Intelligent scheduling method and system for tunnel drainage system based on multi-source data fusion

Abstract

The present invention relates to the field of intelligent scheduling of drainage systems, specifically a method and system for intelligent scheduling of tunnel drainage systems based on multi-source data fusion. The method obtains data and performs spatiotemporal alignment to obtain a final rainfall model after spatiotemporal calibration; obtains the tunnel's seepage field and surface runoff field to obtain a nonlinear mapping relationship between rainfall intensity and tunnel runoff; obtains the predicted tunnel runoff through the nonlinear mapping relationship, collects the tunnel drainage system's water collection well level data and the operating output data of the drainage pump group, and generates drainage pump group control reference parameters; and when the difference between the predicted runoff flow and the measured drainage flow of adjacent tunnels is less than a preset difference threshold, outputs a frequency adjustment instruction and a start-stop control sequence based on the drainage pump group control reference parameters. This method improves the spatiotemporal accuracy of rainfall prediction, fully considers the interference effect of runoff flow between tunnels, and improves the precise scheduling of tunnel drainage systems.

Description

Technical Field

[0001] The present invention relates to the field of intelligent scheduling of drainage systems, and in particular to an intelligent scheduling method and system for tunnel drainage systems based on multi-source data fusion. Background Art

[0002] With the acceleration of urbanization and the continuous advancement of transportation infrastructure construction, tunnel projects, as a crucial component of urban underground space development, are attracting increasing attention for the safety and reliability of their drainage systems. Tunnel drainage systems fulfill the crucial functions of removing water accumulation within tunnels, preventing groundwater leakage, and ensuring the safety of tunnel structures. Intelligent scheduling of drainage systems is crucial for ensuring safe tunnel operations, especially in seasons or regions with heavy rainfall.

[0003] The operating conditions of tunnel drainage systems change dynamically with rainfall intensity, tunnel seepage characteristics, and the influence of adjacent tunnel drainage systems. However, existing technologies lack effective dynamic adjustment mechanisms. Traditional methods often rely on relatively single data sources, such as rain gauges or water level sensors within tunnels, and are unable to fully and accurately reflect rainfall conditions around the tunnel or seepage conditions within the tunnel. Especially at the intersection of tunnel clusters, water interference between different drainage systems exists. Existing hydrological models cannot accurately reflect the flow coupling characteristics under dynamic boundary conditions, resulting in inaccurate water volume predictions and irrational drainage scheduling.

[0004] Therefore, there is an urgent need for an intelligent scheduling method for tunnel drainage systems based on multi-source data fusion to realize the intelligent scheduling of drainage pump groups and improve the safety and reliability of tunnel drainage systems. Summary of the Invention

[0005] (1) Technical problems to be solved

[0006] The purpose of the present invention is to provide an intelligent scheduling method and system for tunnel drainage systems based on multi-source data fusion, so as to solve the problem that traditional tunnel drainage system scheduling methods have low drainage prediction accuracy due to a single data source, insufficient spatiotemporal calibration capabilities, failure to consider the nonlinear mapping relationship between rainfall and water collection, and the interference effect of adjacent tunnel drainage, resulting in unreasonable drainage scheduling.

[0007] (2) Technical solution

[0008] To achieve the above objectives, the present invention provides, on the one hand, an intelligent scheduling method for a tunnel drainage system based on multi-source data fusion, the method comprising:

[0009] S1. Obtain meteorological satellite data, ground-based radar data, and seepage pressure data in the tunnel, and perform spatiotemporal alignment through spatiotemporal nested data fusion method and dynamically assign weight coefficients to obtain the final rainfall model after spatiotemporal calibration.

[0010] S2. Obtain the tunnel seepage field and surface runoff field, and obtain the nonlinear mapping relationship between rainfall intensity and tunnel runoff flow through alternating iterative calculation.

[0011] S3. The spatiotemporal distribution data of rainfall intensity of the final rainfall model is used to obtain the tunnel predicted water flow through the nonlinear mapping relationship, the water level data of the collection well of the tunnel drainage system and the operating output data of the drainage pump group are obtained, and the drainage pump group control benchmark parameters are generated; the frequency adjustment instructions and start-stop control sequence of the drainage pump group are output according to the tunnel predicted water flow and the drainage pump group control benchmark parameters.

[0012] Furthermore, the method of performing spatiotemporal alignment by spatiotemporal nested data fusion and dynamically allocating weight coefficients to obtain a spatiotemporally calibrated final rainfall model includes:

[0013] The spatial resolution of the meteorological satellite data was increased to the same grid scale as the ground-based radar data using the bilinear interpolation method, and the timestamps of the ground-based radar data were synchronized to match the sampling frequency of the seepage pressure data. A multidimensional feature matrix was constructed based on the spatiotemporally aligned meteorological satellite data, ground-based radar data, and seepage pressure data. Each grid node in the multidimensional feature matrix contained the rainfall intensity of the meteorological satellite data, the rainfall intensity of the ground-based radar data, the seepage pressure, the spatial coordinates, and the timestamp.

[0014] The weight coefficients of meteorological satellite data, ground-based radar data and seepage pressure data are dynamically calculated, wherein the weight coefficient of meteorological satellite data is determined according to the exponential decay characteristics of its historical error, the weight coefficient of ground-based radar data is calculated according to the spatial variance of rainfall intensity within its grid, and the weight coefficient of seepage pressure data is dynamically adjusted according to its real-time Pearson correlation coefficient with surface runoff data, and the surface runoff data is obtained by inversion of surface runoff section geometric parameters.

[0015] The rainfall intensity of meteorological satellite data, the rainfall intensity of ground-based radar data, and the seepage pressure of each grid node in the multidimensional feature matrix are weighted and calculated with corresponding weight coefficients to generate an initial calibration rainfall model.

[0016] The correlation coefficient between the seepage rate time series data and the rainfall intensity time series data in the preset fan-shaped area in the direction of the tunnel axis is detected in real time. When the correlation coefficient is detected to be lower than the preset correlation coefficient threshold, the weight coefficient of the seepage pressure data is recalculated and the spatial variance of the ground-based radar data is locally corrected. The rainfall intensity of the grid nodes in the intersection area of ​​the tunnel group in the initial calibration rainfall model is iteratively optimized according to the corrected weight coefficient until the difference between the predicted rainfall intensity and the measured rainfall intensity is less than the preset rainfall intensity difference threshold. The final rainfall model after spatiotemporal calibration is then output.

[0017] Furthermore, the method of detecting in real time the correlation coefficient between the seepage rate time series data and the rainfall intensity time series data in a preset sector-shaped area in the tunnel axis direction, and when the correlation coefficient is detected to be lower than a preset correlation coefficient threshold, recalculating the weight coefficient of the seepage pressure data and performing local correction on the spatial variance of the ground-based radar data includes:

[0018] An infiltration rate sensor array is arranged in the preset sector area according to a preset sampling interval and infiltration rate time series data is collected. The infiltration rate time series data and the rainfall intensity time series data of the corresponding sector area in the initial calibration rainfall model are used to calculate the correlation coefficient in the current time window through a sliding time window.

[0019] When the correlation coefficient is lower than the preset correlation coefficient threshold, the real-time Pearson correlation coefficient of the seepage pressure data and the surface runoff data in the current time window is obtained and dynamically corrected by the correction factor to obtain the weight coefficient of the seepage pressure data. The correction factor is the sliding average deviation of the seepage pressure data.

