Continuous monitoring device for water level of underground water

Abstract

The invention relates to the technical field of environmental monitoring, and discloses an underground water level continuous monitoring device, which comprises a monitoring probe module arranged at the bottom area of a shell and used for monitoring the water level of underground water and related environmental parameters; the data processing module is arranged in the middle area in the shell and used for collecting, processing, analyzing and storing the monitoring data; the wireless transmission module is connected with the data processing module and used for uploading the processed monitoring data to a remote monitoring system; the energy management module is arranged on the upper layer of the data processing module in the shell; the shell and installation module comprises a shell, a bottom plate and multiple layers of partition plates in the shell. Through time sequence analysis based on a long short-term memory network (LSTM), the change trend (such as water level fluctuation in the future 7 days) of the groundwater level in the future can be predicted, and through dynamic comparison of monitoring data and historical data modes, an alarm is triggered and early warning information is provided when abnormal water level changes (such as sudden drop and sudden rise) are found.

Description

Technical Field

[0001] The invention relates to the technical field of environmental monitoring, and in particular to a device for continuous monitoring of groundwater level. Background Art

[0002] Groundwater is one of the important water resources, and the dynamic changes of its water level are closely related to environmental protection, agricultural irrigation, urban water supply and industrial production. Therefore, continuous monitoring of groundwater levels is of great significance for resource management and geological disaster warning. Traditional groundwater monitoring devices mainly rely on simple physical sensors to obtain water level data through manual reading or low-frequency sampling. This method has the following disadvantages:

[0003] Existing technologies mainly rely on manual post-analysis of historical data to obtain regular information on water level changes, and lack real-time detection capabilities. Once water level anomalies occur (such as drastic changes caused by geological disasters), traditional systems cannot detect and trigger alarms in time, which may lead to significant losses or risks.

[0004] Some existing systems use simple mathematical models to analyze groundwater level changes, but these models cannot be optimized online with new data and are difficult to adapt to changing environmental conditions (such as seasonal rainfall or changes in human water use behavior). This will lead to reduced prediction accuracy and fail to meet dynamic monitoring needs in complex environments. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a groundwater level continuous monitoring device, which solves the problem that the prior art mainly relies on manual post-analysis of historical data to obtain regular information on water level changes and lacks real-time detection function.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a groundwater level continuous monitoring device, comprising:

[0007] A monitoring probe module, disposed in the bottom area of ​​the housing, for monitoring the groundwater level and related environmental parameters;

[0008] A data processing module, disposed in the middle area of ​​the housing, for collecting, processing, analyzing and storing monitoring data;

[0009] A wireless transmission module, connected to the data processing module, for uploading the processed monitoring data to a remote monitoring system;

[0010] An energy management module, arranged on a layer above the data processing module in the housing, for providing power support for the groundwater level continuous monitoring device;

[0011] The housing and installation module include a housing, a base plate and internal multi-layer partitions, which are used to protect the device structure and adapt to different installation environments.

[0012] Preferably, the monitoring probe module includes:

[0013] The pressure sensor, temperature sensor and water quality sensor are respectively fixed at the plug-in interface of the bottom partition to monitor the water level, temperature and water quality parameters of the groundwater;

[0014] The plug plate is connected to the partition plate, the outer walls of the pressure sensor, the temperature sensor and the water quality sensor are arranged inside the partition plate, and the partition plate is fixedly connected to the shell through a sealing ring to enhance the sealing performance.

[0015] Preferably, the data processing module includes:

[0016] A data acquisition unit, connected to the pressure sensor, the temperature sensor and the water quality sensor, for receiving water level, water quality and temperature data;

[0017] The edge computing unit uses the Kalman filter algorithm to perform real-time denoising on the collected data and fits the water level change trend through the least squares method;

[0018] The artificial intelligence analysis unit builds a groundwater level prediction model based on the long short-term memory network (LSTM), predicts water level changes in the next seven days, and detects abnormal fluctuations;

[0019] A data storage unit, fixed on the upper surface of the partition, with a built-in high-capacity storage device, supporting data compression storage and breakpoint resume functions;

[0020] The data encryption and security unit uses the AES-256 encryption algorithm to protect data storage and transmission.

[0021] Preferably, the wireless transmission module includes:

[0022] Multi-mode communication unit, supporting NB-IoT, LoRa, 4G and 5G communication modes, and can dynamically switch communication protocols according to the communication environment;

[0023] The signal amplification and stabilization unit is connected to the data processing module to enhance the strength and stability of signal transmission;

[0024] Automatic reconnection unit, used to automatically restore the connection after communication interruption and upload cached data;

[0025] Low power consumption optimization unit reduces energy consumption by optimizing communication protocols.

[0026] Preferably, the energy management module includes:

[0027] A lithium battery, fixed on the third partition of the shell, is used to provide power support;

[0028] an ambient energy harvesting unit, including a high-efficiency monocrystalline silicon solar panel, disposed in a top region of the housing;

[0029] The power monitoring unit monitors the battery power in real time and triggers an alarm when the power level falls below the threshold;

[0030] The low power control unit switches to low power mode to extend the operating time when the device is idle.

