the porosity, Hh is the aquifer thickness, A is the aquifer area; STxs is the permeability coefficient, and are weight coefficients, and are greater than zero.

[0021] As a further description of the above technical solution: The training method of the machine learning model for predicting the groundwater recharge amount at the future time P includes:

[0022] Converting the collected historical groundwater recharge training data into a corresponding set of feature vectors;

[0023] Using each set of feature vectors as the input of the machine learning model, the machine learning model takes the groundwater recharge amount corresponding to each set of target area groundwater level, groundwater loss rate and groundwater storage capacity coefficient as the output, takes the actual corresponding groundwater recharge amount of each set of target area groundwater level, groundwater loss rate and groundwater storage capacity coefficient as the prediction target, and takes minimizing the loss function value of the machine learning model as the training target; Stop training when the loss function value of the machine learning model is less than or equal to the preset target loss value.

[0024] As a further description of the above technical solution: It further includes:

[0025] A recharge optimization module, which uses a groundwater flow numerical model to simulate the groundwater level changes under different recharge rates and optimize the recharge amount and rate;

[0026] The groundwater flow numerical model is the MODFLOW model. MODFLOW simulates the Darcy flow law of groundwater, that is, the water flow is proportional to the hydraulic gradient and the medium permeability. The MODFLOW model predicts the groundwater level and flow direction by solving the groundwater flow equation.

[0027] As a further description of the above technical solution: The method of using the groundwater flow numerical model to simulate the groundwater level changes under different recharge rates and optimize the recharge amount and rate includes:

[0028] Define the model domain and mesh generation. Select and define the study area, including the groundwater recharge area and its surrounding influence range, and divide the model domain into regular grid cells;

[0029] Set the initial conditions and input the comprehensive parameters of the groundwater level aquifer in the target area. The comprehensive parameters include the thickness, area, porosity, and permeability coefficient of the aquifer;

[0030] Set the recharge conditions and input the real-time rainfall S in the target area and the real-time evaporation Z in the target area;

[0031] Run the baseline simulation and observe the natural change of the groundwater level over time without groundwater recharge;

[0032] Set up scenarios for running simulations, that is, for different recharge rates, increase the recharge rate sequentially, run the simulations respectively, and record the changes in the groundwater level in each case;

[0033] Evaluate the change in the water level after recharge to see if the expected water level recovery is achieved. Label the recharge scenarios that achieve the expected water level recovery and summarize them to establish a set of recharge scenarios;

[0034] Based on the predicted recharge volume and the preset groundwater recharge rate threshold, analyze and optimize the recharge scenarios in the set of recharge scenarios. Based on the recharge scenarios in the remaining set of recharge scenarios, select a suitable recharge scenario according to the environmental state in the target area.

[0035] As a further description of the above technical solution: Denote the set of recharge scenarios as YH:

[0036] YH ∈ {(a1, b1, c1), (a2, b2, c2),..., (an, bn, cn)}, where an is the groundwater recharge volume corresponding to the set label n, bn is the groundwater recharge time corresponding to the set label n, and cn is the groundwater recharge rate corresponding to the set label n.

[0037] As a further description of the above technical solution: The method for analyzing and optimizing the recharge scenarios in the set of recharge scenarios based on the predicted recharge volume and the preset groundwater recharge rate threshold includes:

[0038] When an in the recharge scenario is greater than the predicted recharge volume, remove this recharge scenario from the set of recharge scenarios;

[0039] When bn in the recharge scenario is greater than the P moment, remove this recharge scenario from the set of recharge scenarios;

[0040] When cn in the recharge scenario is greater than the preset groundwater recharge rate threshold, remove this recharge scenario from the set of recharge scenarios.

[0041] As a further description of the above technical solution: the rainfall S in the target area at the future time P and the evaporation Z in the target area are obtained through weather forecasts.

