A method and system for predicting water inflow of a large longitudinal slope water-rich tunnel during operation period

Abstract

The present application relates to a kind of big longitudinal slope water-rich tunnel operation period water inflow prediction method and system using, belong to hydrogeology field, prediction method includes: after the data collected is preprocessed, according to the distribution state of tunnel surrounding rock and monitoring data, using Goodmann formula, in combination with geological parameters and the structure characteristics of each layer, establish layered water inflow physical calculation model;Multiple source measured data is used as network input, and Goodmann theory solution is used as physical constraint term, and the PINN model established is trained;Using the PINN model optimized after training, the geological, hydrological and water inflow data of each monitoring point along the tunnel are intelligently calculated in real time, and under different working conditions and multiple time periods, the dynamic prediction value of the water inflow of each monitoring point is generated.The present application gives consideration to the double target of tunnel drainage safety and energy utilization, makes full use of the natural elevation of tunnel longitudinal slope, without additional excavation of high-cost water diversion slope shaft, with good economy and generalizability.

Description

Technical Field

[0001] This invention relates to the field of hydrogeology, and in particular to a method and system for predicting and utilizing water inflow during the operation of a water-rich tunnel with a steep longitudinal slope. Background Technology

[0002] Tunnel engineering plays a vital role in transportation infrastructure construction and regional economic development. As highway, railway, and urban rail transit construction expands into mountainous, hilly, and complex hydrogeological environments, the proportion of tunnels traversing water-rich strata is increasing. Currently, the prediction of tunnel water inflow largely relies on simplified empirical formulas or idealized numerical simulation methods, resulting in significant deviations in predictions. Inaccurate predictions directly impact the design and layout of drainage systems—the selection of drainage equipment and pipeline capacity often fails to meet actual needs due to overestimation or underestimation, potentially leading to resource waste and increased risk of water inflow disasters. Simultaneously, most existing tunnel drainage systems primarily rely on simple discharge, failing to develop and utilize the potential energy of water in high-drop, water-rich tunnels, resulting in substantial energy loss. This not only reduces the economic efficiency of tunnel operation but also contradicts the direction of green and low-carbon development.

[0003] Traditional tunnel water inrush prediction mainly relies on numerical simulation and physical models. However, these methods are limited by incomplete geological data, high parameter uncertainty, and poor model adaptability, failing to meet the dynamic monitoring and real-time early warning needs under complex geological conditions and variable operating environments. Furthermore, existing tunnel drainage systems mostly rely on discharge, failing to effectively develop and utilize the hydroelectric energy resources inherent in the inrush water, resulting in energy waste and environmental burden. Especially in water-rich tunnels with steep longitudinal slopes, the water flow possesses high potential energy. If this energy could be converted into electricity on-site, it would provide green energy support for tunnel operation, achieving a dual improvement in energy conservation, emission reduction, and economic benefits. Therefore, addressing these issues is currently a crucial consideration. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for predicting and utilizing water inflow during the operation of a water-rich tunnel with a large longitudinal slope, thus solving the deficiencies of the prior art.

[0005] The objective of this invention is achieved through the following technical solution: a method for predicting water inflow during the operation of a water-rich tunnel with a steep longitudinal slope, the prediction method comprising:

[0006] Step 1: After preprocessing the collected data, based on the distribution of the surrounding rock in the tunnel and the monitoring data, a physical calculation model for the layered water inflow is established using the Goodman formula, combined with geological parameters and the structural characteristics of each layer.

[0007] Step 2: Use multi-source measured data as network input and Goodman's theoretical solution as physical constraint term to train the established PINN model;

[0008] Step 3: Use the trained and optimized PINN model to intelligently calculate the geological, hydrological and water inflow data collected in real time at each monitoring point along the tunnel, and generate dynamic prediction values ​​of water inflow for each monitoring point under different working conditions and multiple time periods.

[0009] The physical calculation model for layered water inflow, established based on the distribution of surrounding rock in the tunnel and monitoring data, utilizes the Goodman formula, combined with geological parameters and the structural characteristics of each layer, and includes:

[0010] Based on the measured geological exploration data along the tunnel during its operation, the tunnel cross-section was divided into layers;

[0011] Based on the initial water head data H obtained from historical monitoring, and drawing upon Goodman's theory and the layered structure, a physical calculation model for the layered water inflow of the tunnel cross-section was established. ,in, Let be the permeability coefficient of the surrounding rock. The permeability coefficient of the grouting ring is... The lining permeability coefficient, For the inner diameter radius of the lining, For the outer radius of the lining, The outer radius of the grouting ring, The radius of the outer boundary of the surrounding rock;

[0012] By using a physical calculation model of stratified inflow volume combined with parameters from multiple monitoring points, batch calculations of the theoretical solutions for stratified inflow volume at each monitoring point and in each section were completed.

[0013] Step two specifically includes the following:

[0014] A1. Organize and aggregate multi-source measured data as the input dataset for the network. The multi-source measured data includes geological structure parameters, environmental parameters, historical and real-time water inflow at each monitoring point;

[0015] A2. Design the PINN structure, and determine the number of neurons and activation functions in the input layer, hidden layer, several fully connected layers, and output layer.

[0016] A3. The analytical solution of the layered inflow calculation model established based on the Goodman formula is used as the physical constraint term of the PINN model to ensure that the prediction results of the PINN model fit the monitoring data and conform to the tunnel hydraulic-geophysical mechanism.

