Urban inland inundation emergency roadblock automatic control system and method based on water level induction

Abstract

The invention relates to the technical field of urban flood control and emergency management, and discloses an automatic control system and method for an urban inland inundation emergency roadblock based on water level induction, and the method comprises the steps: constructing an urban hydrological digital twinborn model, integrating the urban geographic data and the hydrological parameters to form a digital model; constructing a network fluid dynamics solver of a graph structure based on a digital twin model, and abstracting an urban road network and a water flow channel into a flow network of a node edge structure; establishing a roadblock control influence propagation model, and analyzing the influence of different control strategies on water flow distribution; designing a multi-step prediction optimization controller, and calculating an optimal roadblock control sequence according to a prediction result; according to the method, the calculation complexity is reduced, rapid response and active prevention of waterlogging and ponding are realized, the water flow evolution trend can be predicted in advance, the control strategy of a plurality of roadblock nodes can be coordinated, and the chain reaction problem caused by local control is effectively avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban flood control and emergency management, more specifically, it relates to an urban waterlogging emergency roadblock automatic control system and method based on water level sensing. BACKGROUND

[0002] With the acceleration of urbanization and the frequent occurrence of extreme weather events, urban waterlogging has become an important problem affecting the safe operation of cities. Traditional urban waterlogging prevention and control mainly relies on drainage systems and simple road control measures, but it is often difficult to respond in time when facing sudden heavy rainfall, which can easily cause loss of life and property. Especially in low-lying areas of cities, underpasses and other waterlogging points, how to achieve rapid sensing and intelligent control of accumulated water has become a technical problem that needs to be solved.

[0003] In recent years, with the development of Internet of Things technology and artificial intelligence, smart city construction has provided new technical means for urban waterlogging prevention and control. By deploying water level sensors, intelligent roadblocks and other equipment at key locations in the city, combined with weather forecasts and traffic monitoring data, real-time monitoring of urban waterlogging risks can be achieved. However, due to the complexity of urban hydrological systems, how to accurately predict water flow trends and make optimal control decisions remains an important challenge in this field.

[0004] In practical applications, urban waterlogging prevention and control also need to consider traffic evacuation needs, which makes roadblock control decisions more complex. Traditional control methods based on experience rules are difficult to balance flood control and traffic needs, often leading to local optimal but globally suboptimal results. Therefore, developing an intelligent roadblock system that can consider hydrodynamic characteristics and traffic flow characteristics and achieve predictive control has important practical significance. SUMMARY

[0005] The present application provides an urban waterlogging emergency roadblock automatic control system and method based on water level sensing, which solves the technical problems of independent control lacking system, passive response delay, lack of global optimization decision, high computational complexity and poor adaptability of static control strategy in related technologies.

[0006] The present application provides an urban waterlogging emergency roadblock automatic control method based on water level sensing, comprising the following steps: Constructing a digital twin model of urban hydrology, integrating city geographic data and hydrological parameters to form a digital model; Based on the digital twin model, a network fluid dynamics solver of graph structure is constructed, the city road network and water flow channel are abstracted as a flow network of node edge structure, and the shallow water equation set is converted into a discrete form based on graph structure by using an improved finite volume method; Based on the calculation results of the network fluid dynamics solver, a roadblock control influence propagation model is constructed to quantitatively analyze the influence of roadblock control operation on water flow distribution, and a mathematical correlation between roadblock state and water flow change is established. Based on the roadblock control influence propagation model, a multi-step predictive optimal controller is realized, and an optimal roadblock control strategy is calculated by using the model predictive control theory to realize active prevention and dynamic regulation of waterlogging.

[0007] In a preferred embodiment, the step of constructing a digital twin model of urban hydrology comprises: Obtaining high-precision digital elevation model data, including urban surface elevation, road geometry, and drainage ditch location; Collecting urban drainage network data and constructing a drainage system topology graph structure to establish a connection relationship with the surface flow model; Labeling all controllable roadblock node positions and their control parameters in the digital model; Collecting historical waterlogging event data and calibrating model parameters using a parameter optimization algorithm.