[0020] The grid cells with abnormal spatial variance of rainfall intensity in the ground-based radar data within the fan-shaped area are synchronously located, the residual distribution is calculated based on the measured seepage rate and the rainfall intensity of the corresponding grid cells, and the residual distribution is spatially interpolated using a Gaussian kernel function to obtain a spatial variance correction coefficient matrix. The spatial variance of the ground-based radar data is locally weighted corrected according to the spatial variance correction coefficient matrix.

[0021] Furthermore, the method of obtaining the seepage field and surface runoff field of the tunnel and obtaining the nonlinear mapping relationship between rainfall intensity and tunnel water flow through alternating iterative calculation includes:

[0022] The method comprises obtaining dynamic change data of the permeability coefficient of the tunnel lining and geometric parameters of the surface runoff section; calculating the water pressure distribution behind the tunnel lining based on the dynamic change data of the permeability coefficient of the tunnel lining and the seepage pressure data, and dynamically correcting the permeability coefficient based on the water pressure distribution behind the tunnel lining; obtaining the surface runoff flow rate through data analysis of the corrected permeability coefficient and the geometric parameters of the surface runoff section; recalculating the water pressure distribution of the seepage field using the surface runoff flow rate as the boundary condition of the seepage field; and calculating the seepage field and the surface runoff field through alternating iterations until the difference in water pressure distribution between two iterations is less than a preset water pressure distribution difference threshold, thereby generating a steady-state coupling relationship between the seepage field and the surface runoff field.

[0023] Real-time water level gradient data of adjacent tunnel drainage systems are obtained, and a water collection interference intensity coefficient is calculated based on the adjacent tunnel spacing and the water level gradient change rate. The water collection interference intensity coefficient is used as a dynamic weight factor to correct the surface runoff boundary conditions in the tunnel group intersection area in the surface runoff model. The coupling relationship between the seepage field and the surface runoff field is updated based on the corrected surface runoff boundary conditions to obtain a corrected coupling relationship.

[0024] According to the steady-state coupling relationship and the modified coupling relationship, a nonlinear mapping relationship between rainfall intensity and tunnel water flow is established.

[0025] Furthermore, the method of obtaining the surface runoff flow rate by analyzing the corrected permeability coefficient and the surface runoff section geometric parameters includes:

[0026] A surface runoff model of the surface runoff field is established based on the corrected permeability coefficient and the surface runoff section geometric parameters; the spatiotemporal distribution data of the rainfall intensity at the current time step is input into the surface runoff model, and then a spatial discretization solution is performed to obtain the water depth and flow velocity distribution data of each grid unit; and the surface runoff flow rate is calculated based on the water depth and flow velocity distribution data and the surface runoff section geometric parameters.

[0027] Furthermore, the method of obtaining real-time water level gradient data of adjacent tunnel drainage systems and calculating the water confluence interference intensity coefficient based on the distance between adjacent tunnels and the water level gradient change rate includes:

[0028] By deploying an array of pressure-type water level sensors in the connecting channels of the adjacent tunnel drainage systems, the water level gradient data of each drainage well along the tunnel axis is collected in real time.

[0029] The geometric center spacing between adjacent tunnel drainage systems is calculated based on the spatial coordinate data in the tunnel group design drawings, and then converted into an equivalent hydraulic conduction distance in the hydrological model. The temporal variation characteristics of the water level gradient data within the current time window are simultaneously extracted, and the sliding difference method is used to calculate the water level gradient change rate between adjacent tunnel collection wells per unit time. The reciprocal of the equivalent hydraulic conduction distance and the water level gradient change rate are dimensionlessly processed to obtain the water collection interference intensity coefficient.

[0030] Furthermore, the method for establishing a nonlinear mapping relationship between rainfall intensity and tunnel water flow based on the steady-state coupling relationship and the modified coupling relationship includes:

[0031] The water pressure distribution characteristics of the seepage field under different permeability coefficients and the surface runoff flow time series data of the surface runoff field are extracted from the steady-state coupling relationship, and the surface runoff boundary condition parameters of the tunnel group intersection area after correction by the water interference intensity coefficient are obtained from the corrected coupling relationship.

[0032] The water pressure distribution characteristics, surface runoff flow time series data and modified surface runoff boundary condition parameters are fused with the spatiotemporal distribution data of rainfall intensity of the final rainfall model to generate an input feature vector including the spatiotemporal distribution of rainfall intensity, the dynamic change of tunnel lining permeability coefficient, the water level gradient of adjacent tunnels and the water interference intensity coefficient.

[0033] Based on the correspondence between the measured data of tunnel runoff flow in historical rainfall events and the input feature vector, a neural network is used for nonlinear regression modeling, where the number of input layer nodes is the dimension of the input feature vector, and the number of output layer nodes is the predicted tunnel runoff flow; when the root mean square error between the predicted runoff flow and the measured runoff flow data is less than a preset error threshold, a nonlinear mapping relationship between rainfall intensity and tunnel runoff flow is generated.

[0034] Furthermore, the method of outputting frequency adjustment instructions and start-stop control sequences of the drainage pump group according to the tunnel predicted water flow and the drainage pump group control reference parameters includes:

[0035] The flow difference between the predicted tunnel water flow and the current actual drainage flow of the drainage pump group is calculated in real time, and the frequency adjustment direction and frequency adjustment amplitude are generated according to the positive and negative signs and change rate of the flow difference.

[0036] A start-stop priority sequence is generated based on the liquid level gradient of the water collection well and the current output ratio of the drainage pump group. When the rising rate of the liquid level gradient in the water collection well exceeds the preset safety threshold, the standby drainage pump group is started first and the maximum frequency adjustment range is allocated; the frequency adjustment instruction is logically coupled with the start-stop priority sequence and then output to the drainage pump group controller for execution.

[0037] On the other hand, based on the same inventive concept, the present invention also provides an intelligent scheduling system for a tunnel drainage system based on multi-source data fusion, the system comprising: a final rainfall model acquisition module, a nonlinear mapping relationship generation module, and a drainage scheduling management module, wherein the modules are sequentially connected to each other;

[0038] The final rainfall model acquisition module is used to obtain meteorological satellite data, ground-based radar data and seepage pressure data in the tunnel, and to obtain the final rainfall model after spatiotemporal alignment through spatiotemporal nested data fusion method and dynamic allocation of weight coefficients.

[0039] The nonlinear mapping relationship generation module is used to obtain the tunnel seepage field and surface runoff field, and obtain the nonlinear mapping relationship between rainfall intensity and tunnel water flow through alternating iterative calculation.

[0040] The drainage scheduling management module is used to obtain the tunnel predicted water flow rate through the nonlinear mapping relationship of the temporal and spatial distribution data of the rainfall intensity of the final rainfall model, obtain the water level data of the water collection well of the tunnel drainage system and the operating output data of the drainage pump group and generate the drainage pump group control benchmark parameters; output the frequency adjustment instructions and start-stop control sequence of the drainage pump group according to the tunnel predicted water flow rate and the drainage pump group control benchmark parameters.