[0017] A4. The total loss function during the training of the PINN model is set as a weighted combination of data fitting loss and physical constraint loss. The data fitting loss measures the error between the output of the PINN model and the actual monitored inflow, while the physical constraint loss measures the deviation between the output of the PINN model and the theoretical solution of the Goodman formula, thus realizing the organic integration of mechanism-driven and data-driven approaches.

[0018] A5. Iteratively train the network using multi-source measured data, minimize the total loss function through the backpropagation algorithm to improve the model's prediction accuracy, introduce cross-validation during training to prevent overfitting, and evaluate the training effect through the hold-out method to ensure the model's generalization ability.

[0019] A6. Based on the validation results, adjust the network architecture, hyperparameters, and loss weights, and repeatedly optimize the model performance.

[0020] The prediction method also includes data acquisition and preprocessing steps, which specifically include the following:

[0021] B1. Based on the geological construction and operation risk zoning along the tunnel, set up water inflow monitoring points, groundwater level monitoring wells, seepage point observation points and meteorological data collection points, and clarify the collection objects, collection frequency, sensor type and data transmission path of each type of data to ensure the spatiotemporal continuity and coverage representativeness of the monitoring data.

[0022] B2. During the tunnel construction period, geological drilling equipment was used to obtain soil and rock samples from various monitoring points for stratigraphic stratification and depth analysis; ground-penetrating radar was used for rapid non-destructive exploration of the tunnel and surrounding underground structures; seismic detectors and seismographs were used to assist in determining faults and the integrity of rock mass structures; and underground 3D scanners were used to construct a 3D model of the strata surrounding the tunnel to supplement spatial distribution information; the collected data covered surrounding rock type, joints and fissures, surrounding rock integrity, lining thickness, and grouting ring parameters.

[0023] B3. During the tunnel operation period, an automatic groundwater level monitoring instrument is used to collect groundwater head height data at each monitoring point in real time in monitoring wells deployed along and around the tunnel; a permeability coefficient tester is used to conduct in-situ permeability tests on the surrounding rock and aquifer in different sections to obtain the permeability coefficient changes in each section; combined with a GPRS / IoT remote transmission module, automatic collection and remote transmission of groundwater level and permeability data are realized.

[0024] B4. During the tunnel operation period, automatic flow meters are used to continuously monitor the water inflow at each monitoring point along the tunnel to obtain real-time water inflow data for each section; seepage point observation devices are used to accurately record the location and flow changes of each seepage point in the tunnel; and data acquisition terminals and IoT remote transmission modules are used to realize online automatic acquisition and remote transmission of water inflow and seepage data.

[0025] B5. During the tunnel operation period, the aggregated water inflow, groundwater level, and permeability coefficient are managed in a unified manner through an automated processing system. Extreme values ​​are automatically removed using anomaly detection tools, and missing data is supplemented by interpolation or estimation using historical data. All data is converted to a unified dimension and standard format, normalized, and stored in the database to ensure information security and reliability.

[0026] A hydropower resource utilization system based on a method for predicting water inflow during the operation of a water-rich tunnel with a steep longitudinal slope, the system comprising an engineering condition confirmation and data input device, a water diversion and collection system deployment device, a bypass and safe discharge system setup device, and a monitoring and feedback device.

[0027] The engineering conditions and data input device is used to receive the water inflow prediction results of different tunnel sections, convert the prediction results into input parameters that can be used for system control, establish a water use condition database, and determine safety boundary conditions.

[0028] The water collection system deployment device is used to be laid along the longitudinal slope of the tunnel, to collect the water gushing from each section and guide it into the main water collection pipe, to reduce the fluctuation of the water flow and regulate the water storage, to convert the potential energy of the water gushing into electrical energy, and to supply the generated electrical energy to the tunnel ventilation, lighting and monitoring facilities.

[0029] The bypass and safe discharge system is equipped with a device for directly discharging the gushing water into the tunnel when the unit is under maintenance or the water flow exceeds the power generation capacity.

[0030] The monitoring and feedback device is used to collect water flow, water level and power generation parameters in real time, and adjust the operation status of the water diversion and collection system and the bypass and safe discharge system based on the water inrush prediction results.

[0031] The water collection system deployment module specifically includes the following:

[0032] Collect information on the tunnel's longitudinal slope alignment, key elevations, existing drainage structures within the tunnel, and the location of equipment rooms.

[0033] Input the obtained water inflow prediction results and water catchment locations from different cross sections into the data module to establish a cross section-time water inflow condition database;

[0034] Target power / load curves are generated based on the load type during the operation period, serving as a benchmark for subsequent operation and scheduling;

[0035] Determine safety boundary conditions: allowable water level range inside the tunnel, maximum allowable back pressure, minimum stable flow conditions, tunnel structure waterproofing level, and operation and maintenance windows.

[0036] The water collection system deployment device includes a water collection system, a pressure stabilization and energy storage unit, and an energy conversion unit;

[0037] The water collection system is distributed along the longitudinal slope of the tunnel to collect the gushing water generated at each section and guide it into the main water collection pipe.

[0038] The pressure stabilization and energy storage unit is connected to the main water supply pipe and is used to reduce fluctuations in the inflow rate and regulate the water storage capacity.

[0039] The energy conversion unit is located downstream of the tunnel's longitudinal slope and is used to convert the potential energy of the gushing water into electrical energy, which is then used to supply the tunnel's ventilation, lighting, and monitoring facilities.