[0031] Preferably, the housing and the mounting module include:

[0032] The outer shell is made of corrosion-resistant material, and the inner bottom is fixedly connected with a bottom plate, and the bottom plate is provided with filter holes for filtering impurities in groundwater;

[0033] The partition is connected to the shell through a sealing ring to form a multi-layer modular structure for separating different functional modules;

[0034] The gear and the rack, the gear rotates on the upper surface of the second layer of the partition and the rack slides on the inner wall of the shell, and the gear is driven to rotate by pulling the rack so that the two racks are retracted at the same time;

[0035] The observation window and indicator light are respectively fixed inside the top partition to display the equipment operation status and fault warning.

[0036] Preferably, the application process of the artificial intelligence analysis unit based on the long short-term memory network LSTM model includes a data preparation stage, specifically:

[0037] The monitoring data are preprocessed. The monitoring data include water level data, temperature data, and environmental factors. The preprocessing includes data cleaning, interpolation to fill missing values, and normalization. All data are normalized to the range of [0,1] or [-1,1]. A time window is constructed, and the data of the past k time points are used as the model input to predict the water level changes at the next m time points. The input format is:

[0038] X input ={(H t-k , H t-k+1 , ..., H t-1 )},Y output ={H t , H t-1 , ..., H t-m}

[0039] Among them, X input The input data is the water level at the past k time points;

[0040] Ht-k ,H t-k+1 ,…,H t-1 The water level data in the past;

[0041] Y output To output data, the water level at m future time points;

[0042] H t ,H t+1 ,…,H t+m is the target value of the future water level height.

[0043] Preferably, the long short-term memory network LSTM structure in the artificial intelligence analysis unit includes:

[0044] The forget gate is used to decide whether to keep or forget historical information. Its calculation formula is:

[0045] f t =σ(W f ·[h t-1 , x t ]+b f )

[0046] Among them, f t is the output value of the forget gate, ranging from [0,1], which determines the forgetting ratio of historical information;

[0047] σ is the activation function, using the sigmoid function, and the output range is [0,1];

[0048] W f is the weight matrix of the forget gate;

[0049] h t-1 is the hidden state at the previous moment;

[0050] xt is the input at the current moment;

[0051] b f is the bias vector of the forget gate;

[0052] The input gate is used to update the memory cell state. Its calculation formula is:

[0053] i t =σ(W i ·[h t-1 ,x t ]+b i ),

[0054]

[0055] Among them, it is the output value of the input gate, ranging from [0,1], indicating the memory ratio of the current new information;

[0056] is the value of the candidate memory unit;

[0057] W i ,W C is the weight matrix of the input gate and the candidate memory unit;

[0058] b i , b C is the bias vector of the input gate and the candidate memory unit;

[0059] C t is the state of the memory unit at the previous moment;

[0060] C t-1 is the memory unit state at the previous moment;

[0061] ⊙ is the element-wise multiplication operation;

[0062] The output gate is used to generate the current hidden state for use at the next moment or in the next layer. The calculation formula is:

[0063] o t =σ(W o ·[h t-1 , x t ]+b o ), h t =o t ⊙tanh(C t )

[0064] Among them, t is the output value of the output gate, ranging from [0,1];

[0065] W o is the weight matrix of the output gate;

[0066] b o is the bias vector of the output gate;

[0067] h t is the hidden state at the current moment;

[0068] tanh(C t ) is the nonlinear activation value of the memory unit.

[0069] Preferably, the long short-term memory network LSTM of the artificial intelligence analysis unit performs model training and optimization through the following steps:

[0070] Use mean square error MSE as the loss function to calculate the error between the predicted value and the true value. The loss function formula is:

[0071]

[0072] in, MSE is the mean square error, which represents the average sum of square errors between the model prediction value and the true value;

[0073] N is the total number of samples;

[0074] H i,pred is the predicted water level height of the i-th sample;

[0075] H i,true is the true water level height of the i-th sample;

[0076] The Adam optimization algorithm is used to iteratively optimize the model parameters through the following formula:

[0077]

[0078] Among them, θ t is the model parameter at the current moment;

[0079] θ t-1 is the model parameter of the previous moment;

[0080] η is the learning rate, which indicates the step size of parameter update;

[0081] is the gradient of the loss function with respect to the model parameters.

[0082] Preferably, the long short-term memory network LSTM of the artificial intelligence analysis unit performs water level change trend prediction and model evaluation based on the model prediction results, specifically including:

[0083] Using the input data of the time window, the water level changes at the next m time points are predicted, and the output is:

[0084]

[0085] in, is the predicted water level at the i-th time point in the future;

[0086] f is the function expression of the LSTM model;

[0087] θ is the trained model parameter;

[0088] H t-k ,H t-k+1 ,…,H t-1 Historical input data for the time window;

[0089] The model performance is evaluated using the following evaluation indicators:

[0090] Mean Square Error (MSE):

[0091]

[0092] Mean absolute error (MAE):

[0093]

[0094] Coefficient of determination:

[0095]

[0096] in, is the mean of the true water level height;

[0097] R 2 is the coefficient of determination, and the closer the value is to 1, the better the model prediction performance is.

