Intelligent reservoir spillway water level monitoring and spilling volume early warning system

Abstract

The invention relates to the technical field of water level monitoring and early warning, in particular to an intelligent reservoir spillway water level monitoring and spillway early warning system, which comprises an anti-interference water level sensing module, an anti-siltation sensing array is arranged on a key section of a spillway, and the key section comprises a curve section and a gate section; the dynamic hydraulic twinning module is used for processing through a lightweight fluid network algorithm on the basis of the water surface transverse slope, the absolute water level and the gate vibration displacement in combination with a hydraulic BIM model; the edge cloud cooperative flow calculation module comprises an edge node and a cloud server; the risk self-adaptive early warning module is used for executing early warning when the calibrated real-time spilling volume breaks through a spilling threshold value; according to the method, the flow calculation error under the high turbulence working condition is remarkably reduced, the accuracy of flood discharge amount calculation is effectively improved, and it is ensured that flood discharge control is more accurate especially under the scene that the curve flow state is complex and the gate opening degree changes remarkably.

Description

Technical Field

[0001] The present invention relates to the technical field of water level monitoring and early warning, and particularly to an intelligent reservoir spillway water level monitoring and spillway discharge early warning system. Background Art

[0002] As an important hydraulic structure for regulating the reservoir water level and ensuring the safe operation of the reservoir, the spillway's flood discharge capacity and regulation method directly affect the flood control efficiency and the safety of the downstream area. Under the traditional reservoir management mode, the calculation of the spillway discharge depends on a static hydraulic model, and the spillway discharge is calculated using a fixed design roughness coefficient and empirical formula. However, this method has several limitations:

[0003] The traditional method does not consider the dynamic changes in the spillway flow pattern. Especially under complex working conditions such as non-linear flow patterns in the bend section, changes in the gate opening, and enhanced turbulence, the fixed roughness coefficient model is difficult to accurately describe the water flow characteristics, resulting in a large error in the calculation of the spillway discharge and potentially causing risks of over-flood discharge or insufficient regulation.

[0004] Traditional reservoir flood discharge management relies on historical experience and fixed scheduling rules and cannot effectively respond to sudden heavy rainfall or a sharp increase in the upstream flood inflow. Due to the lack of a flow prediction mechanism, managers usually adopt an ex-post adjustment method for gate regulation, resulting in problems such as response lag and water level overshoot, which affect flood control safety. Summary of the Invention

[0005] The present invention provides an intelligent reservoir spillway water level monitoring and spillway discharge early warning system.

[0006] The intelligent reservoir spillway water level monitoring and spillway discharge early warning system includes:

[0007] An anti-interference water level sensing module: An anti-silting sensing array is arranged at key sections of the spillway. The key sections include the bend section and the gate section, where:

[0008] The bend section uses a dual-frequency millimeter-wave radar (24 GHz / 77 GHz) to scan the lateral water surface slope;

[0009] The gate section deploys a pressure-ultrasonic fusion sensing group to synchronously obtain the absolute water level and the gate vibration displacement;

[0010] A dynamic hydraulic digital twin module: Based on the lateral water surface slope, the absolute water level, the gate vibration displacement, and combined with the hydraulic BIM model, the following are processed through a lightweight fluid network algorithm:

[0011] ① Calculate the bend centrifugal force correction coefficient based on the lateral slope;

[0012] ② Convert the vibration displacement into the dynamic shrinkage rate of the flow-through section;

[0013] ③Generate a digital twin including the turbulence intensity every 5 minutes;

[0014] Edge-cloud collaborative flow calculation module: It includes an edge node and a cloud server, where:

[0015] The edge node executes: Calculate the benchmark spillway discharge using the improved Manning formula, where the roughness coefficient is dynamically updated by the digital twin;

[0016] The cloud server executes: Predict the flow deviation value in the future time (30 minutes) through a long short-term memory network (LSTM), and send it back to the edge node for calibration. The edge node calibration includes adding the flow deviation value to the benchmark spillway discharge to generate the calibrated real-time spillway discharge;

[0017] Risk adaptive warning module:

[0018] When the calibrated real-time spillway discharge exceeds the spillway threshold, execute the following:

[0019] ①Divide the risk level according to the turbulence intensity of the digital twin;

[0020] ②Based on the risk level, trigger the gate opening control instruction and synchronously push it to the reservoir operation terminal.