[0042] Beneficial effects:

[0043] The intelligent control device for dynamic balance of groundwater recharge provided by the present invention obtains the characteristic data of groundwater in the target area, converts the historical characteristic data of groundwater into a training set by using a sliding window method to train a machine learning model for predicting the groundwater level in the target area at the future time P, predicts the groundwater level in the target area at the future time P, obtains the groundwater loss rate in the target area at the future time P according to the groundwater level in the target area at the future time P, then calculates the groundwater storage capacity coefficient of the target area by obtaining the thickness, area, porosity and permeability coefficient of the aquifer in the target area, and then inputs the obtained groundwater loss rate, groundwater storage capacity coefficient, and the real-time collected groundwater level in the target area into the trained machine learning model to predict the groundwater recharge volume, thereby accurately predicting the groundwater recharge volume, ensuring the control accuracy of the groundwater recharge volume, and realizing the dynamic balance of groundwater;

[0044] Furthermore, the present invention also adds a recharge optimization module, which uses a groundwater flow numerical model (MODFLOW model) to simulate the groundwater level changes under different recharge rates. Through model simulation, different recharge strategies can be tested in a virtual environment, avoiding expensive and time-consuming experiments in reality. By simulating different recharge schemes, the optimal groundwater recharge volume and rate can be determined to ensure that while meeting the demand, over-recharge or under-recharge is avoided. The optimized recharge strategy helps to achieve the sustainable management of water resources and ensure the long-term availability of groundwater resources. Description of the Drawings

[0045] The present invention will be further explained below in conjunction with the drawings and embodiments:

[0046] Figure 1 It is a module connection diagram of the intelligent control device for dynamic balance of groundwater recharge in Embodiment 1 of the present invention;

[0047] Figure 2 It is a module connection diagram of the intelligent control device for dynamic balance of groundwater recharge provided in Embodiment 2 of the present invention. Detailed Embodiments

[0048] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific illustrations. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0049] Example 1

[0050] Please refer to Figure 1 , the embodiment of the present invention provides a technical solution: a smart control device for dynamic balance of groundwater recharge, including:

[0051] A data acquisition module collects characteristic data of groundwater in the target area through i observation wells in the target area, where the characteristic data of the groundwater includes the thickness Hh of the aquifer, the groundwater level h, as well as the rainfall S and evaporation Z of the target area at the future P moment;

[0052] The rainfall S and evaporation Z of the target area at the future P moment are obtained through weather forecasts.

[0053] It should be noted that the observation wells are the observation wells drilled in the target area, and the depth of the wells should reach the groundwater layer, and the well pipes should be installed to ensure the representativeness of the water level data.

[0054] A loss prediction module analyzes the characteristic data of groundwater in the target area, predicts the groundwater level in the target area at the future P moment, and obtains the groundwater loss rate in the target area at the future P moment according to the groundwater level in the target area at the future P moment;

[0055] The method for analyzing the characteristic data of groundwater in the target area and predicting the groundwater level in the target area at the future P moment includes:

[0056] Using a sliding window method to convert the historical characteristic data of groundwater into a training set to train a machine learning model for predicting the groundwater level in the target area at the future P moment;

[0057] Using the training set generated by the sliding window method as the input of the machine learning model, the machine learning model outputs the predicted value of the groundwater level in the next P hours, takes the real-time groundwater level as the prediction target, and takes minimizing the sum of the prediction accuracies of all training data as the training target; where the calculation formula of the prediction accuracy is: zk = (ak - wk)2, where k is the number of all training data, zk is the prediction accuracy, ak is the predicted groundwater level corresponding to the kth group of training data, and wk is the actual groundwater level corresponding to the kth group of training data; training the machine learning model until the sum of the prediction accuracies reaches convergence and then stopping the training; the machine learning model is a recurrent neural network model.

[0058] The groundwater loss rate in the target area at the future P moment obtained according to the groundwater level in the target area at the future P moment is:

[0059]

[0060] Wherein, Qp is the groundwater loss rate, h is the groundwater level, hp is the groundwater level at the future time P predicted, and p is the time interval.