[0098] The present invention provides a groundwater level continuous monitoring device, which has the following beneficial effects:

[0099] 1. The present invention can predict the future trend of groundwater level changes (such as water level fluctuations in the next 7 days) through time series analysis based on long short-term memory network (LSTM). By dynamically comparing monitoring data and historical data patterns, an alarm is triggered when abnormal water level changes (such as sudden drops or sudden rises) are found, and early warning information is provided. The model can be continuously optimized with the input of new data to enhance its ability to predict long-term changes in groundwater levels.

[0100] 2. The present invention can measure the groundwater level in real time with an accuracy of millimeters, ensuring the accuracy of the monitoring data. It can also monitor the groundwater temperature and water quality (such as pH value, conductivity, dissolved oxygen, etc.), provide comprehensive groundwater environmental information, and provide support for water resource management and environmental protection. By collecting data in real time, it ensures that the monitoring system can respond quickly to water level changes, and is suitable for monitoring needs in dynamic environments.

[0101] 3. The present invention allows operators to easily place the device into the well by adjusting the sliding distance and direction of the rack, without the need for complex lifting tools, thus simplifying the installation process of the device.

[0102] 4. The present invention effectively intercepts impurities such as mud, particles, etc. when water enters the device through the filter holes, protects internal sensors (such as pressure sensors, temperature sensors and water quality sensors) and electronic modules from being contaminated or damaged, extends the service life of the equipment and improves monitoring accuracy.

[0103] 5. The present invention uses modular design to miniaturize the monitoring device, making it easy to install in various narrow monitoring wells, while reducing production costs and improving cost performance, which is conducive to large-scale promotion and application, especially in areas where a large number of monitoring points are required. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] Figure 1 It is a three-dimensional schematic diagram of a groundwater level continuous monitoring device of the present invention;

[0105] Figure 2 A schematic diagram of the internal structure of a groundwater level continuous monitoring device of the present invention;

[0106] Figure 3 It is a schematic diagram of the partial structure of a plug board of a groundwater level continuous monitoring device of the present invention;

[0107] Figure 4 It is a schematic diagram of the partial structure of the gear of a groundwater level continuous monitoring device of the present invention;

[0108] Figure 5 A system architecture diagram of a groundwater level continuous monitoring device of the present invention;

[0109] Figure 6 This is a data processing module architecture diagram of a groundwater level continuous monitoring device of the present invention;

[0110] Figure 7 This is a structural diagram of a wireless transmission module of a groundwater level continuous monitoring device of the present invention.

[0111] Among them, 1. Shell; 2. Bottom plate; 3. Sealing ring; 4. Partition; 5. Pressure sensor; 6. Temperature sensor; 7. Water quality sensor; 8. Plug-in board; 9. Data processing module; 10. Wireless transmission module; 11. Lithium battery; 12. Filter hole; 13. Energy management module; 14. Gear; 15. Rack; 16. Solar panel; 17. Observation window; 18. Indicator light. DETAILED DESCRIPTION

[0112] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0113] Please see attached Figure 1 To Attachment Figure 7 The embodiment of the present invention provides a groundwater level continuous monitoring device, comprising:

[0114] A monitoring probe module is arranged in the bottom area of ​​the housing 1 and is used to monitor the groundwater level and related environmental parameters;

[0115] The data processing module 9 is arranged in the middle area of ​​the housing 1 and is used to collect, process, analyze and store the monitoring data;

[0116] The wireless transmission module 10 is connected to the data processing module 9 and is used to upload the processed monitoring data to the remote monitoring system;

[0117] The energy management module 13 is arranged on the upper layer of the data processing module 9 in the housing 1 and is used to provide power support for the groundwater level continuous monitoring device;

[0118] The housing and installation module includes a housing 1, a base plate 2, and an internal multi-layer partition 4, which are used to protect the device structure and adapt to different installation environments.

[0119] Specifically, the device realizes the intelligent prediction and early warning function of groundwater level changes through the built-in artificial intelligence analysis unit. Using the long short-term memory network (LSTM) model, the device can conduct deep learning and analysis of historical groundwater level data, so as to accurately capture the trend of water level changes and predict water level fluctuations in the future (for example, the next 7 days). This prediction capability based on time series analysis not only improves the adaptability of the device to dynamic water level changes, but also provides a scientific basis for decision-making for water resources management departments. At the same time, the device is equipped with an anomaly detection function, which can quickly identify sudden abnormal changes (such as rapid increase or decrease in water level) by comparing the current water level data with historical patterns in real time. When an abnormality is detected, the device will automatically trigger a local or remote alarm to provide users with timely warnings, which helps to avoid geological disasters, environmental risks or water resource utilization problems that may be caused by drastic changes in groundwater levels. In addition, the system supports online optimization, which can continuously improve the prediction accuracy of the model based on the newly added data, ensure the long-term prediction ability and applicability of the device, and bring intelligent upgrades to the field of groundwater monitoring.

[0120] The monitoring probe module includes:

[0121] The pressure sensor 5, the temperature sensor 6 and the water quality sensor 7 are respectively fixed at the interface of the plug plate 8 of the bottom partition plate 4, and are used to monitor the water level, temperature and water quality parameters of the groundwater;

[0122] The plug plate 8 is connected to the partition 4. The outer walls of the pressure sensor 5, the temperature sensor 6 and the water quality sensor 7 are arranged inside the partition 4. The partition 4 is fixedly connected to the housing 1 through the sealing ring 3 to enhance the sealing performance.

