Water conservancy gate valve flow dynamic adjustment management and control system based on deep learning

Abstract

The application belongs to the technical field of water conservancy gate valve supervision, and specifically relates to a water conservancy gate valve flow dynamic adjustment and control system based on deep learning, which comprises a gate valve flow multi-dimensional sensing and collecting module, a collected data adaptive preprocessing module, a deep learning flow prediction modeling module, a dynamic opening degree precise adjustment module, a whole-process control and scheduling module and a water hammer risk adaptive suppression module; the application collects parameters such as water level, flow and opening degree through multi-dimensional collection, and carries out adaptive denoising, abnormality repair and standardization processing; the application also improves a deep learning model to predict natural flow without intervention, calculates the optimal opening degree and adjustment rate according to the flow deviation classification, and outputs a compensation coefficient to correct the adjustment rate, so as to realize precise dynamic adjustment of the gate valve flow, effectively suppress the pipeline water hammer risk, and greatly improve the safety and precision of the water conservancy gate valve control.

Description

Technical Field

[0001] This invention relates to the field of hydraulic gate valve monitoring technology, specifically a hydraulic gate valve flow dynamic adjustment and control system based on deep learning. Background Technology

[0002] As core control equipment for water conveyance, agricultural irrigation, and flood control, the accuracy of flow regulation, response time, and operational safety of hydraulic gate valves directly affect the scheduling efficiency, water conveyance stability, and service life of pipeline equipment in water conservancy projects. With the advancement of smart water conservancy construction, the traditional gate valve control mode has the following defects in actual operation:

[0003] Currently, fixed opening manual adjustment or simple PID closed-loop adjustment are commonly used. No deep learning non-intervention natural flow prediction model has been built, so it is impossible to predict the flow evolution trend. It can only passively respond to real-time deviations, and the adjustment is not forward-looking enough. Moreover, the opening adjustment relies on experience fixed rate and parameters, and does not combine flow deviation to optimize the optimal opening and the matching rate, resulting in poor adjustment accuracy and matching degree.

[0004] At the same time, it lacks real-time calculation, classification and adaptive suppression mechanisms for water hammer risk, which can easily cause water hammer impact and damage to equipment, and cannot take into account both the accuracy of flow regulation and the safety of water conservancy operation.

[0005] Therefore, developing a dynamic flow regulation and control system for hydraulic gate valves that integrates deep learning flow prediction, dynamic precise opening adjustment, and adaptive suppression of water hammer risk has become an urgent problem to be solved in the field of intelligent control of hydraulic gate valves. Summary of the Invention

[0006] The purpose of this invention is to provide a deep learning-based dynamic flow regulation and control system for hydraulic gate valves to address the aforementioned technical deficiencies.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a deep learning-based dynamic regulation and control system for hydraulic gate valve flow, comprising a gate valve flow multi-dimensional sensing and acquisition module, an acquisition data adaptive preprocessing module, a deep learning flow prediction and modeling module, a dynamic opening degree precise adjustment module, and a full-process control and scheduling module;

[0008] The gate valve flow multi-dimensional sensing and acquisition module is used to collect multi-dimensional key parameters of gate valve operation. It has the ability to adaptively adjust the acquisition frequency and realizes real-time data transmission, interruption caching and recovery retransmission through the LoRa-WiFi dual-mode transmission module. The acquired data adaptive preprocessing module is used to adaptively denoise, repair outliers, standardize data and normalize dimensions of raw data, and output a standardized dataset adapted to deep learning models.

[0009] The deep learning-based flow prediction modeling module predicts natural flow under no-intervention conditions during a preset future period based on an improved deep learning model. The dynamic opening precision adjustment module calculates the main adjustment deviation based on the predicted natural flow and the preset target flow, determines the optimal valve opening and tiered adjustment rate, and achieves precise flow adjustment using a feedforward predictive adjustment mode. The full-process control and scheduling module is used to set flow target values, coordinate the collaborative operation of various modules, monitor the system status in real time, trigger abnormal warnings and global scheduling, and store all system operation data.