[0041] (3) Beneficial effects

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. By fusing multi-source data such as meteorological satellite data, ground-based radar data, and tunnel seepage pressure data, and applying the spatiotemporal nested data fusion method for spatiotemporal alignment and dynamic weight allocation, a final rainfall model after spatiotemporal calibration is generated, which significantly improves the spatiotemporal accuracy of rainfall forecasts and effectively overcomes the problems of traditional methods such as single data source and insufficient spatiotemporal calibration capabilities.

[0044] 2. Based on the final rainfall model after temporal and spatial calibration, an iterative calculation determines the nonlinear mapping relationship between rainfall intensity and tunnel runoff, making tunnel runoff predictions more accurate and reliable. This provides a key basis for intelligent scheduling of drainage pumps, enabling dynamic adjustment of pump operating parameters based on real-time predictions, significantly enhancing the tunnel drainage system's ability to cope with complex rainfall conditions.

[0045] 3. Taking into account the mutual interference effects of adjacent tunnel drainage systems and combining the control baseline parameters of the drainage pump groups, the system outputs frequency adjustment instructions and start-stop control sequences for the drainage pump groups, achieving coordinated and optimized scheduling between the tunnel drainage system and the drainage systems of adjacent tunnels. This not only improves the operating efficiency of individual tunnel drainage systems but also optimizes the drainage layout of the entire tunnel cluster, effectively avoiding drainage problems caused by interference from adjacent tunnel drainage systems and further improving the safety and reliability of the tunnel drainage system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flowchart of an intelligent scheduling method for a tunnel drainage system based on multi-source data fusion according to Example 1 of the present invention.

[0047] Figure 2 This is a schematic diagram of the module composition of the intelligent scheduling system for a tunnel drainage system based on multi-source data fusion according to Example 2 of the present invention. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] Before giving examples, it is necessary to explain the application scenarios of the present invention. The present invention is an intelligent scheduling method and system for tunnel drainage systems based on multi-source data fusion, which is used to solve the problem that traditional tunnel drainage system scheduling methods have low drainage prediction accuracy due to single data source, insufficient time and space calibration capabilities, failure to consider the nonlinear mapping relationship between rainfall and runoff, and the interference effect of adjacent tunnel drainage, resulting in unreasonable drainage scheduling.

[0050] Example 1: Figure 1 As shown, this embodiment provides an intelligent scheduling method for a tunnel drainage system based on multi-source data fusion, the method comprising:

[0051] S1. Acquire meteorological satellite data, ground-based radar data, and tunnel seepage pressure data. Using a spatiotemporal nested data fusion method, these data are aligned temporally and spatially, and weight coefficients are dynamically assigned to obtain a final, spatiotemporally calibrated rainfall model. Meteorological satellite data provides broad, macroscopic meteorological information, while ground-based radar data provides more precise local meteorological characteristics. Tunnel seepage pressure data reflects the hydrological conditions within the tunnel. Because different data sources vary in temporal and spatial accuracy and coverage, spatiotemporal nested data fusion integrates the strengths of each data source, eliminates spatiotemporal inconsistencies, and more accurately reflects the spatiotemporal distribution of rainfall.

[0052] S2. Obtain the tunnel's seepage and surface runoff fields, and through alternating iterative calculations, derive a nonlinear mapping relationship between rainfall intensity and tunnel runoff flow. The seepage field describes the flow of groundwater around the tunnel, while the surface runoff field reflects the distribution and flow of surface water. Rainfall intensity is a key factor affecting tunnel runoff flow, and the relationship between the two is not a simple linear one. Alternating iterative calculations can more accurately capture this complex nonlinear relationship.

[0053] S3. Calculate the predicted tunnel runoff flow rate using the nonlinear mapping relationship for the spatiotemporal distribution data of rainfall intensity from the final rainfall model. Obtain the water level data for the tunnel drainage system's water collection wells and the operating output data for the drainage pump groups, and generate control benchmark parameters for the drainage pump groups. Output frequency adjustment instructions and a start / stop control sequence for the drainage pump groups are generated based on the predicted tunnel runoff flow rate and the control benchmark parameters for the drainage pump groups. The water collection well level data reflects the water accumulation within the tunnel, while the operating output data for the drainage pump groups reflects their operating status. The control benchmark parameters for the drainage pump groups comprehensively consider the actual drainage demand within the tunnel and the operating capacity of the drainage pump groups.

[0054] The method of performing spatiotemporal alignment by a spatiotemporal nested data fusion method and dynamically allocating weight coefficients to obtain a final rainfall model after spatiotemporal calibration includes:

[0055] The spatial resolution of the meteorological satellite data was increased to the same grid scale as the ground-based radar data using bilinear interpolation. The timestamps of the ground-based radar data were synchronized to match the sampling frequency of the seepage pressure data. A multidimensional feature matrix was constructed based on the spatiotemporally aligned meteorological satellite, ground-based radar, and seepage pressure data. Each grid node in the multidimensional feature matrix contains the rainfall intensity from the meteorological satellite data, the rainfall intensity from the ground-based radar data, the seepage pressure, spatial coordinates, and timestamp. Meteorological satellite data typically has a lower spatial resolution, while ground-based radar data has a relatively higher spatial resolution. The spatial resolution of the meteorological satellite data was increased to the same grid scale as the ground-based radar data using bilinear interpolation, making the two comparable in spatial dimension. Furthermore, the timestamps of the ground-based radar data were synchronized to ensure temporal consistency between the different data sources.

[0056] Dynamically calculate weight coefficients for meteorological satellite data, ground-based radar data, and seepage pressure data. The weight coefficient for meteorological satellite data is determined based on the exponential decay characteristics of its historical error. The weight coefficient for ground-based radar data is calculated based on the spatial variance of rainfall intensity within its grid. The weight coefficient for seepage pressure data is dynamically adjusted based on its real-time Pearson correlation coefficient with surface runoff data, which is obtained by inverting the geometric parameters of surface runoff sections. The greater the historical error, the smaller the weight coefficient for meteorological satellite data. As the impact of the error decreases over time, the weight coefficient will also change accordingly, enabling a dynamic assessment of the reliability of historical data. The weight coefficient will be adjusted accordingly if the reliability of ground-based radar data in a particular area may be affected. Surface runoff data is obtained by inverting the geometric parameters of surface runoff sections. A higher correlation between seepage pressure data and surface runoff data indicates that the seepage pressure data more accurately reflects rainfall conditions and thus has a higher weight coefficient.

[0057] The rainfall intensity of meteorological satellite data, the rainfall intensity of ground-based radar data, and the seepage pressure of each grid node in the multidimensional feature matrix are weightedly calculated with the corresponding weight coefficients to generate an initial calibrated rainfall model; the weighted calculation comprehensively considers the contribution and reliability of different data sources, making the initial calibrated rainfall model more accurate.

[0058] The correlation coefficient between the seepage rate time series data and the rainfall intensity time series data within a preset sector-shaped area along the tunnel axis is detected in real time. If the correlation coefficient is detected to be below a preset correlation coefficient threshold, the weight coefficient of the seepage pressure data is recalculated and the spatial variance of the ground-based radar data is locally corrected. Based on the corrected weight coefficient, the rainfall intensity at the grid nodes in the tunnel cluster intersection area of ​​the initial calibrated rainfall model is iteratively optimized until the difference between the predicted rainfall intensity and the measured rainfall intensity is less than a preset rainfall intensity difference threshold. The final rainfall model, after spatiotemporal calibration, is then output. If the correlation coefficient is detected to be below the preset correlation coefficient threshold, it indicates that the relationship between the current seepage pressure data and rainfall intensity may have changed, requiring the weight coefficient of the seepage pressure data to be recalculated and the spatial variance of the ground-based radar data to be locally corrected. By continuously adjusting the weight coefficient and performing iterative calculations, the difference between the predicted and measured rainfall intensity is kept below the preset rainfall intensity difference threshold.