[0040] The bypass and safe emission system device specifically includes the following:

[0041] A bypass pipe and a quick-switching valve are installed before the generator set to automatically switch to bypass direct discharge when the unit fails, is under maintenance, or the flow rate is lower than the minimum stable value.

[0042] The tailwater from the generator set is incorporated into the tailwater channel or the existing drainage system to ensure that the principle of prioritizing drainage safety inside the tunnel is always met.

[0043] Equipped with manual or electric dual-circuit control and mechanical limit switches to ensure reliable switching in emergency situations.

[0044] The monitoring and feedback device specifically includes a monitoring point deployment unit and an operation scheduling and control strategy unit:

[0045] The monitoring point layout unit includes: flow meters and valve position feedback points in each branch, pressure / flow monitoring points in the main pipe, and liquid level and overflow monitoring points in the pressure stabilizing unit; pressure and water hammer event recording in the pressure pipeline, and speed, power, temperature rise and vibration monitoring points in the generator set; the monitoring signals are connected to the controller and linked with the data and scheduling module to achieve remote or local two-level control.

[0046] The operation scheduling and control strategy unit:

[0047] Power Point Tracking Mode: When it is necessary to track the target load during the operation period, the target inflow rate is calculated based on the real-time head and limited, and closed-loop regulation is achieved through guide vanes or valves;

[0048] Water level stabilization mode: When drainage safety is prioritized, the control objective is to maintain the overflow water level. exist Between, inflow rate and bypass rate Based on the water level deviation, hS,min is the minimum overflow water level, and hS,max is the maximum overflow water level;

[0049] Branch selection: Determine branch priority based on available head and loss, and give priority to branches with high head and low loss;

[0050] Safety interlock: When any monitored quantity exceeds the limit, the system triggers the corresponding interlock, including load reduction, bypass opening, shutdown, and overflow discharge.

[0051] This invention offers the following advantages: a method and system for predicting and utilizing water inflow during the operation of a water-rich tunnel with a steep longitudinal slope. This system balances the dual objectives of tunnel drainage safety and energy utilization, fully utilizing the natural elevation differences of the tunnel's longitudinal slope without requiring the excavation of costly water diversion shafts, thus exhibiting good economic efficiency and scalability. Based on known water inflow prediction results, the system can plan water energy utilization schemes in advance during the tunnel design and construction phases. During operation, it dynamically adjusts operating parameters using monitoring and control units, effectively addressing the spatial distribution differences and temporal fluctuations of water inflow at different cross-sections, and achieving continuous and stable power output. The generated electricity directly powers ventilation, lighting, monitoring, and safety facilities during tunnel operation, significantly improving the tunnel's energy self-sufficiency and sustainable operation capabilities. Simultaneously, it reduces energy waste on the surrounding environment during tunnel drainage, demonstrating a comprehensive advantage in safety, economy, and environmental protection. Attached Figure Description

[0052] Figure 1 This is a schematic flowchart of the method of the present invention;

[0053] Figure 2 This is a schematic diagram of the PINN model principle;

[0054] Figure 3 This is a schematic diagram of the overall system layout;

[0055] Figure 4 A schematic diagram of the hydraulic system and head;

[0056] Figure 5 For monitoring-control-scheduling block diagram;

[0057] Figure 6 This is a schematic diagram of energy conversion and bypass safety. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application provided below with reference to the accompanying drawings is not intended to limit the scope of protection of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. The present invention will be further described below with reference to the accompanying drawings.

[0059] One embodiment of this invention relates to an intelligent prediction method for water inflow during the operation of a water-rich tunnel with a steep longitudinal slope, combining the Goodman formula with the PINN (Physical Neural Network) model. This method effectively addresses the shortcomings of existing water inflow prediction technologies in terms of accuracy, real-time performance, and dynamic modeling in response to complex hydrological and meteorological conditions, thereby improving the level of intelligent assessment of water inflow changes during tunnel operation.

[0060] like Figure 1 and Figure 2 As shown, it specifically includes the following:

[0061] S1: Determine the study area along the tunnel and collect data to obtain raw data related to the prediction of water inflow in water-rich tunnels on steep longitudinal slopes. This data includes geological data, hydrogeological parameters, tunnel water inflow monitoring data, and tunnel structural parameters. Specifically, this includes: geological data such as regional stratigraphic distribution, lithological type, joint and fracture characteristics, surrounding rock type, lining thickness, and grouting ring characteristics; hydrogeological parameters such as groundwater head height, permeability coefficient, and distribution of aquifers and impermeable layers; and tunnel operation and maintenance data such as water inflow, leakage point distribution, and flow rate variation curves at various monitoring points along the tunnel. This includes the following:

[0062] S101: Based on the geological construction and operational risk zoning along the tunnel route, rationally deploy water inflow monitoring points, groundwater level monitoring wells, seepage point observation points, and meteorological data collection points. Clearly define the collection targets, collection frequency, sensor type, and data transmission path for each type of data to ensure the spatiotemporal continuity and representative coverage of the monitoring data.

[0063] S102: During the tunnel construction phase, geological drilling equipment was used to obtain soil and rock samples from various monitoring points for stratigraphic stratification and depth analysis; ground-penetrating radar was used for rapid, non-destructive exploration of the tunnel and surrounding underground structures; seismic detectors and seismographs were used to assist in determining faults and the integrity of the rock mass structure; and a subsurface 3D scanner was used to construct a 3D model of the strata surrounding the tunnel, supplementing spatial distribution information. The collected data covered surrounding rock type, joints and fissures, surrounding rock integrity, lining thickness, and grouting ring parameters.