[0010] Furthermore, the gate valve flow multi-dimensional sensing and acquisition module integrates a submersible static pressure water level sensor, an ultrasonic Doppler flow sensor, a high-precision angle sensor, a differential pressure sensor, and an environmental compensation sensor; the water level sensor is located 5m upstream and 5m downstream of the gate valve, the flow sensor is located 10m downstream of the gate valve, the angle sensor is located at the rotating shaft of the gate valve actuator, and the differential pressure sensor is located at the inlet and outlet ends of the gate valve.

[0011] Furthermore, the adaptive preprocessing module for the acquired data adopts an improved wavelet threshold denoising algorithm for adaptive denoising. It identifies outliers through a 3σ confidence interval and repairs them using a quadratic interpolation method. The data is then converted to the [0,1] interval to complete standardization, and core features are extracted to complete dimensional regularization.

[0012] Furthermore, the deep learning traffic prediction modeling module divides the dataset into training, validation, and test sets in a 7:2:1 ratio. The model training uses an adaptive learning rate optimization algorithm and predicts the output of natural traffic flow prediction values ​​for future time periods without intervention.

[0013] Furthermore, the dynamic opening precision adjustment module uses the difference between the preset target flow and the predicted natural flow as the sole main adjustment deviation, and classifies the deviation into three levels: small, medium, and large.

[0014] Furthermore, the dynamic opening precision adjustment module combines the real-time opening of the gate valve, the water flow impact correction coefficient, and the water flow disturbance coefficient to calculate the optimal target opening of the gate valve, and corrects the adjustment accuracy in real time based on the opening feedback from the actuator.

[0015] Furthermore, the dynamic opening precision adjustment module introduces a water hammer suppression compensation coefficient to correct the graded adjustment rate, matching the valve's mechanical response with the pipeline's water hammer protection requirements.

[0016] Furthermore, it also includes a water hammer risk adaptive suppression module. The water hammer risk adaptive suppression module constructs a water hammer pressure calculation model based on the pipeline transient flow theory, and calculates the instantaneous water hammer pressure in the pipeline in real time by combining the opening change and the pressure difference between the inlet and outlet of the gate valve.

[0017] Furthermore, the water hammer risk adaptive suppression module is divided into three risk levels—low, medium, and high—based on instantaneous water hammer pressure, and outputs corresponding water hammer suppression compensation coefficients for different levels.

[0018] Compared with the prior art, the beneficial effects of the present invention are:

[0019] 1. In this invention, multi-dimensional sensing and acquisition, adaptive data preprocessing and deep learning flow prediction are used to adopt feedforward predictive opening adjustment to solve the problems of traditional gate valve regulation lag and flow overshoot, greatly improve regulation accuracy and response speed, and realize intelligent dynamic control of gate valve flow throughout the entire process.

[0020] 2. In this invention, the water hammer risk adaptive suppression function calculates the water hammer pressure in real time, classifies the risk level and corrects the adjustment rate, effectively suppressing water hammer impact in pipelines, protecting the safety of gate valves and pipeline equipment, and taking into account both the accuracy of flow regulation and the safety of water conservancy operation. Attached Figure Description

[0021] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0022] Figure 1 This is a schematic diagram of the overall system structure of the present invention;

[0023] Figure 2 This is a schematic diagram of the operation method of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Example 1: Refer to Figure 1-2 As shown, the hydraulic gate valve flow dynamic regulation and control system based on deep learning proposed in this invention includes a gate valve flow multi-dimensional sensing and acquisition module, an acquisition data adaptive preprocessing module, a deep learning flow prediction and modeling module, a dynamic opening degree precise adjustment module, and a full-process control and scheduling module.