[0059] The method of detecting the correlation coefficient between the seepage rate time series data and the rainfall intensity time series data in a preset sector-shaped area in the tunnel axis direction in real time, and recalculating the weight coefficient of the seepage pressure data and locally correcting the spatial variance of the ground-based radar data when the correlation coefficient is detected to be lower than a preset correlation coefficient threshold, includes:

[0060] An array of seepage rate sensors is deployed within the pre-set sector-shaped area at preset sampling intervals to collect seepage rate time series data. This seepage rate time series data is then compared with the rainfall intensity time series data for the corresponding sector-shaped area in the initial calibration rainfall model using a sliding time window to calculate a correlation coefficient within the current time window. The seepage rate sensors can accurately measure the seepage rate at different locations within the area in real time. The sampling interval depends on the required monitoring accuracy and the scope of the area. A smaller sampling interval can provide more detailed seepage rate distribution information, but it also increases cost and data processing complexity. The sensor array continuously collects seepage rate data, generating seepage rate time series data that reflects the temporal changes in the seepage rate within the area. Sliding time windows are a commonly used time series analysis method that divides the entire time series data into multiple continuous time windows of a certain length. Within each time window, the seepage rate time series data and the rainfall intensity time series data are statistically analyzed to calculate a correlation coefficient. The correlation coefficient measures the degree of linear correlation between the two time series data. By calculating the correlation coefficient within the current time window through a sliding time window, the changes in the relationship between the seepage rate and rainfall intensity can be monitored in real time. If the correlation coefficient is high, it means that there is a strong linear correlation between the two, that is, changes in rainfall intensity will cause corresponding changes in the seepage rate; if the correlation coefficient is low, it means that the linear correlation between the two is weak.

[0061] When the correlation coefficient is lower than a preset correlation coefficient threshold, the real-time Pearson correlation coefficient between the seepage pressure data and the surface runoff data within the current time window is obtained and dynamically corrected using a correction factor to obtain a weight coefficient for the seepage pressure data. The correction factor is the moving average deviation of the seepage pressure data. The Pearson correlation coefficient is a statistic used to measure the degree of linear correlation between two variables. Within the current time window, the seepage pressure data and the surface runoff data are collected separately and the correlation coefficient is calculated using the Pearson correlation coefficient formula. The correction factor is the moving average deviation of the seepage pressure data. The moving average deviation refers to the average deviation between each data point and the average value of the data within the time period. By calculating the moving average deviation of the seepage pressure data, the fluctuation of the seepage pressure data is reflected, providing a reference for dynamically correcting the weight coefficient of the seepage pressure data. The specific correction method can be designed according to actual conditions. For example, the weight of the correction factor can be adjusted according to the magnitude of the correlation coefficient. The correction factor can then be multiplied by a coefficient and added or subtracted from the original weight coefficient to obtain a new weight coefficient.

[0062] Grid cells within the sector-shaped area with abnormal spatial variance of rainfall intensity in the ground-based radar data are simultaneously located. A residual distribution is calculated based on the measured seepage rate and the rainfall intensity of the corresponding grid cell. A Gaussian kernel function is used to spatially interpolate the residual distribution to obtain a spatial variance correction coefficient matrix. Based on the spatial variance correction coefficient matrix, a local weighted correction is performed on the spatial variance of the ground-based radar data. The ground-based radar data within the sector-shaped area is divided into a number of grid cells at a certain spatial resolution, with each grid cell corresponding to a specific spatial location. This gridding facilitates spatial analysis and processing of the data, calculating the spatial variance of rainfall intensity within each grid cell. Spatial variance reflects the degree of dispersion of rainfall intensity within a grid cell; a larger variance indicates a more dramatic variation in rainfall intensity within the grid cell. Based on the sensor's location information, the grid cell corresponding to each measured seepage rate data point is determined, and rainfall intensity data for that grid cell is obtained. For each measured seepage rate data point, the residual between it and the rainfall intensity of the corresponding grid cell is calculated. In this embodiment, the linear or nonlinear relationship between rainfall intensity and seepage rate is determined through statistical analysis of historical data, and the residual reflects the deviation between the two. All residual data are statistically analyzed according to their spatial positions to obtain the spatial distribution of the residuals. The Gaussian kernel function is a commonly used spatial interpolation function with good smoothness and locality. The Gaussian kernel function is used to perform spatial interpolation on the residual distribution to obtain the spatial variance correction coefficient corresponding to each grid cell in the fan-shaped area. The spatial variance correction coefficient reflects the degree of correction of the spatial variance of rainfall intensity in the grid cell. The larger the spatial variance correction coefficient, the greater the degree to which the spatial variance of rainfall intensity in the grid cell needs to be corrected. The spatial variance correction coefficients of all grid cells are arranged in the order of the grid to construct a spatial variance correction coefficient matrix. For each grid cell, its original spatial variance is multiplied by the corresponding spatial variance correction coefficient to obtain the corrected spatial variance. The spatial variance of the ground-based radar data is locally weighted corrected, so that the grid cells with abnormal spatial variance of rainfall intensity are properly adjusted, thereby improving the accuracy and reliability of the ground-based radar data.

[0063] The method of obtaining the tunnel seepage field and the surface runoff field and obtaining the nonlinear mapping relationship between rainfall intensity and tunnel water flow through alternating iterative calculation includes:

[0064] The method obtains dynamic data on the tunnel lining's permeability coefficient and geometric parameters of the surface runoff cross-section. The water pressure distribution behind the tunnel lining is calculated based on the dynamic data and seepage pressure data, and the permeability coefficient is dynamically corrected based on the water pressure distribution behind the tunnel lining. The corrected permeability coefficient and geometric parameters of the surface runoff cross-section are analyzed to obtain the surface runoff flow rate. The surface runoff flow rate is used as the boundary condition of the seepage field to recalculate the water pressure distribution of the seepage field. The seepage field and the surface runoff field are calculated through alternating iterations until the difference in water pressure distribution between two iterations is less than a preset water pressure distribution difference threshold. A steady-state coupling relationship between the seepage field and the surface runoff field is then generated. The permeability coefficient of the tunnel lining is affected by various factors, such as lining material degradation and groundwater erosion, which can cause the permeability coefficient to change over time. Obtaining this dynamic data can more accurately reflect the actual permeability performance of the tunnel lining. The geometric parameters of the surface runoff cross-section, including the shape and size of the cross-section, are crucial for calculating the surface runoff flow rate. For example, cross-sectional shapes such as rectangular, trapezoidal, or circular can cause variations in water velocity and flow rate through the cross-section. Based on the dynamic variation of the tunnel lining's permeability coefficient and seepage pressure data, and utilizing basic principles of seepage mechanics, such as Darcy's law, the water pressure distribution behind the tunnel lining is calculated. Seepage pressure data is acquired by installing pressure sensors behind the tunnel lining. Because the water pressure distribution affects the permeability of the tunnel lining, the permeability coefficient of the tunnel lining is dynamically corrected based on the calculated water pressure distribution. For example, increasing water pressure causes changes in the porosity of the lining material, thereby affecting the permeability coefficient. The corrected permeability coefficient and surface runoff cross-sectional geometric parameters are input into a suitable hydraulic model, such as the Manning equation, and the surface runoff flow rate is calculated through data analysis. Surface runoff flow rate is a critical parameter in the surface runoff field, reflecting the strength of the surface water flow. The calculated surface runoff flow rate is used as the boundary condition of the seepage field to recalculate the water pressure distribution within the seepage field. Surface runoff flow will have an impact on the seepage field. For example, surface runoff may enter the rock mass around the tunnel through infiltration and other means, changing the water pressure distribution of the seepage field. The recalculated water pressure distribution of the seepage field and the surface runoff field are alternately iterated. In each iteration, the boundary conditions of the surface runoff field, such as the infiltration amount, are updated according to the water pressure distribution of the seepage field. At the same time, the boundary conditions of the seepage field, such as the surface runoff recharge, are updated according to the changes in the surface runoff field. The difference in water pressure distribution between the two iterations is calculated. When the difference is less than the preset water pressure distribution difference threshold, the iterative process has converged. At this time, a steady-state coupling relationship between the seepage field and the surface runoff field is generated, reflecting the interaction and equilibrium state between the seepage field and the surface runoff field.