[0064] S103: During tunnel operation, automatic groundwater level monitoring instruments are used in monitoring wells deployed along and around the tunnel to collect real-time groundwater head height data at each monitoring point. Permeability coefficient testers are used to conduct in-situ permeability tests on the surrounding rock and aquifers in different sections to determine permeability coefficient variations. Combined with a GPRS / IoT remote transmission module, automatic collection and remote transmission of groundwater level and permeability data are achieved. The collected data covers groundwater head height, permeability coefficient, etc., providing fundamental data support for tunnel water inflow prediction models and hydrodynamic analysis.

[0065] S104: During tunnel operation, automatic flow meters are used to continuously monitor the water inflow at various monitoring points along the tunnel, acquiring real-time water flow data for each section. Leakage point observation devices are used to accurately record the location and flow changes of each leakage point within the tunnel. Data acquisition terminals and IoT remote transmission modules are applied to achieve online automatic acquisition and remote transmission of water inflow and leakage data. The collected data includes water inflow at each monitoring point, leakage point distribution, and flow change curves, providing dynamic data support for assessing tunnel water hazard risks and intelligently predicting water inflow.

[0066] S105: During tunnel operation, an automated processing system is used to uniformly manage the collected raw monitoring data, including water inflow, groundwater level, and permeability coefficient. Anomaly detection tools automatically remove extreme values, and missing data is supplemented using interpolation or historical data estimation. All data is converted to a unified dimension and standard format, normalized, and stored in a database to ensure information security and reliability, providing high-quality data support for the intelligent water inflow prediction model.

[0067] S2: Taking a water-rich tunnel with a steep longitudinal slope as an example, a physical calculation model for layered water inflow was constructed. First, based on the measured geological exploration data along the tunnel route during its operation, the tunnel cross-section was divided into layers. Taking a typical tunnel cross-section as an example, the burial depth is 40m, the tunnel inner diameter radius is 5m, the lining thickness is 0.6m, and the grouting ring thickness is 2m. The structural parameters and hydrogeological parameters of each layer obtained through measured and design data are shown in Table 1.

[0068] Table 1. Structural parameters and hydrogeological parameters of each layer

[0069]

[0070] Based on historical monitoring data, the initial water head H is 35m. Referring to the Goodman formula and combining it with the layered structure, a physical calculation model for the layered water inflow of the tunnel cross-section is established:

[0071] ,

[0072] in, Let be the permeability coefficient of the surrounding rock. The permeability coefficient of the grouting ring is... The lining permeability coefficient, For the inner diameter radius of the lining, For the outer radius of the lining, The outer radius of the grouting ring, The radius of the outer boundary of the surrounding rock is denoted as .

[0073] Substituting the parameters, the inflow rate is calculated as follows:

[0074] ,

[0075] The calculated water inflow at this cross-section is 1.82 × 10⁻⁵ m³ / s.

[0076] Using the calculation formulas of the aforementioned physical calculation model for stratified inflow, and combining parameters from multiple monitoring points, batch calculations of the theoretical solutions for stratified inflow at each monitoring point and in each section were completed. Based on this, an accurate theoretical foundation is provided for the subsequent construction of physical constraint terms and model training in the PINN model, realizing the coupling of engineering mechanisms between structural parameters, geological factors, and inflow.

[0077] S3: Intelligent prediction of tunnel water inflow using PINN. This method uses multi-source measured data along the tunnel as input to PINN, fits the water inflow pattern using a deep network, and introduces layered Goodman theory solutions as physical constraints to ensure that the model satisfies the physical mechanism of tunnel water inflow while maintaining data fitting capability. Specifically, it includes the following:

[0078] S301: Model Input Data Preparation. Combining the data preprocessing and layered physical modeling from the previous stage, multi-source measured data during tunnel operation, such as geological structure parameters, burial depth, hydrogeological parameters, historical and real-time water inflow at each monitoring point, are integrated into the network input dataset. Normalization is then applied to unify the numerical range of all input parameters, improving the stability and convergence speed of model training.

[0079] S302: Build the PINN model, which includes an input layer containing geological structure parameters (such as permeability coefficient, thickness, and burial depth of each layer), environmental parameters (hydraulic head, monitoring point coordinates, etc.), and historical inflow. Hidden layers consist of several fully connected layers, using ReLU or Tanh as activation functions; the number of layers and neurons is determined through cross-validation. The output layer contains the predicted inflow value for the corresponding monitoring point and operating conditions. Hyperparameter settings include learning rate and regularization coefficient, which are iteratively optimized based on the validation set performance.

[0080] S303: Physical constraints and boundary conditions are introduced, and the physical constraint term of the network is combined with the layered inflow physical calculation model. This includes physical constraints, requiring the inflow output of the network to be consistent with the analytical solution of the layered inflow physical calculation model, and using the residual of the analytical solution to constrain the network prediction results; boundary conditions are set, and for monitoring intervals or tunnel endpoints, the inflow must meet the actual boundary; for points with special geological structures, the boundary parameters need to be dynamically adjusted; spatial distribution information is embedded, and the coordinates of monitoring points are used as input features, enabling the network to characterize the differences and distribution of physical inflow at different spatial points.

[0081] S304: Define the total loss function for training the PINN model, which includes data-driven terms and physical constraint terms:

[0082] Data fitting loss: measures the mean square error (MSE) between the network output and the actual monitored water flow.