[0026] The gate valve flow multi-dimensional sensing and acquisition module is used to collect multi-dimensional key parameters during the operation of the gate valve, realize real-time data capture, anti-interference transmission and cache backup, provide high-quality raw data source for subsequent data processing, prediction and regulation of the system, and has the ability to adaptively adjust the acquisition frequency to adapt to complex water conservancy site environments.

[0027] Specifically, the gate valve flow multi-dimensional sensing and acquisition module is deployed in the upstream and downstream areas of the gate valve and at the gate valve actuator, integrating a submersible static pressure water level sensor, an ultrasonic Doppler flow sensor, a high-precision angle sensor, a differential pressure sensor and an environmental compensation sensor.

[0028] The water level sensors are installed 5m upstream and 5m downstream of the gate valve to capture the upstream water level Hu and the downstream water level Hd in real time. Preferably, the acquisition frequency can be adaptively adjusted according to the water flow fluctuation amplitude. When the water flow fluctuation amplitude is >0.1m, the acquisition frequency is increased from 1 time / 10s to 1 time / 2s. When the fluctuation amplitude is <0.05m, the acquisition frequency is reduced to 1 time / 30s, which effectively reduces the redundancy of invalid data.

[0029] The flow sensor is installed 10m downstream of the gate valve to collect the actual flow rate Qr of the gate valve in real time. During the collection process, abnormal pulse signals generated by water flow vortices are automatically filtered to ensure the authenticity of the flow data. The angle sensor is installed at the shaft of the gate valve actuator to collect the current opening degree θ of the gate valve in real time with an accuracy of 0.1°. The differential pressure sensor is installed at the inlet and outlet of the gate valve to collect the pressure difference ΔP between the inlet and outlet. The environmental compensation sensor collects the ambient temperature T and humidity S to compensate for the collection errors of various sensors.

[0030] Furthermore, all the raw data collected is transmitted in real time through the LoRa-WiFi dual-mode transmission module, taking into account both transmission stability and transmission rate in complex water conservancy environments. At the same time, the module has a built-in data buffer unit with a capacity of 1 hour. When the transmission is interrupted, the collected data is automatically buffered and retransmitted after the transmission is restored, which helps to avoid data loss. Finally, all the raw collected data is transmitted to the data adaptive preprocessing module.

[0031] The adaptive preprocessing module for data acquisition performs noise reduction, outlier repair, data standardization, and dimensional regularization on the raw data transmitted by the gate valve flow multi-dimensional sensing acquisition module. This eliminates the impact of interference factors on data quality and outputs standardized data that meets the input requirements of the deep learning model. It also has the ability to self-check data quality to ensure the accuracy of subsequent predictions and adjustments.

[0032] Specifically, the adaptive preprocessing module receives the raw data (Hu, Hd, Qr, θ, ΔP, T, S) transmitted by the gate valve flow multi-dimensional sensing acquisition module. It first starts adaptive denoising processing and adopts an improved wavelet threshold denoising algorithm. It does not require manual setting of the threshold and automatically adjusts the threshold parameters according to the data fluctuation characteristics to remove noise data caused by environmental interference and sensor errors. This is different from the existing fixed threshold denoising method and improves the denoising targeting.

[0033] Then, outlier repair is performed. By calculating the 3σ confidence interval of each parameter, outlier data outside the interval is identified. Combined with the valid data at adjacent time points, the outlier repair is completed by using the quadratic interpolation method to avoid the deviation of subsequent model predictions due to outlier data.

[0034] Next, data standardization is performed to convert all parameters to the [0,1] range, eliminating the impact of differences in the dimensions of different parameters on model training. Finally, data dimensionality is normalized to extract the core features related to gate valve flow regulation, remove redundant features, form a standardized dataset, and transmit the standardized dataset to the deep learning flow prediction modeling module.

[0035] The deep learning flow prediction modeling module is based on all real-time standardized operating data at the current moment. Through an improved deep learning model, it completes the flow prediction for a future preset time period (preferred, 10 minutes). The prediction object is the future flow of water when the valve is kept at its current opening without any adjustment or intervention. This provides data support for the operation of the dynamic opening precision adjustment module and adapts to the dynamic changes in water conservancy conditions.