[0065] Real-time water level gradient data from adjacent tunnel drainage systems is obtained, and the water level interference intensity coefficient is calculated based on the distance between adjacent tunnels and the rate of change of the water level gradient. The water level interference intensity coefficient is used as a dynamic weighting factor to modify the surface runoff boundary conditions at the intersection of the tunnel group in the surface runoff model. The coupling relationship between the seepage field and the surface runoff field is updated based on the modified surface runoff boundary conditions to obtain a modified coupling relationship. Water level sensors are installed in the drainage systems of adjacent tunnels. These sensors can accurately measure the water level at different locations within the tunnels in real time. By regularly collecting this water level data, the water level distribution of the adjacent tunnel drainage systems at each moment can be obtained. The water level gradient refers to the rate of change of the water level in space, that is, the amount of change in the water level per unit distance. For adjacent tunnels, the water level gradient at different locations between the adjacent tunnels can be calculated based on the water level data at different locations. For example, at key locations such as the junction and intersection of adjacent tunnels, the calculated water level gradient can reflect the flow trend and intensity of water at these locations. Measuring the actual distance between adjacent tunnels is one of the important parameters for calculating the water level interference intensity coefficient. The distance between adjacent tunnels affects the flow and interaction of water between tunnels. In addition to the water level gradient itself, the rate of change of the water level gradient over time also needs to be analyzed. This rate of change can reflect the dynamics of water flow between adjacent tunnels, such as whether the flow is accelerating or decelerating. The water flow interference intensity coefficient comprehensively reflects the degree of interference between adjacent tunnels. A larger water flow interference intensity coefficient indicates greater interference intensity. Surface runoff models are typically used to describe the movement and distribution of surface water flow. The surface runoff boundary conditions at the intersection of a tunnel cluster are crucial to the accuracy of the model. The surface runoff boundary conditions at the intersection of a tunnel cluster are modified based on the magnitude of the water flow interference intensity coefficient. For example, a large water flow interference intensity coefficient indicates strong interference between adjacent tunnels. In this case, the surface runoff boundary conditions need to be adjusted to more accurately reflect the flow in this area. There is a close interaction between the seepage field and the surface runoff field. Surface runoff enters the ground through infiltration and other means, affecting the water pressure distribution and flow in the seepage field. Changes in the seepage field, in turn, affect the generation and flow of surface runoff. The revised surface runoff boundary conditions are input into the coupled model of the seepage and surface runoff fields, recalculating parameters such as the water pressure distribution and flow rate in these fields. The revised coupled relationship between the seepage and surface runoff fields more accurately reflects the actual interference between flows in adjacent tunnels.

[0066] According to the steady-state coupling relationship and the modified coupling relationship, a nonlinear mapping relationship between rainfall intensity and tunnel water flow is established.

[0067] The method of obtaining the surface runoff flow rate by analyzing the corrected permeability coefficient and the surface runoff cross-section geometric parameters includes:

[0068] A surface runoff model for the surface runoff field is established based on the corrected permeability coefficient and the geometric parameters of the surface runoff cross-section. The spatiotemporal distribution data of rainfall intensity at the current time step is input into the surface runoff model, which is then spatially discretized to obtain water depth and flow velocity distribution data for each grid cell. The surface runoff flow rate is calculated based on this water depth and flow velocity distribution data and the geometric parameters of the surface runoff cross-section. The surface runoff model is based on hydraulic principles, such as the continuity equation and the momentum equation, and is used to describe the generation, flow, and distribution of surface runoff. Surface runoff is a dynamic process, and the spatiotemporal distribution of rainfall intensity changes over time. Therefore, rainfall intensity data for the current time step is required to accurately simulate surface runoff. Rainfall intensity varies not only temporally but also spatially. For example, rainfall intensity may be high in some areas but low in others. Inputting this spatiotemporal distribution data of rainfall intensity into the surface runoff model can more realistically reflect the generation of surface runoff. To solve the surface runoff model, the study area must be divided into multiple grid cells through spatial discretization. Each grid cell represents a small spatial region. Calculations for each grid cell reveal the surface runoff for the entire study area. After inputting the spatiotemporal distribution of rainfall intensity into the surface runoff model, numerical methods are used to perform the spatial discretization. The solution considers various factors, such as topography, soil properties, and vegetation cover, to obtain water depth and flow velocity distribution data for each grid cell. Water depth reflects the degree of surface water accumulation, while flow velocity reflects the speed of water flow. Based on this water depth and flow velocity distribution data for each grid cell and combined with the geometric parameters of the surface runoff cross-section, the surface runoff flow rate for each grid cell can be calculated. For example, for a rectangular cross-section, the flow rate can be calculated based on the water depth and cross-section width; for a trapezoidal cross-section, factors such as the cross-section slope coefficient must also be considered. By summing the surface runoff flow rates for all grid cells, the surface runoff flow rate for the entire study area can be determined, which is important for assessing flood risk and designing drainage systems.

[0069] The method of obtaining real-time water level gradient data of adjacent tunnel drainage systems and calculating the water collection interference intensity coefficient based on the distance between adjacent tunnels and the water level gradient change rate includes:

[0070] By deploying an array of pressure-type water level sensors within the connecting channels of adjacent tunnel drainage systems, real-time data on the water level gradient of each drainage well along the tunnel axis is collected. By rationally arranging the array of pressure-type water level sensors within the connecting channels of adjacent tunnel drainage systems, the water pressure at the drainage wells can be accurately measured. Based on the corresponding relationship between pressure and water level, the water level data of each drainage well along the tunnel axis is obtained. By taking the difference between the water level data of adjacent drainage wells, the water level gradient between them is determined.