[0021] Optionally, the layout method of the anti-silting sensor array includes:

[0022] Arrange a dual-frequency millimeter-wave radar group along the bend section. Among them, the 24GHz radar array scans the transverse water surface slope at a 15° elevation angle, and calculates the slope angle through the phase difference; the 77GHz radar captures the micro-deformation of the water surface in a vertical incidence manner, and the dual-frequency data is fused by Kalman filtering to generate the water surface transverse slope value α resistant to haze interference fused ;

[0023] Install a pressure-ultrasonic fusion sensor group on the upstream face of the gate section. Among them, the pressure sensor array is distributed according to the isobar, and MEMS piezoresistive probes are arranged at intervals of 50cm to measure the absolute water pressure value and calculate the water level. The ultrasonic transmitter is embedded in the center of the pressure sensor and emits a 1MHz pulsed wave through a 10° deflection angle, and calculates the gate vibration displacement Δd based on the echo time delay difference.

[0024] Optionally, the dynamic hydraulic twin module specifically includes:

[0025] Centrifugal force correction coefficient calculation unit: Based on the transverse water surface slope angle α of the bend section fused And the hydraulic radius R in the hydraulic BIM model, introduce a time decay factor, and calculate the centrifugal force correction coefficient K through a lightweight fluid network algorithm c ;

[0026] Dynamic shrinkage rate conversion unit: Combining the main vibration frequency and the designed opening of the gate, maps the gate vibration displacement Δd to the cross-sectional shrinkage rate η of the flow-through section;

[0027] Digital twin generation unit: Combining the centrifugal force correction coefficient and the cross-sectional shrinkage rate of the flow-through section, inputs them into the lightweight fluid network for numerical solution to obtain the flow field distribution. The lightweight fluid network is based on the Reynolds-averaged N-S equations, calculates the velocity changes of the water flow in the bend and gate areas, and quantifies the turbulence intensity. After the calculation is completed, writes the key parameters into the property table of the digital twin.

[0028] Optionally, in the digital twin generation unit, the following operations are performed every 5 minutes:

[0029] (1) Input K c , η into the lightweight fluid network and solve the Reynolds-averaged N-S equations:

[0030] where, τ ij is the Reynolds stress term, δ iθ is the centrifugal force direction tensor, is the Reynolds-averaged velocity component, representing the average velocity of the water flow in the i direction, is the Reynolds-averaged velocity component, representing the average velocity of the water flow in the j direction, combined with the spillway flow field, is the local velocity change rate, is the convection term, ρ is the water body density, is the Reynolds-averaged pressure, representing the average pressure of the fluid, is the pressure gradient force, ν is the dynamic viscosity, representing the dynamic viscosity of water, taking 1.0×10 -6 m 2 / s, is the viscous dissipation term, τ ij is the Reynolds stress term, representing the additional stress caused by turbulence;

[0031] (2) Output the global turbulence intensity I t through the embedded solver;

[0032] (3) Write K c , η, I t into the digital twin property table and fuse with the geometric data of the hydraulic BIM model to generate a visual flow field atlas.

[0033] Optionally, in the edge node, the improved Manning formula is used to calculate the reference spillway discharge Q b :

[0034] where, the update of the dynamic roughness coefficient n dyn is expressed as:

[0035] Among them, n0 is the reference roughness coefficient in the hydraulic BIM model (extracted from the digital twin property library), I t is the current turbulent intensity output by the digital twin, η is the contraction rate of the flow cross-section, A represents the flow cross-sectional area, R is the hydraulic radius, S is the energy slope, which represents the energy loss rate per unit length of the water flow and is approximately calculated by the transverse water surface slope, R 2 / 3 represents the 2 / 3 power of the hydraulic radius and is used to describe the dependence of the water flow on the channel geometry in the Manning formula, S 1 / 2 represents the square root of the energy slope and is used to characterize the influence of the gravity driving force of the water flow on the flow rate in the Manning formula.

[0036] Optionally, the long short-term memory network in the cloud server predicts the flow deviation value in the next 30 minutes. The input features of the long short-term memory network include:

[0037] The time series data of the reference spillway discharge in the previous 1 hour (sampling interval: 5 minutes);

[0038] I output by the digital twin t , k c , η;

[0039] The rainfall intensity in the next 30 minutes predicted by the weather forecast;

[0040] The output of the long short-term memory network includes the flow deviation value ΔQ(t) in the next 30 minutes, with a time resolution of 5 minutes.

[0041] Optionally, the edge node executes every 5 minutes:

[0042] Receive the ΔQ(t) sequence sent from the cloud;

[0043] Align the reference spillway discharge Q b (t) according to the timestamp;

[0044] Generate the calibrated real-time spillway discharge: Q cal (t) = Q b (t) + ΔQ(t)·sigmoid(t / 30), where sigmoid(t / 30) is the time weight function, which gives higher confidence to proximal predictions, making the prediction weight 0.82 for the next 5 minutes → 0.12 for 30 minutes, conforming to the physical law that the uncertainty of hydrological prediction increases with time.