[0083] Physical constraint loss: measures the error between the network output and the hierarchical Goodman theory solution;

[0084] Total loss function: a weighted combination of two parts, in the form of:

[0085] ,

[0086] in, For the total loss function, For data loss items, For data loss weighting coefficients, For physical constraint loss terms, This represents the weighting coefficient for physical constraint loss.

[0087] S305: Network Training and Validation. Using prepared sample data, the PINN model is iteratively trained using gradient descent and backpropagation algorithms to minimize the total loss function. Cross-validation is introduced during training, and some monitoring data is reserved for evaluating the model's generalization ability to prevent overfitting.

[0088] S306: Model optimization and iteration. By combining historical data with newly acquired real-time data, the model prediction error is continuously monitored, and the network structure and loss weights are dynamically adjusted based on error feedback. For water inrush behavior under special working conditions or extreme weather conditions, the physical boundary parameters are updated in a timely manner to ensure model adaptability and engineering reliability.

[0089] S4: Intelligent Prediction and Dynamic Response. Utilizing the previously trained and optimized PINN model, this system intelligently calculates geological, hydrological, and water inflow data collected in real-time from monitoring points along the tunnel. Through automatic model processing and input analysis, it can quickly generate dynamic prediction results for water inflow at each monitoring point under different operating conditions and at multiple time periods.

[0090] S5: Model Validation and Improvement. The predicted output of the PINN model is systematically compared with historical observed water inflow data along the tunnel route. Error analysis methods are used to quantify the model's predictive performance. The feasibility and accuracy of the model are analyzed in conjunction with actual engineering conditions. Based on the error distribution and prediction deviations under extreme conditions, model parameters and physical constraint weights are iteratively optimized to continuously improve prediction effectiveness.

[0091] S6: Model Application. The validated and selected PINN model will be integrated into the tunnel operation and management decision-making platform to enable real-time intelligent prediction of water inrush during tunnel operation. The model can be linked with the on-site monitoring system to automatically receive real-time data, enabling dynamic early warning of water inrush risks, automatic formulation of response strategies, and resource allocation.

[0092] like Figures 3-6As shown, another embodiment of the present invention relates to a hydropower resource utilization system based on the prediction results of water inflow in a water-rich tunnel with a large longitudinal slope, including an engineering condition confirmation and data input device, a water diversion and collection system layout device, a bypass and safe discharge system setting device, and a monitoring and feedback device.

[0093] Engineering conditions and data input device: used to receive water inflow prediction results from different sections of the tunnel, convert the prediction results into input parameters that can be used for system control, establish a water use condition database, and determine safety boundary conditions;

[0094] Water collection system deployment device: used to be deployed along the longitudinal slope of the tunnel, used to collect the water gushing from each section and guide it into the main water collection pipe, used to reduce the fluctuation of water flow and regulate the water storage, used to convert the potential energy of the water gushing into electrical energy, and use the generated electrical energy to supply the tunnel ventilation, lighting and monitoring facilities.

[0095] Bypass and safety discharge system device: used to directly discharge the gushing water into the tunnel when the unit is under maintenance or the water flow exceeds the power generation capacity;

[0096] Monitoring and feedback device: used to collect water flow, water level and power generation parameters in real time, and adjust the operation status of water diversion and collection system and bypass and safe discharge system based on the water inrush prediction results.

[0097] Furthermore, the water collection system deployment device includes a water collection system, a pressure stabilization and energy storage unit, and an energy conversion unit.

[0098] Water collection system: Distributed along the longitudinal slope of the tunnel, used to collect the water gushing from each section and guide it into the main water collection pipe; Pressure stabilization and energy storage unit: Connected to the main water collection pipe, used to reduce fluctuations in water flow and regulate water storage; Energy conversion unit: Located downstream of the tunnel longitudinal slope, used to convert the potential energy of the water gushing into electrical energy, and to supply the generated electrical energy to the tunnel ventilation, lighting and monitoring facilities.

[0099] Furthermore, the engineering condition verification and data input device specifically includes the following:

[0100] S101: Collect the longitudinal slope alignment of the tunnel, key elevations, and the location of existing drainage structures and equipment rooms inside the tunnel.

[0101] S102: The obtained water inrush prediction results from different cross-sections... The water catchment location is input into the data module to establish a cross-section-time water inflow condition database.

[0102] S103: Generate target power / load curves based on the load type during the operation period (lighting, ventilation, monitoring, etc.). This serves as the benchmark for subsequent operation and scheduling.

[0103] S104: Determine safety boundary conditions: allowable water level range inside the tunnel, maximum allowable back pressure, minimum stable flow conditions, waterproofing level of the tunnel structure, and operation and maintenance windows, etc.

[0104] Furthermore, the deployment of the water diversion and collection system specifically achieves the following:

[0105] S201: Set up collection wells / troughs in the predicted concentrated water inflow area to collect scattered water inflow on-site; the top / surroundings of the collection wells (troughs) are connected to the lining drainage layer and are equipped with inspection holes.

[0106] S202: Branch pipes are drawn from each water collection point and merge into the main water collection pipe laid along the longitudinal slope. One-way valves and electric regulating valves are installed on the branch pipes. A rectifier section is set at the junction with the main pipe to reduce local losses.

[0107] S203: Main pipe (laid along the longitudinal slope, with necessary vent valves and maintenance valves installed to ensure stable gravity flow and facilitate venting and maintenance).

[0108] S204: Determine the priority of each branch: Calculate the equivalent available head for each branch and establish a start-stop strategy.