[0071] The geometric center distance between adjacent tunnel drainage systems is calculated based on the spatial coordinate data in the tunnel group design drawings, and the geometric center distance is converted into the equivalent hydraulic conduction distance in the hydrological model; the temporal variation characteristics of the water level gradient data in the current time window are simultaneously extracted, and the sliding difference method is used to calculate the water level gradient change rate between adjacent tunnel water collection wells per unit time; the inverse of the equivalent hydraulic conduction distance and the water level gradient change rate are dimensionlessly processed to obtain the watershed interference intensity coefficient. The watershed interference intensity coefficient is obtained through the watershed interference intensity calculation formula. , the interference intensity calculation formula is: ;in, is the water level gradient change rate, is the equivalent hydraulic conduction distance; This is an empirical coefficient calibrated based on measured data of interference between adjacent tunnel drainage systems during historical rainfall events. The spatial locations of adjacent tunnel drainage systems are determined based on the spatial coordinate data in the tunnel group design drawings. The geometric center coordinates of adjacent tunnel drainage systems are calculated to determine their geometric center spacing. This geometric center spacing is converted into an equivalent hydraulic conduction distance in the hydrological model. In actual hydrological processes, water flow between tunnel drainage systems is not a simple linear process and is influenced by multiple factors, such as soil permeability and drainage system structure. The equivalent hydraulic conduction distance takes these factors into account and more accurately reflects the flow characteristics between adjacent tunnel drainage systems. The temporal variation characteristics of the water level gradient data within the current time window are simultaneously extracted. The time window can be set according to actual needs. By analyzing the water level gradient data within the time window, the temporal trend of the water level gradient can be understood. The sliding difference method is used to calculate the rate of change of the water level gradient between adjacent tunnel sump wells per unit time. The sliding difference method is a commonly used time series data analysis method. It calculates the rate of change of the water level gradient by calculating the ratio of the water level gradient difference between adjacent time points to the time interval. For example, if at time t1 and t2 (t2>t1), the water level gradients between adjacent tunnel water collection wells are h1 and h2 respectively, and the time interval is =t2−t1, then the water level gradient change rate Dimensionless processing is used to eliminate the dimensional effects between different physical quantities, allowing comparison and calculation on the same scale. Dimensionless processing is usually achieved through methods such as standardization and normalization.

[0072] The method for establishing a nonlinear mapping relationship between rainfall intensity and tunnel water flow based on the steady-state coupling relationship and the modified coupling relationship includes:

[0073] The steady-state coupling relationship extracts the water pressure distribution characteristics of the seepage field under different permeability coefficients and the surface runoff flow time series data of the surface runoff field. The modified coupling relationship also obtains the surface runoff boundary condition parameters at the tunnel intersection area, corrected for the water flow interference intensity coefficient. By analyzing the steady-state coupling relationship, the water pressure values ​​and their distribution patterns at each point in the seepage field under different permeability coefficients can be extracted. For example, a larger permeability coefficient facilitates groundwater flow and may result in a relatively smaller water pressure gradient; conversely, a smaller permeability coefficient may result in a larger water pressure gradient. The steady-state coupling relationship can be used to obtain runoff flow data at different time points in the surface runoff field, forming a surface runoff flow time series. This data reflects the temporal variation of surface runoff. Affected by various factors such as precipitation intensity, topography, and soil type, a mutual recharge relationship exists between surface runoff and groundwater. Changes in surface runoff affect groundwater recharge, thereby affecting tunnel water flow. The modified coupling relationship is derived by modifying the steady-state coupling relationship to account for more practical factors, such as the water flow interference intensity coefficient. Surface runoff boundary condition parameters, such as boundary water level and flow rate, are crucial for describing the boundary characteristics of the surface runoff field. In the modified coupling relationship, these parameters change accordingly due to the consideration of the watershed interference intensity coefficient. The modified parameters, which account for the watershed interference intensity coefficient between adjacent tunnel drainage systems, more accurately describe the characteristics of the surface runoff field at the intersection of tunnel clusters.

[0074] The water pressure distribution characteristics, surface runoff flow time series data and modified surface runoff boundary condition parameters are subjected to multi-dimensional feature fusion with the spatiotemporal distribution data of rainfall intensity of the final rainfall model to generate an input feature vector containing the spatiotemporal distribution of rainfall intensity, the dynamic change of tunnel lining permeability coefficient, the water level gradient of adjacent tunnels and the watershed interference intensity coefficient; the above-mentioned multiple feature data are subjected to multi-dimensional feature fusion to generate an input feature vector containing rich information, which can comprehensively reflect the various factors affecting the tunnel watershed flow.

[0075] Based on the correspondence between the measured tunnel runoff data from historical rainfall events and the input feature vector, a neural network is used for nonlinear regression modeling. The number of input layer nodes is the dimension of the input feature vector, and the number of output layer nodes is the predicted tunnel runoff flow. When the root mean square error (RMS) between the predicted runoff flow and the measured runoff flow data is less than a preset error threshold, a nonlinear mapping relationship between rainfall intensity and tunnel runoff flow is generated. The number of input layer nodes is equal to the dimension of the input feature vector, with each node corresponding to a feature in the input feature vector. Multidimensional feature data can be input into the neural network, allowing it to learn the relationship between these features and tunnel runoff flow. The number of output layer nodes is the predicted tunnel runoff flow. The goal of the neural network is to predict the tunnel runoff flow based on the input feature vector. The RMS error (RMS) is a measure of the difference between the predicted runoff flow and the measured runoff flow data, reflecting the accuracy of the prediction model. When the RMS error between the predicted runoff flow and the measured runoff flow data is less than the preset error threshold, it indicates that the neural network model has learned the nonlinear relationship between factors such as rainfall intensity, tunnel lining permeability, adjacent tunnel water level gradient, and runoff interference intensity coefficient, and tunnel runoff flow. At this point, the neural network model can be used as a nonlinear mapping relationship to predict the tunnel runoff under different rainfall intensities and tunnel conditions, and can more accurately reflect the actual hydrogeological conditions and the complexity of the tunnel drainage system.

[0076] The method for outputting frequency adjustment instructions and start-stop control sequences of the drainage pump group according to the tunnel predicted water flow and the drainage pump group control reference parameters includes:

[0077] The flow difference between the predicted tunnel water flow and the current actual drainage flow of the drainage pump group is calculated in real time. The frequency adjustment direction and frequency adjustment amplitude are generated according to the positive and negative signs and the change rate of the flow difference. The frequency adjustment amplitude is mapped to the target frequency increment or decrement of the drainage pump group through the fuzzy control algorithm, and the target frequency adjustment amount is constrained by the liquid level safety threshold in the drainage pump group control reference parameters.

[0078] A start-stop priority sequence is generated based on the liquid level gradient of the water collection well and the current output ratio of the drainage pump group. When the rising rate of the liquid level gradient in the water collection well exceeds the preset safety threshold, the standby drainage pump group is started first and the maximum frequency adjustment amplitude is allocated; the frequency adjustment instruction is logically coupled with the start-stop priority sequence, and after eliminating the instantaneous fluctuation interference through the preset delay filter module, it is output to the drainage pump group controller for execution, and after execution, the flow difference value between the predicted and actual drainage flow is re-detected. If the flow difference value does not return to the threshold range, the membership function parameters of the fuzzy control algorithm are dynamically optimized according to the previous adjustment effect and the adjustment process is iterated.