[0045] Optionally, the spillway threshold in the risk adaptive warning module is expressed as Q th , when the calibrated real-time spillway discharge Q cal ≥Q th a warning is triggered, Among them, Q designFor the designed flood discharge capacity in the hydraulic BIM model, β is the working condition adjustment factor (1.2 for the flood season and 0.9 for the non-flood season), and I t,max is the maximum allowable turbulent intensity in the material durability database (25% for concrete structures).

[0046] Optionally, the risk level division according to the turbulent intensity of the digital twin specifically includes:

[0047] I t When < 12%, it is set as level I;

[0048] 12% ≤ I t < 18%, it is set as level II;

[0049] I t > 18%, it is set as level III.

[0050] Optionally, the trigger gate opening control instruction includes:

[0051] Input the current risk level;

[0052] Output the gate priority regulation sequence:

[0053] If the risk level = level I: Maintain the current situation and increase the monitoring frequency;

[0054] If the risk level = level II: Prioritize opening the gate farthest from the bend (reduce the impact of secondary flow);

[0055] If the risk level = level III: Alternately open the gates to form staggered flood discharge (interval ≤ 30 seconds).

[0056] Advantages of the present invention:

[0057] In the present invention, by introducing a dynamic hydraulic digital twin, key parameters such as the lateral water surface slope and gate vibration displacement are calculated in real time, and the Reynolds-averaged N-S equation is solved by combining a lightweight fluid network. The turbulent intensity, centrifugal force correction coefficient, and cross-sectional shrinkage rate are dynamically updated. The improved Manning formula is adopted and calibrated by the dynamic roughness coefficient fed back by the twin, enabling the flow rate calculation to be adaptively adjusted according to the changes in the water flow state. Compared with the traditional static roughness method, the flow rate calculation error of the present invention is significantly reduced under high-turbulent conditions, effectively improving the accuracy of the spillway discharge calculation. Especially in scenarios with complex bend flow patterns and significant changes in gate openings, it ensures more accurate flood discharge control.

[0058] The present invention adopts an edge computing + cloud intelligent prediction architecture. The edge nodes execute the calculation of the benchmark spillway discharge, and the cloud predicts the future flow deviation based on LSTM, enabling the system to adjust the spillway strategy in advance. During the data fusion process, a time-weighted correction mechanism is introduced to adjust the trust level according to the distance of the prediction time, ensuring that the proximal prediction is more reliable. Compared with the traditional flood discharge scheme based on a static hydraulic model, the present invention can predict the abnormal trend of spillway in advance, dynamically adjust the gate opening, reduce the risk of sudden over-flow, and improve the intelligent level of reservoir flood discharge scheduling.

[0059] The present invention adopts a dynamic threshold determination model based on the correction of turbulent intensity. According to the ratio of the current turbulent intensity to the historical maximum tolerance intensity, the early warning trigger threshold is dynamically adjusted to avoid the problems of premature or late warning caused by the increase of turbulent intensity. Combined with a hierarchical early warning mechanism, the early warning level can be divided according to the amplitude of the spillway discharge exceeding the design value, and the adjustment of the gate opening and the reservoir dispatching terminal are linked to achieve full-process automatic control from mild spillway regulation to emergency flood discharge response. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0061] Figure 1 It is a schematic diagram of the system function module of the embodiment of the present invention;

[0062] Figure 2 It is a schematic diagram of the dynamic hydraulic twin module of the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] The following will describe the present invention in detail with reference to the drawings and specific embodiments. At the same time, it should be noted that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; and the drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0064] It should be noted that in the specification, it is mentioned that "an embodiment", "embodiment", "exemplary embodiment", "some embodiments", etc. indicate that the described embodiment may include specific features, structures or characteristics, but not necessarily every embodiment includes the specific feature, structure or characteristic. In addition, when combining an embodiment to describe a specific feature, structure or characteristic, implementing such a feature, structure or characteristic in combination with other embodiments (whether explicitly described or not) should be within the knowledge scope of those skilled in the relevant art.

[0065] Generally, terms can be understood at least in part from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or property in the singular sense, or can be used to describe a combination of features, structures, or properties in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey a set of exclusive factors, but rather can alternatively, depending at least in part on the context, allow for the existence of other factors that are not necessarily explicitly described.