[0109] S205: Complete the civil engineering and waterproofing of the water collection well / trough, and set up the grid and sedimentation structure according to the standard to facilitate the settling and cleaning of impurities.

[0110] S206: After the branch pipes and main pipes are installed in place, water pressure and leakage tests are conducted; valve linkage tests are performed to ensure flexible opening and closing.

[0111] S207: Controllable connection nodes are set up with existing transverse drainage ditches and central drainage pipelines to achieve a smooth functional transition from the construction period to the operation period.

[0112] Furthermore, the voltage stabilization and energy storage unit specifically includes the following:

[0113] S301: Install surge tanks / pressure regulating wells and stabilizing tanks on the downstream side of the longitudinal slope, and configure overflow / bypass structures and overflow weirs to limit the maximum operating water level.

[0114] S302: The voltage stabilizing volume is checked using the mass conservation principle, and the following conditions are met during operation:

[0115] ,

[0116] Among them, V S (t) represents the flow rate of the voltage stabilizing unit. For the i-th branch, For the set of branches that have been opened, Inbound traffic This refers to overflow / bypass flow.

[0117] S303: Install a liquid level sensor and overflow alarm; convert the volume-level function... Incorporate into control logic and limit h S (t) represents the overflow water level, h S，min For the minimum overflow level, h S，max This is the maximum overflow water level.

[0118] Furthermore, the energy conversion unit is specifically configured to include the following:

[0119] S401: A pressure pipeline is installed from the self-stabilizing unit to the unit's water inlet, with a pipeline length of... Pipe inner diameter It is equipped with an inlet valve and waterproof hammer components (such as a slow-closing valve / air valve).

[0120] S402: Downstream equipment room is equipped with turbines and generator sets (units), the type of which is selected based on the available head and flow rate (impact type / reaction type), including guide vane / nozzle adjustment, generator and control cabinet.

[0121] S403: Head loss check along the line and in local areas:

[0122] ,

[0123] ,

[0124] in, h f This refers to head loss, i.e. Figure 4 Pipe loss in the middle, λ This is the pipe resistance coefficient. v p The flow velocity inside the pipe. g It is the acceleration due to gravity. q p For the main flow rate of the pipeline, A p The cross-sectional area of ​​the pipe. It is the sum of all local drag coefficients. This represents the local resistance coefficient of the k-th local resistance component.

[0125] S404: Relationship between available water head and power:

[0126] ,

[0127] ,

[0128] in, For the overall efficiency of the unit and electrical system, Indicates the water level in the collection well. H(t) represents the turbine inlet water level, H(t) is the available net head at time t, and P(t) is the system output power at time t.

[0129] S405: Perform net positive suction head (NPSH) and minimum stable flow rate verification to ensure that the unit operates within the available operating conditions.

[0130] Furthermore, the bypass and safe emission system installation device specifically includes the following:

[0131] S501: A bypass pipe and a quick-switching valve are installed before the generator set to automatically switch to bypass direct discharge when the unit fails, is under maintenance, or the flow rate is lower than the minimum stable value.

[0132] S502: The tailwater of the generator set is connected to the tailwater channel / existing drainage system to ensure that the principle of prioritizing drainage safety inside the tunnel can be met at any time.

[0133] S503: Equipped with manual / electric dual-circuit control and mechanical limit switches to ensure reliable switching in emergency situations.

[0134] Furthermore, the monitoring and feedback device specifically includes a monitoring point layout unit and an operation scheduling and control strategy unit; wherein the monitoring point layout unit specifically includes the following:

[0135] S601: Set flow meters and valve position feedback points in each branch, set pressure / flow monitoring points in the main pipe, and set liquid level and overflow monitoring points in the pressure stabilizing unit.

[0136] S602: Set up pressure and water hammer event records in pressure pipelines, and set up speed, power, temperature rise and vibration monitoring points in generator sets.

[0137] S603: Monitoring signal access controller (PLC / edge computing), linked with the data and scheduling module to achieve remote / local two-level control.

[0138] Furthermore, the motion scheduling and control strategy unit specifically includes the following:

[0139] S701: Power Point Tracking Mode: When it is necessary to track the target load during the operating period. At that time, based on the real-time available water head Calculate the target inbound flow rate as follows: Closed-loop regulation is achieved through guide vanes / valve;

[0140] Where clip() is the qualified function, Q min As the lower limit of traffic, Q max This is the maximum traffic limit.

[0141] S702: Water level stabilization mode: When drainage safety is prioritized, the control objective is to maintain the overflow water level. exist Between; inflow rate and bypass rate Distributed according to water level deviation.

[0142] S703: Branch Selection: Determine branch priority based on available head and losses, prioritizing branches with high head and low losses; the equivalent available head of the i-th branch can be expressed as:

[0143]

[0144] Strategy by Determined by weighted sorting based on real-time available traffic , where z i For the upstream elevation of branch road i, z s Elevation of system nodes (main water tank, main canal, etc.) ζ v is the component drag coefficient. i Let be the flow velocity of branch i. In the branch selection strategy, this refers to the set of available branches determined in real time based on weighted priority.

[0145] S704: Safety Interlock: When any monitored quantity exceeds the limit (such as overpressure, overspeed, low liquid level, high liquid level), the system triggers the corresponding interlock: load reduction, bypass opening, shutdown, overflow discharge, etc.