[0123] Specifically, the pressure sensor 5

[0124] Arrangement: The pressure sensor 5 is fixedly connected to the interface of the plug plate 8 of the bottom partition 4, and its outer wall is embedded in the partition 4.

[0125] Functional description: The pressure sensor 5 calculates the water level by detecting the pressure generated by groundwater in real time. It is suitable for a water depth range of 0-100 meters and the measurement accuracy can reach ±1 mm.

[0126] Technical features: The pressure sensor adopts MEMS micro-electromechanical system technology, and the shell material is corrosion-resistant titanium alloy. It can work for a long time in high-salinity, acidic or alkaline groundwater environments.

[0127] Temperature sensor 6

[0128] Arrangement: The temperature sensor 6 and the pressure sensor 5 are installed side by side on the plug board 8, and their detection probes extend downward into the water.

[0129] Functional description: Temperature sensor 6 is used to measure the real-time temperature data of groundwater, with a temperature range of -20℃ to 85℃ and an accuracy of ±0.1℃.

[0130] Technical features: Built-in temperature compensation algorithm can feed back the measured temperature data to the pressure sensor 5 to correct the water level measurement error caused by temperature changes.

[0131] Water quality sensor 7

[0132] Arrangement: The water quality sensor 7 is installed on the other side of the plug board 8, and is arranged at a distance from the pressure sensor 5 and the temperature sensor 6, so as to avoid interference between the sensors.

[0133] Function description: The water quality sensor 7 collects the pH value, conductivity and dissolved oxygen content of groundwater in real time, providing basic data for groundwater quality analysis.

[0134] Technical features: The water quality sensor has a modular design and supports the expansion of other monitoring parameters such as turbidity, total organic carbon, etc. The shell is made of polymer anti-corrosion material to extend the service life of the sensor.

[0135] Plug-in 8

[0136] Design features: The plug-in board 8 is tightly connected to the bottom partition board 4, and the sensor module is fixed by a quick-plug interface, which is convenient for disassembly, replacement and maintenance, and is powered by the energy management module 13.

[0137] The data processing module 9 includes:

[0138] A data acquisition unit, connected to the pressure sensor 5, the temperature sensor 6 and the water quality sensor 7, for receiving water level, water quality and temperature data;

[0139] The edge computing unit uses the Kalman filter algorithm to perform real-time denoising on the collected data and fits the water level change trend through the least squares method;

[0140] The artificial intelligence analysis unit builds a groundwater level prediction model based on the long short-term memory network (LSTM), predicts water level changes in the next seven days, and detects abnormal fluctuations;

[0141] The data storage unit is fixed on the upper surface of the partition 4, has a built-in high-capacity storage device, and supports data compression storage and breakpoint resume functions;

[0142] The data encryption and security unit uses the AES-256 encryption algorithm to protect data storage and transmission.

[0143] Specifically, the data acquisition unit

[0144] Hardware configuration: integrated multiplexer and 16-bit high-precision analog-to-digital converter ADC, used to convert the analog signal output by the sensor into a digital signal.

[0145] Function description: The data of the pressure sensor 5, the temperature sensor 6 and the water quality sensor 7 are collected simultaneously, and the sampling frequency can be set to 1Hz to 10Hz.

[0146] Anti-interference design: Use hardware filtering circuits such as RC low-pass filters and software anti-noise algorithms such as moving average filtering to ensure the stability and accuracy of data transmission.

[0147] Edge computing unit

[0148] Real-time filtering: Use the Kalman filter algorithm to remove noise from the collected data. The calculation process is as follows:

[0149]

[0150] in, is the estimated value at the current moment, K k is the Kalman gain.

[0151] Dynamic trend analysis: Fitting water level change trend based on least squares method:

[0152] H(t)=at+b

[0153] Where a and b are fitting parameters, which are used to describe the trend of water level changes over time.

[0154] Artificial Intelligence Analysis Unit

[0155] LSTM model: A water level prediction model is built based on the long short-term memory network LSTM, which inputs the water level data of the past 7 days and outputs the water level prediction value for the next 7 days.

[0156] Anomaly detection: Compare current data with historical trends and trigger an anomaly alarm when the water level changes beyond the set threshold range.

[0157] Data storage unit

[0158] Storage device: built-in 32GB SD card, supports multiple storage formats such as JSON and CSV.

[0159] Function description: Realize data compression storage and breakpoint resume, and historical data can be saved for no less than 6 months.

[0160] Data Encryption and Security Unit

[0161] Encryption mechanism: AES-256 encryption algorithm is used to encrypt data to ensure data security.

[0162] Integrity verification: The integrity of data during storage and transmission is verified through a hash verification algorithm.

[0163] The wireless transmission module 10 includes:

[0164] Multi-mode communication unit, supporting NB-IoT, LoRa, 4G and 5G communication modes, and can dynamically switch communication protocols according to the communication environment;

[0165] The signal amplification and stabilization unit is connected to the data processing module 9 to enhance the strength and stability of signal transmission;

[0166] Automatic reconnection unit, used to automatically restore the connection after communication interruption and upload cached data;

[0167] Low power consumption optimization unit reduces energy consumption by optimizing communication protocols.