[0079] Example 2: Based on the same inventive concept, Figure 2As shown, this embodiment also provides an intelligent scheduling system for a tunnel drainage system based on multi-source data fusion, the system comprising: a final rainfall model acquisition module, a nonlinear mapping relationship generation module, and a drainage scheduling management module, wherein the modules are sequentially connected to each other;

[0080] The final rainfall model acquisition module is used to obtain meteorological satellite data, ground-based radar data and seepage pressure data in the tunnel, and to obtain the final rainfall model after spatiotemporal alignment through spatiotemporal nested data fusion method and dynamic allocation of weight coefficients.

[0081] The nonlinear mapping relationship generation module is used to obtain the tunnel seepage field and surface runoff field, and obtain the nonlinear mapping relationship between rainfall intensity and tunnel water flow through alternating iterative calculation.

[0082] The drainage scheduling management module is used to obtain the tunnel predicted water flow rate through the nonlinear mapping relationship of the temporal and spatial distribution data of the rainfall intensity of the final rainfall model, obtain the water level data of the water collection well of the tunnel drainage system and the operating output data of the drainage pump group and generate the drainage pump group control benchmark parameters; output the frequency adjustment instructions and start-stop control sequence of the drainage pump group according to the tunnel predicted water flow rate and the drainage pump group control benchmark parameters.

[0083] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0084] Finally, it should be noted that although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent scheduling method for tunnel drainage system based on multi-source data fusion, characterized in that: The method comprises: Acquire meteorological satellite data, ground-based radar data, and tunnel seepage pressure data, and use the spatiotemporal nested data fusion method to perform spatiotemporal alignment and dynamically assign weight coefficients to obtain the final rainfall model after spatiotemporal calibration. Obtain the tunnel's seepage field and surface runoff field, and obtain the nonlinear mapping relationship between rainfall intensity and tunnel runoff flow through alternating iterative calculations; The spatiotemporal distribution data of rainfall intensity of the final rainfall model is used to obtain the tunnel predicted water flow through the nonlinear mapping relationship, the water level data of the water collection well of the tunnel drainage system and the operating output data of the drainage pump group are obtained, and the drainage pump group control reference parameters are generated; the frequency adjustment instruction and the start-stop control sequence of the drainage pump group are output according to the tunnel predicted water flow and the drainage pump group control reference parameters; The method of performing spatiotemporal alignment by a spatiotemporal nested data fusion method and dynamically allocating weight coefficients to obtain a final rainfall model after spatiotemporal calibration includes: The spatial resolution of the meteorological satellite data was increased to the same grid scale as the ground-based radar data using a bilinear interpolation method, and the timestamps of the ground-based radar data were synchronized to match the sampling frequency of the seepage pressure data. A multidimensional feature matrix was constructed based on the spatiotemporally aligned meteorological satellite data, ground-based radar data, and seepage pressure data. Each grid node in the multidimensional feature matrix contained the rainfall intensity of the meteorological satellite data, the rainfall intensity of the ground-based radar data, the seepage pressure, the spatial coordinates, and the timestamp. Dynamically calculate weight coefficients for meteorological satellite data, ground-based radar data, and seepage pressure data, wherein the weight coefficient for meteorological satellite data is determined based on the exponential decay characteristics of its historical error, the weight coefficient for ground-based radar data is calculated based on the spatial variance of rainfall intensity within its grid, and the weight coefficient for seepage pressure data is dynamically adjusted based on its real-time Pearson correlation coefficient with surface runoff data, where the surface runoff data is obtained by inversion of surface runoff section geometric parameters; Performing weighted calculation on the rainfall intensity of meteorological satellite data, rainfall intensity of ground-based radar data, seepage pressure and corresponding weight coefficients for each grid node in the multidimensional feature matrix to generate an initial calibrated rainfall model; The correlation coefficient between the seepage rate time series data and the rainfall intensity time series data in the preset fan-shaped area in the direction of the tunnel axis is detected in real time. When the correlation coefficient is detected to be lower than the preset correlation coefficient threshold, the weight coefficient of the seepage pressure data is recalculated and the spatial variance of the ground-based radar data is locally corrected. The rainfall intensity of the grid nodes in the intersection area of ​​the tunnel group in the initial calibration rainfall model is iteratively optimized according to the corrected weight coefficient until the difference between the predicted rainfall intensity and the measured rainfall intensity is less than the preset rainfall intensity difference threshold. The final rainfall model after spatiotemporal calibration is then output.

2. The intelligent scheduling method for tunnel drainage system based on multi-source data fusion according to claim 1 is characterized in that: The method of detecting the correlation coefficient between the seepage rate time series data and the rainfall intensity time series data in a preset sector-shaped area in the tunnel axis direction in real time, and recalculating the weight coefficient of the seepage pressure data and locally correcting the spatial variance of the ground-based radar data when the correlation coefficient is detected to be lower than a preset correlation coefficient threshold, includes: Arranging a seepage rate sensor array in the preset sector area at a preset sampling interval and collecting seepage rate time series data, calculating a correlation coefficient in a current time window between the seepage rate time series data and the rainfall intensity time series data of the corresponding sector area in the initial calibration rainfall model through a sliding time window; When the correlation coefficient is lower than the preset correlation coefficient threshold, the real-time Pearson correlation coefficient of the seepage pressure data and the surface runoff data in the current time window is obtained and the weight coefficient of the seepage pressure data is obtained by dynamic correction using a correction factor, where the correction factor is the sliding average deviation of the seepage pressure data; The grid cells with abnormal spatial variance of rainfall intensity in the ground-based radar data within the fan-shaped area are synchronously located, the residual distribution is calculated based on the measured seepage rate and the rainfall intensity of the corresponding grid cells, and the residual distribution is spatially interpolated using a Gaussian kernel function to obtain a spatial variance correction coefficient matrix. The spatial variance of the ground-based radar data is locally weighted corrected according to the spatial variance correction coefficient matrix.

3. The intelligent scheduling method for tunnel drainage system based on multi-source data fusion according to claim 1 is characterized in that: The method of obtaining the tunnel seepage field and the surface runoff field and obtaining the nonlinear mapping relationship between rainfall intensity and tunnel water flow through alternating iterative calculation includes: Acquire dynamic change data of the tunnel lining permeability coefficient and geometric parameters of the surface runoff section; calculate the water pressure distribution behind the tunnel lining based on the dynamic change data of the tunnel lining permeability coefficient and the seepage pressure data, and dynamically correct the permeability coefficient based on the water pressure distribution behind the tunnel lining; obtain the surface runoff flow rate through data analysis of the corrected permeability coefficient and the geometric parameters of the surface runoff section; use the surface runoff flow rate as a boundary condition of the seepage field to recalculate the water pressure distribution of the seepage field; calculate the seepage field and the surface runoff field through alternating iterative calculations until the difference in water pressure distribution between two iterations is less than a preset water pressure distribution difference threshold, thereby generating a steady-state coupling relationship between the seepage field and the surface runoff field; Real-time water level gradient data of adjacent tunnel drainage systems are obtained, and a water collection interference intensity coefficient is calculated based on the distance between adjacent tunnels and the water level gradient change rate; the water collection interference intensity coefficient is used as a dynamic weight factor to correct the surface runoff boundary conditions of the tunnel group intersection area in the surface runoff model; and the coupling relationship between the seepage field and the surface runoff field is updated based on the corrected surface runoff boundary conditions to obtain a corrected coupling relationship; According to the steady-state coupling relationship and the modified coupling relationship, a nonlinear mapping relationship between rainfall intensity and tunnel water flow is established.