[0146] This system also includes two parts: integrated debugging and operation phase access and metering. The integrated debugging includes the following:

[0147] S801: No-load and water-load commissioning: Check valve operation, signal on / off, bypass switching and alarm function in sequence.

[0148] S802: Gradual increase in operating condition test run: Gradually open the branch circuit, increase the inlet flow rate, record the water level-power-pressure curve, and verify the consistency between the model and the actual measurement.

[0149] S803: Operating Condition Switching Test: Switching between power tracking and water level stabilization modes to verify water hammer suppression and power smoothing during the transition process.

[0150] Furthermore, the operational phase access and metering specifically include the following:

[0151] S901: The power generation output is connected to the tunnel operation power grid through the parallel / off-grid device, giving priority to basic loads such as lighting, ventilation, monitoring and safety facilities.

[0152] S902: Configure electricity metering and operation reports, and combine them with the monitoring-control system to form energy efficiency assessment and maintenance decision data.

[0153] S903: Develop maintenance and repair plans, including valve opening and closing cycles, sensor calibration, turbine and generator bearing / seal inspection, and drainage system dredging.

[0154] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and improvements, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for predicting water inflow during the operation of a water-rich tunnel with a steep longitudinal slope, characterized in that: The prediction method includes: Step 1: After preprocessing the collected data, based on the distribution of the surrounding rock in the tunnel and the monitoring data, a physical calculation model for the layered water inflow is established using the Goodman formula, combined with geological parameters and the structural characteristics of each layer. Step 2: Use multi-source measured data as network input and Goodman's theoretical solution as physical constraint term to train the established PINN model; Step 3: Use the trained and optimized PINN model to intelligently calculate the geological, hydrological and water inflow data collected in real time at each monitoring point along the tunnel, and generate dynamic predicted values ​​of water inflow at each monitoring point under different working conditions and multiple time periods. The physical calculation model for layered water inflow, established based on the distribution of surrounding rock in the tunnel and monitoring data, using the Goodman formula and combining geological parameters and the structural characteristics of each layer, includes: Based on the measured geological exploration data along the tunnel during its operation, the tunnel cross-section was divided into layers. Based on the initial water head data H obtained from historical monitoring, and drawing upon Goodman's theory and the layered structure, a physical calculation model for the layered water inflow of the tunnel cross-section was established. ,in, Let be the permeability coefficient of the surrounding rock. The permeability coefficient of the grouting ring is... The lining permeability coefficient, For the inner diameter radius of the lining, For the outer radius of the lining, The outer radius of the grouting ring, The radius of the outer boundary of the surrounding rock; By using a physical calculation model of stratified inflow rate combined with parameters from multiple monitoring points, batch calculations of the theoretical solutions for stratified inflow rate at each monitoring point and in each section were completed. Step two specifically includes the following: A1. Organize and aggregate multi-source measured data as the input dataset for the network. The multi-source measured data includes geological structure parameters, environmental parameters, historical and real-time water inflow at each monitoring point; A2. Design the PINN structure, and determine the number of neurons and activation functions in the input layer, hidden layer, several fully connected layers, and output layer. A3. The analytical solution of the layered inflow calculation model established based on the Goodman formula is used as the physical constraint term of the PINN model to ensure that the prediction results of the PINN model fit the monitoring data and conform to the tunnel hydraulic-geophysical mechanism. A4. Set the total loss function during the training of the PINN model as a weighted combination of data fitting loss and physical constraint loss. The data fitting loss measures the error between the output of the PINN model and the actual monitored inflow, and the physical constraint loss measures the deviation between the output of the PINN model and the theoretical solution of the Goodman formula, so as to achieve the organic integration of mechanism-driven and data-driven approaches. A5. Iteratively train the network using multi-source measured data, minimize the total loss function through the backpropagation algorithm to improve the model's prediction accuracy, introduce cross-validation during training to prevent overfitting, and evaluate the training effect through the hold-out method to ensure the model's generalization ability. A6. Based on the validation results, adjust the network architecture, hyperparameters, and loss weights, and repeatedly optimize the model performance.

2. The method for predicting water inflow during operation of a water-rich tunnel with a steep longitudinal slope according to claim 1, characterized in that: The prediction method also includes data acquisition and preprocessing steps, which specifically include the following: B1. Based on the geological construction and operation risk zoning along the tunnel, set up water inflow monitoring points, groundwater level monitoring wells, seepage point observation points and meteorological data collection points, and clarify the collection objects, collection frequency, sensor type and data transmission path of each type of data to ensure the spatiotemporal continuity and coverage representativeness of the monitoring data. B2. During the tunnel construction period, geological drilling equipment will be used to obtain soil and rock samples at each monitoring point for stratigraphic stratification and burial depth analysis; ground-penetrating radar will be used to conduct rapid and non-destructive exploration of the tunnel and surrounding underground structures; and seismic detectors and seismographs will be used to assist in determining the integrity of faults and rock mass structures. An underground 3D scanner was used to construct a 3D model of the strata surrounding the tunnel, supplementing spatial distribution information; the collected data covered the surrounding rock type, joints and fissures, surrounding rock integrity, lining thickness, and grouting ring parameters; B3. During the tunnel operation period, an automatic groundwater level monitoring instrument is used to collect groundwater head height data at each monitoring point in real time in monitoring wells deployed along and around the tunnel; a permeability coefficient tester is used to conduct in-situ permeability tests on the surrounding rock and aquifer in different sections to obtain the permeability coefficient changes in each section; combined with a GPRS / IoT remote transmission module, automatic collection and remote transmission of groundwater level and permeability data are realized. B4. During the tunnel operation period, automatic flow meters are used to continuously monitor the water inflow at each monitoring point along the tunnel to obtain real-time water inflow data for each section; seepage point observation devices are used to accurately record the location and flow changes of each seepage point in the tunnel; and data acquisition terminals and IoT remote transmission modules are used to realize online automatic acquisition and remote transmission of water inflow and seepage data. B5. During the tunnel operation period, the aggregated water inflow, groundwater level, and permeability coefficient are managed in a unified manner through an automated processing system. Extreme values ​​are automatically removed using anomaly detection tools, and missing data is supplemented by interpolation or estimation using historical data. All data is converted to a unified dimension and standard format, normalized, and stored in the database to ensure information security and reliability.