[0036] Specifically, the deep learning traffic prediction modeling module receives the standardized dataset transmitted by the data collection adaptive preprocessing module. First, the dataset is divided into training set, validation set and test set in a ratio of 7:2:1. The training set is used for model training, the validation set is used to adjust model parameters, and the test set is used to verify the model prediction accuracy.

[0037] Subsequently, an improved Attention-LSTM deep learning model was constructed. Unlike existing conventional LSTM models, this model adds a gate valve characteristic attention factor to the hidden layer of the network, which automatically focuses on the real-time parameters with the highest weights that affect water flow. At the same time, a gate valve wear factor ω is introduced to make up for the deficiency of existing models that do not consider the flow attenuation caused by long-term wear of equipment. During the model training process, an adaptive learning rate optimization algorithm is adopted to dynamically adjust the learning rate according to the prediction error of the validation set, so as to avoid the model getting stuck in local optima.

[0038] Furthermore, after training, all real-time operating data is input into the model. Based on the currently collected water level difference, valve opening, and equipment wear parameters, the model is directly extrapolated and calculated. Only the predicted natural flow rate Qp under the uninterrupted state for the next preset period is output. The calculated predicted natural flow rate Qp is then directly transmitted to the dynamic opening precision adjustment module. The prediction formula is calculated based on real-time operating parameters.

[0039] ;

[0040] Qp: The natural flow rate of water when the current valve opening is not adjusted during the future preset time period;

[0041] β: Water level difference influence coefficient, obtained through on-site hydraulic condition calibration test, with the preferred value range being [0.3, 0.5].

[0042] Hu: Water level upstream of the gate valve, collected in real time by the front-end sensing sensor; Hd: Water level downstream of the gate valve, collected in real time by the front-end sensing sensor.

[0043] γ: Valve opening influence coefficient, obtained by on-site calibration based on the gate valve diameter, preferably in the range of [0.1, 0.3].

[0044] θ: Real-time opening degree of the gate valve, which is collected in real time by the angle sensor of the front-end sensing module;

[0045] δ: Equipment wear correction coefficient, obtained by fixed calibration based on the gate valve's factory parameters, preferably in the range of [0.05, 0.1].

[0046] ω: Gate valve wear attenuation factor, with a value range of [0, 0.1]. It is calculated based on the cumulative running time of the equipment: ω = 0.0001t, where t is the cumulative running time of the gate valve. It should be noted that the running time data is obtained by retrieving the system's historical running database.

[0047] The dynamic opening precision adjustment module is based on the future uninterrupted natural flow prediction value and the calculation of the unique main adjustment deviation by combining the preset target flow of the scheduling. Based on this, the optimal adjustment opening and adjustment rate of the gate valve are calculated. The feedforward prediction adjustment mode is adopted to compensate for the subsequent flow deviation in advance, thereby solving the problems of traditional valve adjustment lag and water flow fluctuation overshoot from the root and achieving precise adjustment under safety constraints.

[0048] Specifically, the dynamic opening precision adjustment module receives the predicted flow value Qp from the deep learning flow prediction modeling module, the preset target flow Qt from the full-process control and scheduling module, and the water hammer suppression compensation coefficient μ from the water hammer risk adaptive suppression module. First, the unique main adjustment deviation ΔQ is calculated, which is the sole core basis for the entire adjustment process: ΔQ = Qt - Qp.

[0049] The meaning of the main regulation deviation ΔQ: The difference between the preset target flow rate and the natural flow rate without intervention in the next 10 minutes (i.e., the preset future time period). It represents the flow rate difference between the subsequent water flow and the scheduling target if the valve is not adjusted. Based on this value, the valve opening is adjusted in advance to achieve feedforward advance control and avoid the lag defects of traditional control.