[0066] As Figure 1 - Figure 2 shown, the intelligent reservoir spillway water level monitoring and spill volume early warning system includes:

[0067] Anti-interference water level sensing module: An anti-silting sensing array is arranged at the key cross-sections of the spillway. The key cross-sections include the bend section and the gate section, where:

[0068] In the bend section, a dual-frequency millimeter-wave radar (24GHz / 77GHz) is used to scan the lateral water surface slope;

[0069] In the gate section, a pressure-ultrasonic fusion sensing group is deployed to synchronously obtain the absolute water level and the gate vibration displacement;

[0070] Dynamic hydraulic digital twin module: Based on the lateral water surface slope, absolute water level, gate vibration displacement, and combined with the hydraulic BIM model, the following is processed through a lightweight fluid network algorithm:

[0071] ① Calculate the bend centrifugal force correction coefficient based on the lateral slope;

[0072] ② Convert the vibration displacement into the dynamic shrinkage rate of the flow cross-section;

[0073] ③ Generate a digital twin including the turbulence intensity every 5 minutes;

[0074] Edge-cloud collaborative flow calculation module: Includes edge nodes and cloud servers, where:

[0075] The edge node executes: Use the improved Manning formula to calculate the reference spill volume, where the roughness coefficient is dynamically updated by the digital twin;

[0076] The cloud server executes: Predict the flow deviation value for the future time (30 minutes) through a long short-term memory network (LSTM), and send it back to the edge node for calibration. The edge node calibration includes adding the flow deviation value to the reference spill volume to generate the calibrated real-time spill volume;

[0077] Risk adaptive early warning module:

[0078] When the calibrated real-time spill volume breaks through the spill threshold, the following is executed:

[0079] ① The spillway discharge only reflects the total water volume and cannot characterize the distribution of fluid kinetic energy (such as the local damage risk of high-speed water flow impacting the pier). The sudden change in spillway discharge often lags behind the instability of the flow regime (the sudden increase in turbulence intensity is earlier than the significant change in flow rate). Therefore, the risk level is divided according to the turbulence intensity of the digital twin body;

[0080] ② Based on the risk level, trigger the gate opening control instruction and synchronously push it to the reservoir operation terminal.

[0081] The layout methods of the anti-silting sensor array include:

[0082] Arrange a dual-frequency millimeter-wave radar group along the bend section. Among them, the 24GHz radar array scans the lateral water surface slope at an elevation angle of 15°, and calculates the slope angle through the phase difference; the 77GHz radar captures the micro-deformation of the water surface in a vertical incidence manner. The dual-frequency data is fused by Kalman filtering to generate the lateral water surface slope value α resistant to haze interference fused ;

[0083] Calculation of the lateral slope angle in the bend section:

[0084] Phase difference height measurement formula: Among them, Δh is the height difference between adjacent measuring points on the water surface, c is the speed of light, is the phase difference of the 24GHz radar echo, and f is the radar center frequency (24GHz = 24×10 9 Hz);

[0085] Calculation of the slope angle: Among them, α 24 is the lateral slope angle calculated based on 24GHz, L is the distance between adjacent radar measuring points, and α 77 is calculated in the same way, and then the weight of the 77GHz radar data (value range 0-1) is calculated: Among them, C is the real-time haze concentration, C0 is the concentration threshold (50mg / m 3 ), and k1 is the adjustment factor (take 0.1); the fused lateral water surface slope value α fused is expressed as:

[0086] α fused = w 77 ·α 77 +(1 - w 77 )·α 24 ;

[0087] Install a pressure-ultrasonic fusion sensor group on the water-facing side of the gate section. Among them, the pressure sensor array is distributed according to the isobaric lines, and MEMS piezoresistive probes are arranged at intervals of 50 cm to measure the absolute water pressure value and calculate the water level. The hydrostatic pressure equation is used to calculate the water level: P = ρgh1 + P0, where P is the total pressure measured by the pressure sensor, ρ is the water density, g is the acceleration due to gravity, h1 is the absolute water level height, and P0 is the atmospheric pressure. The calculation result is: The ultrasonic transmitter is embedded in the center of the pressure sensor and emits a 1 MHz pulsed wave through a 10° deflection angle. The vibration displacement Δd of the gate is calculated based on the echo time delay difference.

[0088] Calculation of ultrasonic vibration displacement:

[0089] Range measurement based on time delay difference: Among them, Δd is the vibration displacement of the gate, v is the propagation speed of ultrasonic waves in the medium (1500 m / s in water), and Δt is the time delay difference between adjacent pulse echoes;

[0090] Determination of displacement direction: Displacement direction = sign(Δt1 - Δt2)·cosθ, where Δt1 and Δt2 are the echo time delays of two offset probes, and θ is the ultrasonic emission deflection angle of 10°.