3. A hydropower resource utilization system based on a method for predicting water inflow during the operation of a water-rich tunnel with a steep longitudinal slope, characterized in that: The system includes an engineering condition confirmation and data input device, a water diversion and collection system layout device, a bypass and safe discharge system setting device, and a monitoring and feedback device. The engineering conditions and data input device is used to receive the water inflow prediction results of different tunnel sections, convert the prediction results into input parameters that can be used for system control, establish a water use condition database, and determine safety boundary conditions. The water collection system deployment device is used to be laid along the longitudinal slope of the tunnel, to collect the water gushing from each section and guide it into the main water collection pipe, to reduce the fluctuation of the water flow and regulate the water storage, to convert the potential energy of the water gushing into electrical energy, and to supply the generated electrical energy to the tunnel ventilation, lighting and monitoring facilities. The bypass and safe discharge system is equipped with a device for directly discharging the gushing water into the tunnel when the unit is under maintenance or the water flow exceeds the power generation capacity. The monitoring and feedback device is used to collect water flow, water level and power generation parameters in real time, and adjust the operation status of the water diversion and collection system and the bypass and safe discharge system based on the water inrush prediction results.

4. A hydropower resource utilization system based on a method for predicting water inflow during operation of a water-rich tunnel with a steep longitudinal slope, as described in claim 3, is characterized in that: The water collection system deployment module specifically includes the following: Collect information on the tunnel's longitudinal slope alignment, key elevations, existing drainage structures within the tunnel, and the location of equipment rooms. Input the obtained water inflow prediction results and water catchment locations from different cross sections into the data module to establish a cross section-time water inflow condition database; Target power / load curves are generated based on the load type during the operation period, serving as a benchmark for subsequent operation and scheduling; Determine safety boundary conditions: allowable water level range inside the tunnel, maximum allowable back pressure, minimum stable flow conditions, tunnel structure waterproofing level, and operation and maintenance windows.

5. A hydropower resource utilization system based on a method for predicting water inflow during operation of a water-rich tunnel with a steep longitudinal slope, as described in claim 3, is characterized in that: The water collection system deployment device includes a water collection system, a pressure stabilization and energy storage unit, and an energy conversion unit; The water collection system is distributed along the longitudinal slope of the tunnel to collect the gushing water generated at each section and guide it into the main water collection pipe. The pressure stabilization and energy storage unit is connected to the main water supply pipe and is used to reduce fluctuations in the inflow rate and regulate the water storage capacity. The energy conversion unit is located downstream of the tunnel's longitudinal slope and is used to convert the potential energy of the gushing water into electrical energy, which is then used to supply the tunnel's ventilation, lighting, and monitoring facilities.

6. A hydropower resource utilization system based on a method for predicting water inflow during operation of a water-rich tunnel with a steep longitudinal slope, as described in claim 3, is characterized in that: The bypass and safe emission system device specifically includes the following: A bypass pipe and a quick-switching valve are installed before the generator set to automatically switch to bypass direct discharge when the unit fails, is under maintenance, or the flow rate is lower than the minimum stable value. The tailwater from the generator set is incorporated into the tailwater channel or the existing drainage system to ensure that the principle of prioritizing drainage safety inside the tunnel is always met. Equipped with manual or electric dual-circuit control and mechanical limit switches to ensure reliable switching in emergency situations.

7. A hydropower resource utilization system based on a method for predicting water inflow during operation of a water-rich tunnel with a steep longitudinal slope, as described in claim 3, is characterized in that: The monitoring and feedback device specifically includes a monitoring point deployment unit and an operation scheduling and control strategy unit: The monitoring point layout unit includes: flow meters and valve position feedback points in each branch, pressure / flow monitoring points in the main pipe, and liquid level and overflow monitoring points in the pressure stabilizing unit; pressure and water hammer event recording in the pressure pipeline, and speed, power, temperature rise and vibration monitoring points in the generator set; the monitoring signals are connected to the controller and linked with the data and scheduling module to achieve remote or local two-level control. The operation scheduling and control strategy unit: Power Point Tracking Mode: When it is necessary to track the target load during the operation period, the target inflow rate is calculated based on the real-time head and limited, and closed-loop regulation is achieved through guide vanes or valves; Water level stabilization mode: When drainage safety is prioritized, the control objective is to maintain the overflow water level. exist Between, inflow rate and bypass rate Based on the water level deviation, hS,min is the minimum overflow water level, and hS,max is the maximum overflow water level; Branch selection: Determine branch priority based on available head and loss, and give priority to branches with high head and low loss; Safety interlock: When any monitored quantity exceeds the limit, the system triggers the corresponding interlock, including load reduction, bypass opening, shutdown, and overflow discharge.

Images