[0061] A water storage capacity calculation module calculates the groundwater storage capacity coefficient of the target area by obtaining the thickness, area, porosity of the aquifer in the target area and combining with the permeability coefficient;

[0062] The expression of the groundwater storage capacity coefficient is:

[0063]

[0064] Wherein, DXScs is the groundwater storage capacity coefficient, φ is the porosity, Hh is the aquifer thickness, A is the aquifer area; STxs is the permeability coefficient, and are weight coefficients, and are greater than zero.

[0065] It should be noted that the magnitude of the weight coefficient is a specific value obtained by quantifying each data for subsequent comparison. Regarding the magnitude of the weight coefficient, it depends on the number of comprehensive parameters and the weight coefficients initially set by those skilled in the art for each group of comprehensive parameters.

[0066] It should be noted that the porosity represents the proportion of the pore volume in the aquifer to the total rock volume. The porosity is an important index to measure the water storage capacity of rocks or soils. The higher the porosity, the stronger the water storage capacity of the aquifer, because a high porosity means there is more space in the aquifer that can be filled with water, and vice versa. Secondly, the porosity can be determined through laboratory tests (such as core analysis), logging techniques (such as NMR logging), and geological simulations;

[0067] The area and thickness of the aquifer determine the volume space of the aquifer. The larger the thickness and area, the stronger the water storage capacity, and vice versa; among them, the area and thickness of the aquifer can be obtained through electromagnetic detection technology, that is, by using electromagnetic wave detection technology to identify the electromagnetic property differences between the aquifer and other strata, so as to infer the horizontal distribution area and thickness of the aquifer.

[0068] The permeability coefficient represents the resistance of the formation or rock to the flow of water, and is usually used to describe the water flow ability in the formation. In a formation with a high permeability coefficient, water flows more freely. Although this does not directly increase the water storage volume, it affects the recharge and discharge rates of water, and thus affects the dynamic water storage capacity of the aquifer. The main method for obtaining the permeability coefficient is through on-site experiments, that is, through pumping tests.

[0069] The specific method of the pumping test is:

[0070] Stabilize the water level in the observation well to a certain initial level;

[0071] Start pumping water at a constant flow rate and record the changes in the water levels of the pumping well and adjacent observation wells over time;

[0072] Continue pumping until the water level reaches a relatively stable state;

[0073] Use the Theis formula, Cooper-Jacob simplified method, or Thiem formula to analyze the relationship between the drawdown and time or distance, and thus calculate the hydraulic conductivity;

[0074] The stronger the water storage capacity of the aquifer, the higher the allowable recharge volume and rate can be. However, at the same time, attention should be paid to avoiding exceeding the saturation point of the aquifer to prevent excessive groundwater seepage or land subsidence.

[0075] A data collection module that collects historical groundwater recharge training data for the target area. The historical groundwater recharge training data is collected when the groundwater reaches dynamic equilibrium under recharge. The historical groundwater recharge training data includes the groundwater level in the target area, the groundwater loss rate, the groundwater storage capacity coefficient, and the groundwater recharge volume;

[0076] A model training module that trains a machine learning model to predict the groundwater recharge volume at a future time P based on the historical groundwater recharge training data. It real-time collects the groundwater level, groundwater loss rate, and groundwater storage capacity coefficient in the target area, and predicts the groundwater recharge volume based on the trained machine learning model.

[0077] The training method of the machine learning model for predicting the groundwater recharge volume includes:

[0078] Convert the collected historical groundwater recharge training data into a corresponding set of feature vectors;

[0079] Use each set of feature vectors as the input of the machine learning model. The machine learning model takes the groundwater recharge volume corresponding to each set of groundwater level, groundwater loss rate, and groundwater storage capacity coefficient in the target area as the output, takes the actual corresponding groundwater recharge volume of each set of groundwater level, groundwater loss rate, and groundwater storage capacity coefficient in the target area as the prediction target, and takes minimizing the loss function value of the machine learning model as the training target; stop training when the loss function value of the machine learning model is less than or equal to the preset target loss value.