[0168] Specifically, the multi-mode communication unit

[0169] Communication support: Supports NB-IoT, LoRa, 4G and 5G communication modes.

[0170] Dynamic switching: Built-in signal detection algorithm can automatically switch communication protocols according to network conditions to ensure signal coverage and data transmission stability.

[0171] Signal amplification and stabilization unit

[0172] Gain amplification: Equipped with high-gain omnidirectional antenna and power amplifier to enhance the signal propagation strength in underground environments.

[0173] Anti-interference design: Add an electromagnetic shielding layer to reduce the impact of external electromagnetic interference on signal transmission.

[0174] Automatic reconnection unit

[0175] Disconnection detection: monitor the communication status in real time and trigger the automatic reconnection mechanism when the signal is interrupted.

[0176] Cache recovery: Data during the interruption is temporarily stored locally and uploaded in batches after reconnection to ensure that data is not lost.

[0177] Low power consumption optimized unit

[0178] Timed wake-up: The communication module enters the sleep state during the non-data transmission period and wakes up to transmit data only within the set time period.

[0179] Batch transmission: Reduce the number of communications by sending in batches, further reducing power consumption.

[0180] The energy management module 13 includes:

[0181] The lithium battery 11 is fixed on the third partition plate 4 of the housing 1 and is used to provide power support;

[0182] An environmental energy harvesting unit, including a high-efficiency monocrystalline silicon solar cell panel 16, is disposed in the top area of ​​the housing 1;

[0183] The power monitoring unit monitors the battery power in real time and triggers an alarm when the power level falls below the threshold;

[0184] The low power control unit switches to low power mode to extend the operating time when the device is idle.

[0185] Specifically, the environmental energy harvesting unit

[0186] Solar cell panel 16: installed on the top partition plate 4, a single crystal silicon solar cell panel with an energy conversion efficiency of not less than 20%.

[0187] Auxiliary energy collection: Optional micro wind power generation module and temperature difference power generation module for energy supplement in special environments.

[0188] Lithium battery11

[0189] Arrangement: fixed inside the energy management module 13 and connected to the environmental energy collection unit.

[0190] Function description: Provide backup power for the device. A single full charge can support the device to operate for at least 3 months.

[0191] Power Monitoring Unit

[0192] Real-time monitoring: Monitor the battery status through voltage and current sensors, display the remaining power and upload it to the remote monitoring system.

[0193] Low battery alarm: When the battery level drops below the preset threshold, local and remote alarms are triggered.

[0194] Low power control unit

[0195] Energy-saving design: When the device is idle, non-core modules are turned off and only the data acquisition function is kept running.

[0196] Intelligent energy dispatch unit

[0197] Energy allocation: Prioritize energy allocation to data acquisition and wireless transmission modules to ensure the normal operation of core functions.

[0198] Housing and mounting module including:

[0199] The outer shell 1 is made of corrosion-resistant material, and the inner bottom is fixedly connected with a bottom plate 2, and the bottom plate 2 is provided with filtering holes 12 for filtering impurities in groundwater;

[0200] The partition 4 is connected to the housing 1 through the sealing ring 3 to form a multi-layer modular structure for separating different functional modules;

[0201] The gear 14 and the rack 15, the gear 14 rotates on the upper surface of the second layer of the partition 4 and the rack 15 slides on the inner wall of the shell 1, and the gear 14 is driven to rotate by pulling the rack 15 so that the two racks 15 are retracted at the same time;

[0202] The observation window 17 and the indicator light 18 are respectively fixed inside the top partition 4 and are used to display the equipment operation status and fault warning.

[0203] Specifically, when the device needs to be placed in the monitoring well, first pull the rack 15. When the rack 15 slides, it will drive the gear 14 to rotate. When the rack 15 slides a suitable distance, the device can be placed in the well. The rack 15 is used to support the device at the wellhead, which is convenient for the installation of the device. After the installation is completed, the rack 15 can be pulled out, or a slider can be set on the lower surface of the rack 15. A sliding groove for the sliding of the slider is provided inside the partition 4. The sliding groove limits the distance of the slider to prevent the rack 15 from being drawn out, which is used to continuously support the device, and after the device is disassembled, the rack 15 can be lifted to facilitate the removal of the device from the well. A fixing hole is provided at the end of the rack 15, which can be installed at the side of the well with screws to prevent it from shaking. By adjusting the sliding distance and direction of the rack, the operator can easily put the device into the well without complicated lifting tools, which simplifies the installation process of the device.

[0204] The filter hole 12 is used to prevent water from bringing in impurities when entering the interior of the housing 1. The filter hole effectively intercepts impurities such as silt and particles when water enters the interior of the device, protects internal sensors (such as pressure sensors, temperature sensors, and water quality sensors) and electronic modules from being contaminated or damaged, extends the service life of the device and improves monitoring accuracy;

[0205] The sealing ring is used to prevent water from entering the upper layer of the housing 1. The multiple sealing design of the sealing ring can prevent water vapor or liquid from entering the upper area of ​​the housing, protecting the electronic modules (such as the data processing module 9 and the wireless transmission module 10) from water damage, thereby improving the waterproof performance and operational reliability of the device.