[0091] Example: The hydraulic radius R of the spillway of a certain reservoir is 85 m, and the deployment plan is as follows:

[0092] Radar layout in the bend section: One set of dual-frequency radars is arranged every 10 m along the center line of the bend (a total of 8 sets);

[0093] The 24 GHz radar detects that the lateral slope angle α = 0.8°, and after correction by the 77 GHz radar, α' = 0.78°;

[0094] Probe installation in the gate section: Install a 5-column × 3-row fusion probe matrix on the water-facing side of each of the 4 gates.

[0095] The dynamic hydraulic twin module specifically includes:

[0096] Centrifugal force correction coefficient calculation unit: Based on the lateral slope angle α of the water surface in the bend section fused And the hydraulic radius R in the hydraulic BIM model, introduce a time decay factor, and calculate the centrifugal force correction coefficient K through a lightweight fluid network algorithm c :

[0097] Among them, is the exponential decay term (time decay factor), which quantifies the influence of the adaptation time of the bend flow on the centrifugal force (the traditional method ignores the time accumulation effect), v is the real-time flow velocity, t is the flood duration (extracting the design flood hydrograph from the hydraulic BIM model), and τ is the bend flow regime response time constant (taking R / 2v);

[0098] Dynamic contraction rate conversion unit: Combining the main vibration frequency and the designed gate opening, maps the gate vibration displacement Δd to the flow cross-section contraction rate η:

[0099] where f v is the main vibration frequency (extracting the peak frequency of the vibration waveform through FFT), D is the designed gate opening (obtained from the hydraulic BIM model), t v is the time stamp within the vibration period, and T is the natural vibration period of the gate (provided by the material property library in the hydraulic BIM model);

[0100] Digital twin generation unit: Combining the centrifugal force correction coefficient and the flow cross-section contraction rate, inputs them into the lightweight fluid network for numerical solution to obtain the flow field distribution. The lightweight fluid network is based on the Reynolds-averaged N-S equations, calculates the velocity changes of the water flow in the bend and gate areas, and quantifies the turbulence intensity. After the calculation is completed, writes the key parameters into the property table of the digital twin.

[0101] In the digital twin generation unit, the following operations are performed every 5 minutes:

[0102] (1) Input K c , η into the lightweight fluid network and solve the Reynolds-averaged N-S equations:

[0103] where τ ij is the Reynolds stress term, δ iθ is the centrifugal force direction tensor, is the Reynolds-averaged velocity component, representing the average velocity of the water flow in the i direction, mainly used to describe the main flow direction velocity component of the spillway water flow, is the Reynolds-averaged velocity component, representing the average velocity of the water flow in the j direction. Combining with the spillway flow field, it can be used to describe the vertical or lateral velocity components, is the local velocity change rate, reflecting the velocity gradient changing with time, and is used to capture the acceleration effect of the water flow in sudden spill events, is the convection term, representing the change rate of the water flow with position in space, and is used to calculate the velocity changes at different positions, such as the velocity distribution in the spillway bend area, ρ is the water body density, is the Reynolds-averaged pressure, representing the average pressure of the fluid, and is used to calculate the force exerted by the water flow on the spillway sidewall and the gate, is the pressure gradient force, representing the driving force generated by the water flow due to the pressure difference, and is used to analyze the water pressure distribution in the bend area and the gate section. v is the dynamic viscosity, representing the dynamic viscosity of water, taking 1.0×10 -6 m 2 / s, is the viscous dissipation term, representing the action of the viscous force inside the fluid, and is used to describe the boundary layer effect of the water flow near the spillway wall. τ ij is the Reynolds stress term, representing the additional stress caused by turbulence, and is used to model the turbulence intensity in the bend section. is the centrifugal force direction tensor, used to introduce the action of the centrifugal force. i represents the water flow direction. is the tangential direction of the water flow. When represents that the acting force is along the tangential direction of the bend. takes 1, otherwise takes 0;

[0104] (2) Output the global turbulence intensity I through the embedded solver t : where k2 is the turbulent kinetic energy, U ∞ is the characteristic flow velocity. The Reynolds stress term τ in the Reynolds-averaged N-S equation ij is closed and solved through the turbulence model to obtain the turbulent kinetic energy k2, which is the core input parameter of the turbulence intensity;

[0105] (3) Write K c , η, I t into the digital twin body property table and fuse with the geometric data of the hydraulic BIM model to generate a visual flow field atlas.