[0080] The machine learning model can be one of models such as support vector machine regression, random forest regression, or neural network regression.

[0081] The loss function value of the machine learning model is the mean squared error.

[0082] By minimizing the loss function as the objective, the model is trained to better fit the data, thereby improving the performance and accuracy of the model.

[0083] In the loss function, MSE is the loss function value of the machine learning model, x is the feature vector group number; m is the number of feature vector groups; yx is the groundwater recharge volume corresponding to the x-th group of feature vectors, is the groundwater recharge volume corresponding to the x-th group of feature vectors in real time.

[0084] Other model parameters of the machine learning model, the target loss value, the optimization algorithm, the proportion of the training set, test set, and validation set, and the optimization of the loss function are all obtained through actual engineering implementation and continuous experimental tuning.

[0085] In this implementation, by obtaining the characteristic data of the groundwater in the target area, the historical groundwater characteristic data is converted into a training set by using a sliding window method to train a machine learning model for predicting the groundwater level in the target area at the future P moment. The groundwater level in the target area at the future P moment is predicted, and the groundwater loss rate in the target area at the future P moment is obtained according to the groundwater level in the target area at the future P moment. Then, by obtaining the thickness, area, porosity, and permeability coefficient of the aquifer in the target area, the groundwater storage capacity coefficient of the target area is calculated. Then, the obtained groundwater loss rate, groundwater storage capacity coefficient, and the real-time collected groundwater level in the target area are input into the trained machine learning model to predict the groundwater recharge volume, so as to accurately predict the groundwater recharge volume, ensure the control accuracy of the groundwater recharge volume, and achieve the dynamic balance of groundwater.

[0086] Embodiment 2

[0087] Please refer to Figure 2 , this implementation adds a recharge optimization module on the basis of the above embodiment:

[0088] The recharge optimization module uses a groundwater flow numerical model to simulate the groundwater level changes under different recharge rates and optimize the recharge volume and rate.

[0089] The groundwater flow numerical model is the MODFLOW model. MODFLOW simulates the Darcy flow law of groundwater, that is, the water flow rate is proportional to the hydraulic gradient and the medium permeability. The MODFLOW model predicts the groundwater level and flow direction by solving the groundwater flow equation.

[0090] The method of using a groundwater flow numerical model to simulate the groundwater level changes under different recharge rates and optimize the recharge volume and rate includes:

[0091] Define the model domain and grid division, select and define the study area, including the groundwater recharge area and its surrounding influence range, and divide the model domain into regular grid cells;

[0092] Set the initial conditions, and input the comprehensive parameters of the groundwater level aquifer in the target area. The comprehensive parameters include the thickness, area, porosity, and permeability coefficient of the aquifer;

[0093] Set the recharge conditions, and input the real-time rainfall S in the target area and the real-time evaporation Z in the target area;

[0094] Run the baseline simulation, and observe the natural change of the groundwater level over time without groundwater recharge;

[0095] Set multiple scenarios for running simulations, that is, for different recharge rates, sequentially increase the recharge rate, run the simulations respectively, and record the change of the groundwater level in each case;

[0096] Evaluate the change of the water level after recharge, whether the expected water level recovery is achieved, label the recharge scenarios that achieve the expected water level recovery, and summarize and establish a set of recharge scenarios;

[0097] Based on the predicted recharge volume and the preset groundwater recharge rate threshold, analyze and optimize the recharge scenarios in the set of recharge scenarios. Based on the recharge scenarios in the remaining set of recharge scenarios, select a suitable recharge scenario according to the environmental state in the target area.

[0098] Denote the set of the recharge scenarios as YH:

[0099] YH ∈ {(a1, b1, c1), (a2, b2, c2),..., (an, bn, cn)}, where an is the groundwater recharge volume corresponding to the set label n, bn is the groundwater recharge time corresponding to the set label n, and cn is the groundwater recharge rate corresponding to the set label n.