[0206] The application process of the artificial intelligence analysis unit based on the long short-term memory network LSTM model includes the data preparation stage, specifically:

[0207] The monitoring data are preprocessed. The monitoring data include water level data, temperature data, and environmental factors. The preprocessing includes data cleaning, interpolation to fill missing values, and normalization. All data are normalized to the range of [0,1] or [-1,1]. A time window is constructed, and the data of the past k time points are used as the model input to predict the water level changes at the next m time points. The input format is:

[0208] X input ={(H t-k , H t-k+1 , ..., H t-1 )},Y output ={H t , H t-1 , ..., H t-m}

[0209] Among them, X input The input data is the water level at the past k time points;

[0210] H t-k ,H t-k+1 ,…,H t-1 The water level data in the past;

[0211] Y output To output data, the water level at m future time points;

[0212] H t ,H t+1 ,…,H t+m is the target value of the future water level height.

[0213] Specifically, it defines how monitoring data is standardized and formatted through preprocessing to provide a suitable input form for the Long Short-Term Memory Network (LSTM).

[0214] Ensure consistency between different time points and multi-dimensional data (such as water level, temperature, environmental factors, etc.) to reduce the impact of noise in model training.

[0215] The time window method is used to organize the input-output relationship and build a time series foundation for prediction tasks.

[0216] Effect:

[0217] Data cleaning and normalization can remove outliers and scale differences and enhance the convergence and generalization capabilities of the model.

[0218] The way the time window is constructed enables LSTM to effectively capture the correlation between historical data and future trends, providing higher accuracy for groundwater level prediction.

[0219] The LSTM structure of the long short-term memory network in the artificial intelligence analysis unit includes:

[0220] The forget gate is used to decide whether to keep or forget historical information. Its calculation formula is:

[0221] f t =σ(W f ·[h t-1 , x t ]+b f )

[0222] Among them, f t is the output value of the forget gate, ranging from [0,1], which determines the forgetting ratio of historical information;

[0223] σ is the activation function, using the sigmoid function, and the output range is [0,1];

[0224] W f is the weight matrix of the forget gate;

[0225] h t-1 is the hidden state at the previous moment;

[0226] xt is the input at the current moment;

[0227] b f is the bias vector of the forget gate;

[0228] The input gate is used to update the memory cell state. Its calculation formula is:

[0229] i t =σ(W i ·[h t-1 ,x t +b i ),

[0230]

[0231] Among them, it is the output value of the input gate, ranging from [0,1], indicating the memory ratio of the current new information;

[0232] is the value of the candidate memory unit;

[0233] W i ,W C is the weight matrix of the input gate and the candidate memory unit;

[0234] b i , b C is the bias vector of the input gate and the candidate memory unit;

[0235] C t is the state of the memory unit at the previous moment;

[0236] C t-1 is the memory unit state at the previous moment;

[0237] ⊙ is the element-wise multiplication operation;

[0238] The output gate is used to generate the current hidden state for use at the next moment or in the next layer. The calculation formula is:

[0239] o t =σ(W o ·[h t-1 , x t ]+b o ), h t =o t ⊙tanh(C t )

[0240] Among them, t is the output value of the output gate, ranging from [0,1];

[0241] W o is the weight matrix of the output gate;

[0242] b o is the bias vector of the output gate;

[0243] h t is the hidden state at the current moment;

[0244] tanh(C t ) is the nonlinear activation value of the memory unit.

[0245] Specifically, the core mechanisms of the LSTM model are described, including the forget gate, input gate, and output gate, which enable the model to memorize long-term series relationships while selectively forgetting unimportant information.

[0246] The calculation formula of each gate is given to illustrate how to dynamically adjust the state and hidden state of the memory unit to achieve effective modeling of the dynamic changes of groundwater levels.

[0247] Effect:

[0248] The forget gate can filter out redundant or irrelevant historical information and increase the model's attention to key features.

[0249] The input gate can flexibly introduce new information at the current time point, thereby dynamically updating the state of the memory unit.

[0250] The output gate generates a hidden state for the next calculation or model prediction, enabling the model to adapt to complex tasks with long-term dependencies.

[0251] The introduction of this gating mechanism enables the model to show better robustness and accuracy in responding to groundwater level changes.

[0252] The long short-term memory network LSTM of the artificial intelligence analysis unit is trained and optimized through the following steps:

[0253] Use mean square error MSE as the loss function to calculate the error between the predicted value and the true value. The loss function formula is:

[0254]

[0255] in, MSE is the mean square error, which represents the average sum of square errors between the model prediction value and the true value;

[0256] N is the total number of samples;

[0257] H i,pred is the predicted water level height of the i-th sample;

[0258] H i,true is the true water level height of the i-th sample;

[0259] The Adam optimization algorithm is used to iteratively optimize the model parameters through the following formula:

[0260]

[0261] Among them, θ t is the model parameter at the current moment;

[0262] θ t-1 is the model parameter of the previous moment;

[0263] η is the learning rate, which indicates the step size of parameter update;

[0264] is the gradient of the loss function with respect to the model parameters.

[0265] Specifically, it explains how to use the mean square error (MSE) as the loss function to evaluate the error between the predicted result and the true value, and iteratively update the model parameters through the Adam optimization algorithm.

[0266] The goal of model training is described, which is to minimize the prediction error so that the model can more accurately predict future water level changes.