4. The intelligent scheduling method for tunnel drainage system based on multi-source data fusion according to claim 3 is characterized in that: The method for obtaining the surface runoff flow rate by analyzing the corrected permeability coefficient and the surface runoff cross-section geometric parameters includes: A surface runoff model of the surface runoff field is established based on the corrected permeability coefficient and the surface runoff section geometric parameters; the spatiotemporal distribution data of the rainfall intensity at the current time step is input into the surface runoff model, and then a spatial discretization solution is performed to obtain the water depth and flow velocity distribution data of each grid unit; and the surface runoff flow rate is calculated based on the water depth and flow velocity distribution data and the surface runoff section geometric parameters.

5. The intelligent scheduling method for tunnel drainage system based on multi-source data fusion according to claim 3 is characterized in that: The method of obtaining real-time water level gradient data of adjacent tunnel drainage systems and calculating the water collection interference intensity coefficient based on the distance between adjacent tunnels and the water level gradient change rate includes: By deploying an array of pressure-type water level sensors in the connecting channels of adjacent tunnel drainage systems, the water level gradient data of each drainage well along the tunnel axis is collected in real time. The geometric center spacing between adjacent tunnel drainage systems is calculated based on the spatial coordinate data in the tunnel group design drawings, and then converted into an equivalent hydraulic conduction distance in the hydrological model. The temporal variation characteristics of the water level gradient data within the current time window are simultaneously extracted, and the sliding difference method is used to calculate the water level gradient change rate between adjacent tunnel collection wells per unit time. The reciprocal of the equivalent hydraulic conduction distance and the water level gradient change rate are dimensionlessly processed to obtain the water collection interference intensity coefficient.

6. The intelligent scheduling method for tunnel drainage system based on multi-source data fusion according to claim 3 is characterized in that: The method for establishing a nonlinear mapping relationship between rainfall intensity and tunnel water flow based on the steady-state coupling relationship and the modified coupling relationship includes: The water pressure distribution characteristics of the seepage field under different permeability coefficients and the surface runoff flow time series data of the surface runoff field are extracted from the steady-state coupling relationship, and the surface runoff boundary condition parameters of the tunnel group intersection area after being corrected by the water interference intensity coefficient are obtained from the corrected coupling relationship. The water pressure distribution characteristics, surface runoff flow time series data and modified surface runoff boundary condition parameters are fused with the rainfall intensity spatiotemporal distribution data of the final rainfall model to generate an input feature vector including the spatiotemporal distribution of rainfall intensity, the dynamic change of tunnel lining permeability coefficient, the water level gradient of adjacent tunnels and the water interference intensity coefficient; Based on the correspondence between the measured data of tunnel runoff flow in historical rainfall events and the input feature vector, a neural network is used for nonlinear regression modeling, where the number of input layer nodes is the dimension of the input feature vector, and the number of output layer nodes is the predicted tunnel runoff flow; when the root mean square error between the predicted runoff flow and the measured runoff flow data is less than a preset error threshold, a nonlinear mapping relationship between rainfall intensity and tunnel runoff flow is generated.

7. The intelligent scheduling method for tunnel drainage system based on multi-source data fusion according to claim 1 is characterized in that: The method for outputting frequency adjustment instructions and start-stop control sequences of the drainage pump group according to the tunnel predicted water flow and the drainage pump group control reference parameters includes: Calculate the difference between the predicted tunnel water flow and the actual drainage flow of the drainage pump group in real time, and generate the frequency adjustment direction and frequency adjustment amplitude based on the positive and negative signs and change rate of the flow difference; A start-stop priority sequence is generated based on the liquid level gradient of the water collection well and the current output ratio of the drainage pump group. When the rising rate of the liquid level gradient in the water collection well exceeds the preset safety threshold, the standby drainage pump group is started first and the maximum frequency adjustment range is allocated; the frequency adjustment instruction is logically coupled with the start-stop priority sequence and then output to the drainage pump group controller for execution.

8. The intelligent scheduling system of tunnel drainage system based on multi-source data fusion is characterized by: The system includes: a final rainfall model acquisition module, a nonlinear mapping relationship generation module, and a drainage scheduling management module, and each module is sequentially connected to communicate with each other; The final rainfall model acquisition module is used to obtain meteorological satellite data, ground-based radar data, and seepage pressure data in the tunnel, and to align them in time and space through a spatiotemporal nested data fusion method and dynamically assign weight coefficients to obtain the final rainfall model after spatiotemporal calibration; The nonlinear mapping relationship generation module is used to obtain the tunnel seepage field and surface runoff field, and obtain the nonlinear mapping relationship between rainfall intensity and tunnel water flow through alternating iterative calculation; a drainage scheduling management module for obtaining a tunnel predicted water flow rate from the spatiotemporal distribution data of the rainfall intensity of the final rainfall model through the nonlinear mapping relationship, obtaining the water level data of the water collection well of the tunnel drainage system and the operating output data of the drainage pump group and generating the drainage pump group control reference parameters; and outputting a frequency adjustment instruction and a start-stop control sequence for the drainage pump group based on the tunnel predicted water flow rate and the drainage pump group control reference parameters; The method of performing spatiotemporal alignment by a spatiotemporal nested data fusion method and dynamically allocating weight coefficients to obtain a final rainfall model after spatiotemporal calibration includes: The spatial resolution of the meteorological satellite data was increased to the same grid scale as the ground-based radar data using a bilinear interpolation method, and the timestamps of the ground-based radar data were synchronized to match the sampling frequency of the seepage pressure data. A multidimensional feature matrix was constructed based on the spatiotemporally aligned meteorological satellite data, ground-based radar data, and seepage pressure data. Each grid node in the multidimensional feature matrix contained the rainfall intensity of the meteorological satellite data, the rainfall intensity of the ground-based radar data, the seepage pressure, the spatial coordinates, and the timestamp. Dynamically calculate weight coefficients for meteorological satellite data, ground-based radar data, and seepage pressure data, wherein the weight coefficient for meteorological satellite data is determined based on the exponential decay characteristics of its historical error, the weight coefficient for ground-based radar data is calculated based on the spatial variance of rainfall intensity within its grid, and the weight coefficient for seepage pressure data is dynamically adjusted based on its real-time Pearson correlation coefficient with surface runoff data, where the surface runoff data is obtained by inversion of surface runoff section geometric parameters; Performing weighted calculation on the rainfall intensity of meteorological satellite data, rainfall intensity of ground-based radar data, seepage pressure and corresponding weight coefficients for each grid node in the multidimensional feature matrix to generate an initial calibrated rainfall model; The correlation coefficient between the seepage rate time series data and the rainfall intensity time series data in the preset fan-shaped area in the direction of the tunnel axis is detected in real time. When the correlation coefficient is detected to be lower than the preset correlation coefficient threshold, the weight coefficient of the seepage pressure data is recalculated and the spatial variance of the ground-based radar data is locally corrected. The rainfall intensity of the grid nodes in the intersection area of ​​the tunnel group in the initial calibration rainfall model is iteratively optimized according to the corrected weight coefficient until the difference between the predicted rainfall intensity and the measured rainfall intensity is less than the preset rainfall intensity difference threshold. The final rainfall model after spatiotemporal calibration is then output.

Images