[0050] Then, the deviation levels are determined based on the magnitude of the main deviation, as shown below:

[0051] When |ΔQ| ≤ 0.05m 3 When the deviation is / s, it is considered a small deviation and no significant adjustment is needed.

[0052] When 0.05 m³ / s < |ΔQ| ≤ 0.2 m 3 When the deviation is / s, it is determined to be a moderate deviation, and routine adjustment is performed.

[0053] When |ΔQ| > 0.2m 3 When the deviation is / s, it is determined to be a large deviation, and a rapid adjustment is performed.

[0054] Furthermore, the dynamic opening precision adjustment module combines the current gate valve opening θ and on-site water flow disturbance parameters to calculate the optimal gate valve opening θo in real time using the opening adjustment analysis formula, as follows:

[0055] ;

[0056] θo: Optimal target opening degree of the gate valve;

[0057] θ: Real-time original opening degree of the gate valve, which is collected in real time by the gate valve flow multi-dimensional sensing and acquisition module;

[0058] k: Proportional adjustment coefficient, obtained through PID parameter adaptive tuning calculation, preferably in the range of [0.5, 1.0];

[0059] ΔQ: Main control deviation, which is the sole core basis for opening adjustment;

[0060] λ: Water flow impact correction coefficient, obtained through on-site pipeline water hammer test calibration, preferably in the range of [0.02, 0.05];

[0061] ε: Water flow disturbance coefficient, with a value range of [0, 0.03], obtained by calculation of ε = 0.001ΔP, where ΔP is the pressure difference between the inlet and outlet of the gate valve, which is obtained in real time by the gate valve flow multi-dimensional sensing and acquisition module.

[0062] Preferably, the graded regulation rate v is corrected based on the main regulation deviation level and the superimposed water hammer suppression compensation coefficient μ. The corrected graded rate can be referenced as follows:

[0063] ;

[0064] v: Regulation rate of the gate valve mechanism, in units ;

[0065] ΔQ: Main control deviation;

[0066] μ: Water hammer suppression compensation coefficient.

[0067] This graded speed regulation method matches the mechanical response characteristics of the valve with the water hammer protection requirements of the pipeline. During the regulation process, it relies on the adaptive rate to avoid excessive water flow impact caused by excessive regulation, which could damage the gate valve and pipeline. At the same time, it receives feedback from the gate valve actuator on the actual regulation opening in real time for correction to ensure regulation accuracy. After the parameter calculation is completed, the optimal target opening and regulation rate command of the gate valve are directly sent to the gate valve actuator on site to drive the gate valve to dynamically adjust the opening according to the set parameters.

[0068] The full-process control and scheduling module, as the core control unit of the system, is responsible for receiving the operating data of each module, setting the target value of the gate valve flow, coordinating the workflow of each module, realizing real-time monitoring of the system's operating status, anomaly warning and global scheduling, and ensuring the stable and efficient operation of the system.

[0069] Specifically, the full-process control and scheduling module receives standardized datasets from the adaptive preprocessing module, prediction accuracy reports from the deep learning flow prediction modeling module, and on-site adjustment execution data from the dynamic opening precision adjustment module. It monitors the operating status of each module in real time and presents the real-time changes and historical trends of key parameters such as gate valve flow, opening, water level, and pipeline water hammer pressure through data visualization.

[0070] Furthermore, based on the actual needs of water conservancy projects (such as irrigation, flood control, and water supply), target values ​​Qt for gate valve flow are set for different scenarios. Combined with real-time forecast data and on-site adjustment feedback data, the target values ​​are dynamically adjusted to ensure that the flow regulation meets the project requirements. When an anomaly is detected in a certain module (such as insufficient forecast accuracy, excessive adjustment deviation, or high risk of water hammer in the pipeline), an anomaly warning is immediately triggered, and an emergency dispatch command is generated to adjust the operating parameters of the relevant modules. If the anomaly cannot be resolved on its own, an alarm signal is issued to notify staff to intervene.