[0106] The hydraulic BIM model stores the geometric, physical, and engineering data of the spillway and related hydraulic structures, mainly including:

[0107] Geometric structure information (spillway shape, bend curvature, cross-sectional dimensions);

[0108] Hydraulic design parameters (design flood discharge, flow velocity distribution, energy loss);

[0109] Structural material information (density, elastic modulus, vibration characteristics of materials such as gates and channels);

[0110] Operation control data (gate opening history, spillway regulation rules);

[0111] Extract the bend curve through the spillway geometric modeling data to obtain the curvature hydraulic radius;

[0112] Obtain the design flood hydrograph from the flood discharge-time relationship curve stored in the hydraulic BIM model.

[0113] Extract the maximum opening parameter of the gate directly from the hydraulic BIM model, or read the opening data under different working conditions from the dispatching rule library.

[0114] The natural vibration period of the gate in the material property library is extracted from the material database of the hydraulic BIM model

[0115] These parameters can be extracted and automatically calculated through the API of BIM.

[0116] The improved Manning formula is used in the edge node to calculate the reference spillway discharge Q b :

[0117] Among them, the update of the dynamic roughness coefficient n dyn is expressed as:

[0118] Among them, n0 is the reference roughness in the hydraulic BIM model (extracted from the digital twin property library), I t is the current turbulence intensity output by the digital twin, η is the contraction rate of the flow cross-section, A represents the flow cross-sectional area, R is the hydraulic radius, S is the energy slope, representing the energy loss rate per unit length of the water flow, approximately calculated by the transverse water surface slope, R 2 / 3 represents the 2 / 3 power of the hydraulic radius, which is used to describe the dependence of the water flow on the channel geometry in the Manning formula, S 1 / 2 represents the square root of the energy slope, which is used to characterize the influence of the gravity driving force of the water flow on the flow rate in the Manning formula.

[0119] The long short-term memory network in the cloud server predicts the flow deviation value in the next 30 minutes. The input features of the long short-term memory network include:

[0120] The time series data of the reference spillway discharge in the previous 1 hour (sampling interval 5 minutes);

[0121] I output by the digital twin t , k c , η;

[0122] The rainfall intensity in the next 30 minutes predicted by the weather forecast;

[0123] The output of the long short-term memory network includes the flow deviation value ΔQ(t) in the next 30 minutes, with a time resolution of 5 minutes;

[0124] The architecture of the long short-term memory network is as follows:

[0125] 1. Input layer design:

[0126] Input features: Time series data of the reference spillway discharge; Normalize the data in the past 1 hour (12 5-minute slices); Dimension: 12×1;

[0127] Input features: Digital twin parameters; Concatenate the real-time values of [I t , K c , η] into a feature vector of 3×1;

[0128] Input features: Meteorological rainfall intensity; Normalize the 30-minute future forecast values (6 5-minute slices) to 6×1;

[0129] 2. Bidirectional LSTM layer, structural parameters:

[0130] Number of hidden units: 64 (32 for forward and backward respectively);

[0131] Activation function of the input gate: Hard sigmoid (to accelerate convergence);

[0132] Initialization of the forget gate bias: 1.0 (to alleviate gradient vanishing);

[0133] Calculation equation: h t = LSTM(x t , h t-1 ; W lstm ), where x t is the input feature at time t, and h t is the hidden state;

[0134] 3. Spatiotemporal attention layer: Dynamically focus on key time periods and features:

[0135] (1) Calculate feature energy:

[0136] (2) Generate attention weights:

[0137] where h i is the LSTM hidden state at the i-th time step, d j is the j-th digital twin parameter v a , W a , b a are all trainable parameters;

[0138] 4. Mixture Density Network (MDN) output layer:

[0139] Output structure: G = 3: G is the total number of Gaussian mixture components, g is the Gaussian mixture component, π g is the mixture weight, μ g , σ g are the mean and standard deviation

[0140] Physical constraint: Ensure μ g ≥ 0 (flow deviation is non-negative) through the ReLU activation function.

[0141] 5. Training strategy:

[0142] Training data construction: Historical flood event database (>100 events) and data twin high-fidelity simulation dataset (covering various flow state combinations);

[0143] The loss function uses negative log-likelihood loss;

[0144] Model validation: The MAE prediction performance metric is used.