[0267] Effect:

[0268] The loss function (MSE) can quantify the degree of deviation of model prediction and provide an optimization target for model training.

[0269] The Adam optimization algorithm combines the advantages of momentum and adaptive learning rate, which can quickly and efficiently adjust model parameters and accelerate model convergence.

[0270] Through this training and optimization process, the LSTM model is able to achieve higher prediction accuracy in the complex task of groundwater level change.

[0271] The long short-term memory network LSTM of the artificial intelligence analysis unit predicts the water level change trend and evaluates the model based on the model prediction results, including:

[0272] Using the input data of the time window, the water level changes at the next m time points are predicted, and the output is:

[0273]

[0274] in, is the predicted water level at the i-th time point in the future;

[0275] f is the function expression of the LSTM model;

[0276] θ is the trained model parameter;

[0277] H t-k ,H t-k+1 ,…,H t-1 Historical input data for the time window;

[0278] The model performance is evaluated using the following evaluation indicators:

[0279] Mean Square Error (MSE):

[0280]

[0281] Mean absolute error (MAE):

[0282]

[0283] Coefficient of determination R 2 :

[0284]

[0285] in, is the mean of the true water level height;

[0286] R 2 is the coefficient of determination, and the closer the value is to 1, the better the model prediction performance is.

[0287] Specifically, the application process of the LSTM model in groundwater level prediction is described, including how to predict future water levels based on time window inputs and how to quantify the performance of the model through model evaluation indicators.

[0288] Three evaluation indicators (MSE, MAE, R 2 ), which are used to measure the bias, absolute error and goodness of model fit of the prediction respectively.

[0289] Effect:

[0290] The prediction formula defines the input and output relationship of the LSTM model and clarifies how to use the trained model for practical applications.

[0291] By MSE, MAE and R 2 Evaluation indicators can comprehensively measure the predictive ability of the model and ensure that the model has good applicability and reliability under different environmental conditions.

[0292] This prediction and evaluation process can help users determine whether the model's performance meets actual needs, thereby optimizing groundwater resource management or geological disaster early warning systems.

[0293] The introduction of the LSTM model can effectively process time series data and capture the complex dynamic characteristics of groundwater level changes over time, especially showing better prediction capabilities when dealing with long-term dependencies.

[0294] The comprehensive design of data preprocessing, model optimization and performance evaluation has improved the overall reliability and applicability of the system, and can be widely used in groundwater resource management, geological disaster early warning and other fields.

[0295] Reduce the limitations of traditional monitoring systems in modeling dynamic changes and achieve more accurate groundwater level monitoring and management.

[0296] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A groundwater level continuous monitoring device, characterized in that: include: A monitoring probe module, arranged in the bottom area of ​​the housing (1), for monitoring the groundwater level and related environmental parameters; A data processing module (9), arranged in the middle area of ​​the housing (1), for collecting, processing, analyzing and storing monitoring data; A wireless transmission module (10) is connected to the data processing module (9) and is used to upload the processed monitoring data to a remote monitoring system; An energy management module (13), arranged in a layer above the data processing module (9) in the housing (1), and used to provide power support for the groundwater level continuous monitoring device; The housing and installation module comprises a housing (1), a base plate (2) and an internal multi-layer partition plate (4), which are used to protect the device structure and adapt to different installation environments.

2. A groundwater level continuous monitoring device according to claim 1, characterized in that: The monitoring probe module comprises: The pressure sensor (5), the temperature sensor (6) and the water quality sensor (7) are respectively fixed at the interface of the plug plate (8) of the bottom partition plate (4) and are used to monitor the water level, temperature and water quality parameters of the groundwater; The plug plate (8) is connected to the partition plate (4), the outer walls of the pressure sensor (5), the temperature sensor (6) and the water quality sensor (7) are arranged inside the partition plate (4), and the partition plate (4) is fixedly connected to the housing (1) via a sealing ring (3) to enhance the sealing performance.

3. A groundwater level continuous monitoring device according to claim 2, characterized in that: The data processing module (9) comprises: A data acquisition unit connected to the pressure sensor (5), the temperature sensor (6) and the water quality sensor (7) and used for receiving water level, water quality and temperature data; The edge computing unit uses the Kalman filter algorithm to perform real-time denoising on the collected data and fits the water level change trend through the least squares method; The artificial intelligence analysis unit builds a groundwater level prediction model based on the long short-term memory network (LSTM), predicts water level changes in the next seven days, and detects abnormal fluctuations; A data storage unit, fixed on the upper surface of the partition (4), having a built-in high-capacity storage device and supporting data compression storage and breakpoint resume functions; The data encryption and security unit uses the AES-256 encryption algorithm to protect data storage and transmission.

4. The groundwater level continuous monitoring device according to claim 1, characterized in that: The wireless transmission module (10) comprises: Multi-mode communication unit, supporting NB-IoT, LoRa, 4G and 5G communication modes, and can dynamically switch communication protocols according to the communication environment; A signal amplification and stabilization unit, connected to the data processing module (9), to enhance the strength and stability of signal transmission; Automatic reconnection unit, used to automatically restore the connection after communication interruption and upload cached data; Low power consumption optimization unit reduces energy consumption by optimizing communication protocols.