[0100] The method for analyzing and optimizing the recharge scenarios in the set of recharge scenarios based on the predicted recharge volume and the preset groundwater recharge rate threshold includes:

[0101] When an in the recharge scenario is greater than the predicted recharge volume, remove this recharge scenario from the set of recharge scenarios;

[0102] When bn in the recharge scenario is greater than the moment P, remove this recharge scenario from the set of recharge scenarios;

[0103] When cn in the recharge scenario is greater than the preset groundwater recharge rate threshold, remove this recharge scenario from the set of recharge scenarios.

[0104] In this embodiment, the groundwater flow numerical model (MODFLOW model) is used to simulate the changes in groundwater levels under different recharge rates. Through model simulation, different recharge strategies can be tested in a virtual environment, avoiding expensive and time-consuming experiments in reality. By simulating different recharge scenarios, the optimal groundwater recharge volume and rate can be determined to ensure that while meeting the demand, over-recharge or under-recharge is avoided. The optimized recharge strategy helps to achieve the sustainable management of water resources and ensure the long-term availability of groundwater resources.

[0105] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. The descriptions in the above embodiments and the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. The groundwater recharge dynamic balance intelligent control device is characterized by: include: The data acquisition module collects characteristic data of groundwater in the target area through i observation wells in the target area, wherein the characteristic data of groundwater include the thickness Hh of the aquifer, the groundwater level h, and the rainfall S and evaporation Z of the target area at the future time P; The loss prediction module analyzes the characteristic data of groundwater in the target area, predicts the groundwater level in the target area at time P in the future, and obtains the groundwater loss rate in the target area at time P in the future according to the groundwater level in the target area at time P in the future; The water storage capacity calculation module calculates the groundwater storage capacity coefficient of the target area by obtaining the thickness, area, porosity and permeability coefficient of the aquifer in the target area; The method to obtain the permeability coefficient is: Stabilize the water level in the observation well to a certain initial level; Start pumping water at a constant rate and record the changes in water levels over time in the pumping well and adjacent observation wells; Continue pumping until the water level reaches a relatively stable state; The relationship between water level drawdown and time or distance is analyzed using Theis formula, Cooper-Jacob simplified method or Thiem formula to calculate the permeability coefficient; The expression of the groundwater storage capacity coefficient is: Where DXScs is the groundwater storage capacity coefficient, φ is the porosity, Hh is the aquifer thickness, A is the aquifer area; STxs is the permeability coefficient, and is the weight coefficient, and greater than zero; A data collection module collects historical groundwater recharge training data of the target area. The historical groundwater recharge training data is collected when the groundwater reaches a dynamic balance under recharge. The historical groundwater recharge training data includes the groundwater level, groundwater loss rate, groundwater storage capacity coefficient and groundwater recharge volume of the target area. The model training module trains a machine learning model to predict the groundwater recharge amount at time P in the future based on historical groundwater recharge training data. It collects groundwater levels, groundwater loss rates, and groundwater storage capacity coefficients in the target area in real time, and predicts the groundwater recharge amount based on the trained machine learning model.

2. The groundwater recharge dynamic balance intelligent control device according to claim 1 is characterized in that: The method of analyzing the characteristic data of groundwater in the target area and predicting the groundwater level in the target area at time P in the future includes: The historical groundwater characteristic data are converted into a training set using a sliding window method to train a machine learning model for predicting the groundwater level in the target area at time P in the future; A training set generated by a sliding window method is used as the input of a machine learning model. The machine learning model takes the predicted value of the groundwater level in the next P hours as output, the real-time groundwater level as the prediction target, and minimizing the sum of the prediction accuracies of all training data as the training target; wherein the calculation formula for prediction accuracy is: zk=(ak-wk)2, wherein k is the number of the training data, zk is the prediction accuracy, ak is the predicted groundwater level corresponding to the kth group of training data, and wk is the actual groundwater level corresponding to the kth group of training data; the machine learning model is trained until the sum of the prediction accuracies converges and the training is stopped; the machine learning model is a recurrent neural network model.