[0145] The edge node executes every 5 minutes:

[0146] Receive the ΔQ(t) sequence sent from the cloud;

[0147] Align the benchmark spillway discharge Q b (t) according to the timestamp;

[0148] Generate the calibrated real-time spillway discharge: Q cal (t) = Q b (t) + ΔQ(t)·sigmoid(t / 30), where sigmoid(t / 30) is the time weight function, giving higher confidence to proximal predictions, making the prediction weight from 0.82 in the next 5 minutes to 0.12 in 30 minutes, which conforms to the physical law that the uncertainty of hydrological prediction increases with time.

[0149] The spillway threshold in the risk adaptive warning module is expressed as Q th , when the calibrated real-time spillway discharge Q cal ≥Q th , a warning is triggered. Among them, Q design is the designed flood discharge capacity in the hydraulic BIM model, β is the working condition adjustment factor (taking 1.2 in the flood season and 0.9 in the non-flood season), and I t,max is the maximum allowable turbulent intensity in the material durability library (taking 25% for concrete structures).

[0150] The risk levels are divided according to the turbulent intensity of the digital twin, specifically including:

[0151] I t <12%, set as level I;

[0152] 12% ≤ I t <18%, set as level II;

[0153] I t >18%, set as level III.

[0154] The trigger gate opening control instructions include:

[0155] Input the current risk level;

[0156] Output gate priority regulation sequence:

[0157] If the risk level = Level I: Maintain the current situation and increase the monitoring frequency;

[0158] If the risk level = Level II: Prioritize opening the gate farthest from the bend (reduce the impact of secondary flow);

[0159] If the risk level = Level III: Alternately open the gates to form staggered flood discharge (interval ≤ 30 seconds).

[0160] Encrypt and package the instructions into a message through the industrial Ethernet protocol and push it to the dispatching terminal in real time.

[0161] The present invention covers any substitutions, modifications, equivalent methods, and solutions made to the essence and scope of the present invention. In order to enable the public to have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention. However, those skilled in the art can fully understand the present invention without these detailed descriptions. Additionally, well-known methods, processes, procedures, components, and circuits are not described in detail to avoid unnecessary confusion to the essence of the present invention.

[0162] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. Intelligent reservoir spillway water level monitoring and flood overflow warning system, characterized by: include: Anti-interference water level sensing module: anti-siltation sensor arrays are deployed in key sections of the spillway, including the bend section and the gate section, where: The curved section uses a dual-frequency millimeter-wave radar to scan the lateral slope of the water surface; The gate section deploys a pressure-ultrasonic fusion sensor group to synchronously obtain the absolute water level and gate vibration displacement; Dynamic hydraulic twin module: Based on the water surface lateral slope, absolute water level, gate vibration displacement and combined with the hydraulic BIM model, the following are processed through a lightweight fluid network algorithm: ① Calculate the centrifugal force correction coefficient of the curve based on the transverse slope; ②Convert the vibration displacement into the dynamic contraction rate of the flow section; ③ Generate a digital twin including turbulence intensity every 5 minutes; Edge-cloud collaborative traffic solution module: includes edge nodes and cloud servers, including: Edge node execution: The improved Manning formula is used to calculate the benchmark flood spillway, where the roughness coefficient is dynamically updated by the digital twin; Cloud server execution: predict the flow deviation value in the future through the long short-term memory network and transmit it back to the edge node for calibration. The edge node calibration includes superimposing the flow deviation value to the reference overflow volume to generate the calibrated real-time overflow volume; Risk adaptive early warning module: When the calibrated real-time flood volume exceeds the flood threshold, the following is performed: ① Classify the risk level according to the turbulence intensity of the digital twin; ②, trigger the gate opening control instruction based on the risk level and push it to the reservoir dispatching terminal simultaneously.

2. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 1 is characterized in that: The layout of the anti-stagnation sensor array includes: A dual-frequency millimeter-wave radar group is arranged along the curve section. Among them, the 24GHz radar array scans the horizontal slope of the water surface at an elevation angle of 15°, and the slope angle is calculated by phase difference; the 77GHz radar captures the micro-deformation of the water surface at vertical incidence, and the dual-frequency data is fused by Kalman filtering to generate the horizontal slope value α of the water surface that is resistant to haze interference. fused ; A pressure-ultrasonic fusion sensor group is installed on the water-facing surface of the gate section. The pressure sensor array is distributed according to isobars, and MEMS piezoresistive probes are arranged every 50 cm to measure the absolute water pressure value and calculate the water level. The ultrasonic transmitter is embedded in the center of the pressure sensor and emits a 1MHz pulse wave through a 10° deflection angle. The gate vibration displacement Δd is calculated based on the echo time delay difference.

3. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 1 is characterized in that: The dynamic hydraulic twin module specifically includes: Centrifugal force correction coefficient calculation unit: Based on the transverse slope angle of the water surface in the curved section and the hydraulic radius R in the hydraulic BIM model, the time decay factor is introduced and the centrifugal force correction coefficient K is calculated through the lightweight fluid network algorithm c ; Dynamic contraction rate conversion unit: Combines the vibration main frequency and the gate design opening to map the gate vibration displacement Δd into the flow section contraction rate η; Digital twin generation unit: The centrifugal force correction coefficient and the flow section shrinkage rate are combined and input into the lightweight fluid network for numerical solution to obtain the flow field distribution. The lightweight fluid network calculates the flow velocity changes in the bend and gate areas based on the Reynolds-averaged NS equation, and quantifies the turbulence intensity. After the calculation is completed, the key parameters are written into the attribute table of the digital twin.

4. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 3 is characterized in that: In the digital twin generation unit, the following operations are performed every 5 minutes: (1) K c ,η Input lightweight fluid network and solve the Reynolds averaged NS equation: Among them, τ ij is the Reynolds stress term, δ iθ is the centrifugal force direction tensor, is the Reynolds mean velocity component, representing the average velocity of the water flow in the i direction, is the Reynolds mean velocity component, representing the average velocity of water flow in the j direction. Combined with the spillway flow field, is the local flow velocity change rate, is the convection term, ρ is the water density, is the Reynolds mean pressure, representing the average pressure of the fluid, is the pressure gradient force, v is the dynamic viscosity, representing the dynamic viscosity of water, is the viscous dissipation term, τ ij is the Reynolds stress term, representing the additional stress caused by turbulence; (2) Output the global turbulence intensity I through the embedded solver t ; (3) K c ,η,I t Write the digital twin attribute table and merge it with the geometric data of the hydraulic BIM model to generate a visual flow field map.

5. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 3 is characterized in that: The edge node uses the improved Manning formula to calculate the benchmark flood volume Q b : Among them, the dynamic roughness coefficient n dyn The update is expressed as: Among them, n0 is the benchmark roughness in the hydraulic BIM model, I t is the current turbulence intensity output by the digital twin, η is the flow section contraction rate, A represents the flow section area, R is the hydraulic radius, S is the energy slope, which represents the energy loss rate per unit length of the water flow, which is approximately calculated by the lateral slope of the water surface, and R 23 It represents the 2 / 3 power of the hydraulic radius and is used in the Manning formula to describe the dependence of water flow on channel geometry. 12 It represents the square root of the energy slope and is used in the Manning equation to characterize the effect of the gravity driving force of the water flow on the flow rate.

6. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 5 is characterized in that: The long short-term memory network in the cloud server predicts the traffic deviation value in the next 30 minutes. The input features of the long short-term memory network include: Time series data of the benchmark flood overflow in the previous hour; Digital Twin Output I t ,K c ,η; Weather forecast of rainfall intensity in the next 30 minutes; The output of the LSTM network includes the flow deviation value ΔQ(t) for the next 30 minutes, with a time resolution of 5 minutes.

7. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 6 is characterized in that: The edge node executes every 5 minutes: Receive the ΔQ(t) sequence sent by the cloud; Aligning the baseline overflow volume Q by timestamp b (t); Generate calibrated real-time flood volume: Q cal (t) = Q b (t)+ΔQ(t)·sigmoid(t / 30), where sigmoid(t / 30) is the time weight function, which gives higher confidence to the near-end prediction, making the prediction weight of the next 5 minutes 0.82→0.12 for 30 minutes, which is in line with the physical law that the uncertainty of hydrological prediction increases with time.

8. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 7 is characterized in that: The flood threshold in the risk adaptive warning module is represented by Q th , when the real-time overflow volume Q after calibration cal ≥Q th When the warning is triggered, Among them, Q design is the designed flood discharge capacity in the hydraulic BIM model, β is the working condition adjustment factor, I t,max is the maximum turbulence intensity allowed in the material durability library.

9. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 4 is characterized in that: The risk level classification according to the turbulence intensity of the digital twin specifically includes: I t When <12%, it is set as Level I; 12%≤I t When <18%, it is set as Grade II; I t When >18%, it is set as Grade III.

10. The intelligent reservoir spillway water level monitoring and flood overflow warning system according to claim 1 is characterized in that: The gate opening control instruction is as follows: Enter the current risk level; Output gate priority control sequence: If the risk level = Level I: Maintain the status quo and increase the monitoring frequency; If the risk level = Level II: open the gate farthest from the curve first; If the risk level = Level III: open the gates alternately to form staggered flow discharge.

Images