5. The groundwater level continuous monitoring device according to claim 1, characterized in that: The energy management module (13) comprises: A lithium battery (11), fixed on the third partition (4) of the housing (1), for providing power support; An environmental energy harvesting unit, comprising a high-efficiency monocrystalline silicon solar cell panel (16), is disposed in the top area of ​​the housing (1); The power monitoring unit monitors the battery power in real time and triggers an alarm when the power level falls below the threshold; The low power control unit switches to low power mode to extend the operating time when the device is idle.

6. The groundwater level continuous monitoring device according to claim 1, characterized in that: The housing and the mounting module include: The outer shell (1) is made of corrosion-resistant material, and the inner bottom is fixedly connected to a bottom plate (2), and the bottom plate (2) is provided with filtering holes (12) for filtering impurities in groundwater; The partition (4) is connected to the housing (1) via the sealing ring (3) to form a multi-layer modular structure for separating different functional modules; The gear (14) and the rack (15) are configured such that the gear (14) rotates on the upper surface of the second partition plate (4) and the rack (15) slides on the inner wall of the housing (1), and the gear (14) is driven to rotate by pulling the rack (15) so that the two racks (15) are simultaneously retracted; The observation window (17) and the indicator light (18) are respectively fixed inside the top partition (4) and are used to display the equipment operation status and fault warning.

7. The groundwater level continuous monitoring device according to claim 3, characterized in that: The application process of the artificial intelligence analysis unit based on the long short-term memory network LSTM model includes a data preparation stage, specifically: The monitoring data are preprocessed. The monitoring data include water level data, temperature data, and environmental factors. The preprocessing includes data cleaning, interpolation to fill missing values, and normalization. All data are normalized to the range of [0,1] or [-1,1]. A time window is constructed, and the data of the past k time points are used as the model input to predict the water level changes at the next m time points. The input format is: X input ={(H t-k ,H t-k+1 ,...,H t-1 )},Y output ={H t ,H t-1 ,...,H t+m } Among them, X input The input data is the water level at the past k time points; H t-k ,H t- k +1 ,…,H t-1 The water level data in the past; Y output To output data, the water level at m future time points; H t ,H t+1 ,…,H t+m is the target value of the future water level height.

8. The groundwater level continuous monitoring device according to claim 7, characterized in that: The long short-term memory network LSTM structure in the artificial intelligence analysis unit includes: The forget gate is used to decide whether to keep or forget historical information. Its calculation formula is: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t is the output value of the forget gate, ranging from [0,1], which determines the forgetting ratio of historical information; σ is the activation function, using the sigmoid function, and the output range is [0,1]; W f is the weight matrix of the forget gate; h t-1 is the hidden state at the previous moment; x t is the input at the current moment; b f is the bias vector of the forget gate; The input gate is used to update the memory cell state. Its calculation formula is: i t =σ(W i ·[h t-1 ,x t +b i ), Among them, it is the output value of the input gate, ranging from [0,1], indicating the memory ratio of the current new information; is the value of the candidate memory unit; W i ,W C is the weight matrix of the input gate and the candidate memory unit; b i , b C is the bias vector of the input gate and the candidate memory unit; C t is the state of the memory unit at the previous moment; C t-1 is the memory unit state at the previous moment; ⊙ is the element-wise multiplication operation; The output gate is used to generate the current hidden state for use at the next moment or in the next layer. The calculation formula is: o t =σ(W o ·[h t-1 ,x t ]+b o ),h t =o t ☉tanh(C t ) Among them, t is the output value of the output gate, ranging from [0,1]; W o is the weight matrix of the output gate; b o is the bias vector of the output gate; h t is the hidden state at the current moment; tanh(C t ) is the nonlinear activation value of the memory unit.

9. A groundwater level continuous monitoring device according to claim 8, characterized in that: The long short-term memory network LSTM of the artificial intelligence analysis unit is trained and optimized through the following steps: Use mean square error MSE as the loss function to calculate the error between the predicted value and the true value. The loss function formula is: in, MSE is the mean square error, which represents the average sum of square errors between the model prediction value and the true value; N is the total number of samples; H i,pred is the predicted water level height of the i-th sample; H i,true is the true water level height of the i-th sample; The Adam optimization algorithm is used to iteratively optimize the model parameters through the following formula: Among them, θ t is the model parameter at the current moment; θ t-1 is the model parameter of the previous moment; η is the learning rate, which indicates the step size of parameter update; is the gradient of the loss function with respect to the model parameters.

10. The groundwater level continuous monitoring device according to claim 9, characterized in that: The long short-term memory network LSTM of the artificial intelligence analysis unit performs water level change trend prediction and model evaluation based on the model prediction results, specifically including: Using the input data of the time window, the water level changes at the next m time points are predicted, and the output is: in, is the predicted water level at the i-th time point in the future; f is the function expression of the LSTM model; θ is the trained model parameter; H t-k ,H t-k+1 ,…,H t-1 Historical input data for the time window; The model performance is evaluated using the following evaluation indicators: Mean Square Error (MSE): Mean absolute error (MAE): Coefficient of determination (R 2 ): in, is the mean of the true water level height; R 2 is the coefficient of determination, and the closer the value is to 1, the better the model prediction performance is.

Images