3. The groundwater recharge dynamic balance intelligent control device according to claim 1 is characterized in that: According to the groundwater level in the target area at time P in the future, the groundwater loss rate in the target area at time P in the future is: Where Qp is the groundwater loss rate, h is the groundwater level, hp is the predicted groundwater level at future time P, and p is the time interval.

4. The groundwater recharge dynamic balance intelligent control device according to claim 1 is characterized in that: The training method of the machine learning model for predicting the future groundwater recharge volume at time P includes: Convert the collected historical groundwater recharge training data into a corresponding set of feature vectors; Each group of feature vectors is used as the input of the machine learning model, and the machine learning model takes the groundwater recharge amount corresponding to the groundwater level, groundwater loss rate and groundwater storage capacity coefficient of each group of target areas as output, and takes the groundwater recharge amount actually corresponding to the groundwater level, groundwater loss rate and groundwater storage capacity coefficient of each group of target areas as the prediction target, and takes minimizing the loss function value of the machine learning model as the training target; training is stopped when the loss function value of the machine learning model is less than or equal to the preset target loss value.

5. The groundwater recharge dynamic balance intelligent control device according to claim 1 is characterized in that: Also includes: Recharge Optimization Module, which uses a groundwater flow numerical model to simulate groundwater level changes under different recharge rates and optimize the recharge amount and rate; The groundwater flow numerical model is a MODFLOW model. MODFLOW simulates the Darcy flow law of groundwater, that is, the water flow rate is proportional to the hydraulic gradient and the permeability of the medium. The MODFLOW model predicts the groundwater level and flow direction by solving the groundwater flow equation.

6. The groundwater recharge dynamic balance intelligent control device according to claim 5 is characterized in that: The method of using the groundwater flow numerical model to simulate groundwater level changes under different recharge rates and optimize the recharge amount and rate includes: Define the model domain and grid division, select and define the study area, including the groundwater recharge area and its surrounding influence area, and divide the model domain into regular grid cells; Initial conditions are set, and comprehensive parameters of the groundwater level aquifer in the target area are input, wherein the comprehensive parameters include thickness, area, porosity, and permeability coefficient of the aquifer; Set the compensation conditions and input the real-time rainfall S and real-time evaporation Z of the target area; Run a baseline simulation to observe the natural changes in groundwater levels over time without groundwater recharge; Set up the scheme to run the simulation, that is, for different recharge rates, increase the recharge rate in turn, run the simulation separately, and record the groundwater level changes in each case; Evaluate the change in water level after replenishment to see whether the expected water level recovery has been achieved. Set labels for replenishment plans that have achieved the expected water level recovery and summarize them to establish a replenishment plan set. Based on the predicted recharge volume and the preset groundwater recharge rate threshold, the recharge schemes in the recharge scheme set are analyzed and optimized, and based on the recharge schemes in the remaining recharge scheme set and according to the environmental conditions in the target area, a suitable recharge scheme is selected.

7. The groundwater recharge dynamic balance intelligent control device according to claim 6 is characterized in that: The set of backfill plans is denoted as YH: YH∈{(a1,b1,c1),(a2,b2,c2),...,(an,bn,cn)}, an is the groundwater recharge amount corresponding to the set label n, bn is the groundwater recharge time corresponding to the set label n, and cn is the groundwater recharge rate corresponding to the set label n.

8. The groundwater recharge dynamic balance intelligent control device according to claim 7 is characterized in that: The method for analyzing and optimizing the recharge schemes in the recharge scheme set based on the predicted recharge amount and the preset groundwater recharge rate threshold includes: When an in the replenishment plan is greater than the predicted replenishment amount, the replenishment plan is removed from the replenishment plan set; When bn in the backfill scheme is greater than P, the backfill scheme is removed from the backfill scheme set; When cn in a recharge scheme is greater than the preset groundwater recharge rate threshold, the recharge scheme is removed from the recharge scheme set.

9. The groundwater recharge dynamic balance intelligent control device according to claim 1, characterized in that: The rainfall S in the target area and the evaporation Z in the target area at the future time P are obtained through weather forecast.

Images