[0071] In addition, the full-process control and scheduling module is also responsible for storing all system operation data, model parameters, adjustment records, and historical operation data to form a historical database, providing data support for subsequent system optimization, fault diagnosis, and water conservancy scheduling decisions. At the same time, it transmits the configured Qt to the dynamic opening precision adjustment module in real time to ensure the coordinated linkage of each module and realize the full-process dynamic control of gate valve flow.

[0072] Example 2: Refer to Figure 1-2As shown, the difference between this embodiment and Embodiment 1 is that it also includes a water hammer risk adaptive suppression module. The water hammer risk adaptive suppression module directly calls the upstream determined optimal opening degree and the original opening degree to obtain the actual opening degree change, and performs real-time calculation of pipeline water hammer pressure, risk level classification and compensation coefficient solution. From the perspective of hydraulic safety, it constrains and corrects the previous adjustment command, taking into account both the flow regulation accuracy and pipeline operation safety, avoiding the problems of pipeline water hammer oscillation, pipe wall pressure exceeding limit and equipment erosion damage caused by rapid valve adjustment, and filling the deficiency of the current only flow regulation without active water hammer protection and suppression.

[0073] Specifically, the water hammer risk adaptive suppression module receives the optimal adjustment opening θo and the real-time original opening θ of the gate valve from the dynamic opening precision adjustment module, and simultaneously receives the inlet and outlet pressure difference ΔP collected in real time by the gate valve flow multi-dimensional sensing and acquisition module. Based on the pipeline transient flow theory, it constructs a water hammer pressure calculation model and directly calculates the actual predicted opening change Δθp through the difference between the two, i.e., Δθp = θo - θ:

[0074] Based on the change in pipe opening and the pipeline pressure difference parameter, the predicted water hammer pressure Ps inside the pipeline is calculated in real time. The calculation formula is as follows:

[0075] ;

[0076] Wherein, Ps: instantaneous water hammer pressure in the pipeline;

[0077] η: Water hammer conduction coefficient of pipe material, obtained by on-site calibration based on the material and diameter of the water supply pipe, with a preferred value range of [0.4, 0.7].

[0078] ΔP: Pressure difference between the inlet and outlet of the gate valve, which is collected in real time by the gate valve flow multi-dimensional sensing and acquisition module.

[0079] Furthermore, after calculating the instantaneous water hammer pressure Ps in the pipeline, the risk level can be classified based on the instantaneous water hammer pressure value, as follows:

[0080] When Ps≤15kPa, the risk is low and no parameter compensation is required;

[0081] When 15kPa < Ps ≤ 40kPa is considered medium risk, appropriate adjustment and constraint compensation should be implemented.

[0082] When P > 40 kPa, it is considered high risk; therefore, the adjustment rate is forcibly reduced and the opening offset is corrected.

[0083] Furthermore, the adaptive suppression module for water hammer risk outputs a specific compensation correction coefficient μ (i.e., the water hammer suppression compensation coefficient, dimensionless) based on different risk levels. The segmented value rules are as follows:

[0084] ;

[0085] It should be noted that after the water hammer risk adaptive suppression module completes its calculation, it transmits the water hammer suppression compensation coefficient μ to the dynamic opening precision adjustment module for final safety correction of the initial adjustment rate. This aims to suppress water hammer impact as much as possible while ensuring that the flow adjustment target remains unchanged, thus balancing adjustment accuracy and pipeline operation safety. At the same time, it uploads the water hammer risk level, water hammer pressure value, compensation data, etc., to the full-process control and scheduling module for global monitoring and anomaly alarms.

[0086] The working principle of this invention is as follows: During use, operating parameters such as gate valve water level, flow rate, and opening degree are collected in multiple dimensions. After adaptive noise reduction, anomaly repair, and data standardization, an improved deep learning model is used to predict the natural flow without intervention. Based on the flow deviation classification, the optimal opening degree and graded adjustment rate are calculated. A feedforward prediction mode is used to achieve precise flow regulation. In addition, the water hammer pressure is calculated in real time, risk levels are classified, and compensation coefficients are output to correct the adjustment rate. This solves the problems of traditional hydraulic gate valve regulation lag, flow overshoot, and lack of water hammer protection, significantly improves the accuracy and response speed of flow regulation, effectively suppresses water hammer impact, protects the safety of pipelines and gate valve equipment, adapts to various complex hydraulic working conditions, significantly reduces operation and maintenance risks and manual intervention costs, and realizes intelligent dynamic control of hydraulic gate valve flow.

[0087] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, enabling those skilled in the art to better understand and utilize it. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A deep learning-based dynamic flow regulation and control system for hydraulic gate valves, characterized in that, It includes a gate valve flow multi-dimensional sensing and acquisition module, an acquisition data adaptive preprocessing module, a deep learning flow prediction and modeling module, a dynamic opening degree precise adjustment module, and a full-process control and scheduling module; The gate valve flow multidimensional sensing and acquisition module is used to collect multidimensional key parameters of gate valve operation. The data acquisition adaptive preprocessing module is used to adaptively denoise, repair outliers, standardize data and normalize dimensions of the raw data, and output a standardized dataset adapted to the deep learning model. The deep learning flow prediction modeling module is based on an improved deep learning model to predict the natural flow under no-intervention conditions in the future preset time period. The dynamic opening precision adjustment module calculates the main adjustment deviation based on the natural flow prediction value and the preset target flow, determines the optimal adjustment opening of the gate valve and the graded adjustment rate, and achieves precise flow adjustment by adopting the feedforward prediction adjustment mode. The full-process management and scheduling module is used to set traffic target values, coordinate the collaborative operation of various modules, monitor the system status in real time, trigger abnormal warnings and global scheduling, and store all system operation data.

2. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 1, characterized in that, The gate valve flow multi-dimensional sensing and acquisition module integrates a submersible static pressure water level sensor, an ultrasonic Doppler flow sensor, a high-precision angle sensor, a differential pressure sensor, and an environmental compensation sensor.

3. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 1, characterized in that, The adaptive preprocessing module for the acquired data adopts an improved wavelet threshold denoising algorithm for adaptive denoising. It identifies outliers through a 3σ confidence interval and repairs them using a quadratic interpolation method. The data is then converted to the [0,1] interval to complete standardization, and core features are extracted to complete dimensional regularization.

4. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 1, characterized in that, The deep learning traffic prediction modeling module divides the dataset into training, validation and test sets. The model training uses an adaptive learning rate optimization algorithm and predicts the natural traffic flow for a preset future time period without intervention.

5. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 1, characterized in that, The dynamic opening precision adjustment module uses the difference between the preset target flow and the predicted natural flow as the sole main adjustment deviation, and classifies the deviation into three levels: small, medium, and large.

6. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 5, characterized in that, The dynamic opening precision adjustment module combines the real-time opening of the gate valve, the water flow impact correction coefficient, and the water flow disturbance coefficient to calculate the optimal target opening of the gate valve, and corrects the adjustment accuracy in real time based on the opening feedback from the actuator.

7. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 6, characterized in that, The dynamic opening precision adjustment module introduces a water hammer suppression compensation coefficient to correct the graded adjustment rate, matching the valve's mechanical response with the pipeline's water hammer protection requirements.

8. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 7, characterized in that, It also includes a water hammer risk adaptive suppression module, which constructs a water hammer pressure calculation model based on pipeline transient flow theory, and calculates the instantaneous water hammer pressure in the pipeline in real time by combining the opening change and the pressure difference between the inlet and outlet of the gate valve.

9. The deep learning-based hydraulic gate valve flow dynamic regulation and control system according to claim 8, characterized in that, The water hammer risk adaptive suppression module is divided into three risk levels: low, medium, and high, based on the instantaneous water hammer pressure, and outputs the corresponding water hammer suppression compensation coefficient for each level.

Images