[0008] In a preferred embodiment, the step of constructing a network fluid dynamics solver based on a graph structure comprises: Abstracting the urban road network and water flow channel into a flow network with node and edge structure; Using an improved finite volume method to convert the traditional two-dimensional shallow water equation set into a discrete form based on the graph structure; Adjusting the calculation precision dynamically for high-risk waterlogging areas or complex water flow areas, using fine-grained calculation in key areas and coarse-grained calculation in non-key areas; Constructing a parallel computing framework suitable for graph structure network flow calculation to decompose large-scale fluid calculation tasks into parallel processing subtasks.

[0009] In a preferred embodiment, the water flow state change rate function in the network fluid dynamics solver describes the change rate of the water flow state, considering the water level gradient, gravity, ground friction, and the influence of roadblock blocking on water flow. The calculation of water flow uses an improved Manning formula, which multiplies the effective cross-sectional area of the edge by the square root of two-thirds of the hydraulic radius, then multiplies the absolute value of the water level height difference, divides by the product of the Manning roughness coefficient and the square root of the edge length, and finally multiplies the complement value of the roadblock blocking factor.

[0010] In a preferred embodiment, the step of constructing a roadblock control influence propagation model comprises: For each roadblock control point, the influence coefficient of water level, flow velocity, and other parameters on adjacent areas is calculated by simulating water flow changes under different control states; Analyzing the synergistic effect of simultaneous or sequential operation of multiple barrier control points, identifying control combinations that reinforce or offset each other; Computing the sensitivity of the change of barrier control vector to the water flow state vector, forming a sensitivity matrix; Based on the sensitivity matrix, identifying key control points and optimal control sequences, and determining the combination of barrier operations that can achieve the target water flow regulation effect with the minimum control cost through the minimum spanning tree algorithm in graph theory.

[0011] In a preferred embodiment, the method for generating the sensitivity matrix includes: Apply a small perturbation to each barrier control parameter, calculate the resulting change in water flow state parameters, and take the partial derivative as the sensitivity; Analyzing the change characteristics of sensitivity over time, constructing a dynamic sensitivity matrix for predicting the time delay effect of barrier control; Setting a sensitivity threshold to identify high-sensitivity control relationships and simplify the complexity of the control model.

[0012] In a preferred embodiment, the step of implementing a multi-step predictive optimal controller includes: Establishing a comprehensive optimization function containing water safety targets and traffic evacuation targets, the comprehensive optimization function representing the integral of system loss from the initial time to the termination time, wherein the loss function considers the waterlogging risk, traffic disruption cost and control operation cost; Based on the current water flow state and weather forecast, predicting the water flow evolution trend in the prediction time domain, and calculating the optimal control sequence for multiple time steps; Decomposing the global optimization problem into multiple local sub-problems and solving them using a distributed optimization algorithm; Based on the water flow prediction results and traffic demand, planning dynamically changing emergency evacuation channels to optimize traffic evacuation efficiency under the premise of ensuring safety.

[0013] In a preferred embodiment, the specific definition of the loss function is: The total system loss is equal to the water safety loss multiplied by the first weight coefficient, plus the traffic disruption loss multiplied by the second weight coefficient, plus the control operation loss multiplied by the third weight coefficient, wherein: the water safety loss function calculates the square value of the part of each node in the key monitoring node set that exceeds the safe water level threshold, multiplied by the risk weight of the node, and then summed up; The traffic disruption loss function calculates the ratio of the current traffic capacity to the maximum traffic capacity for each road edge in the key road edge set, subtracts the ratio from 1 to get the traffic capacity loss rate, multiplies it by the traffic importance weight of the road, and then sums up; The control operation loss function includes the conversion cost of control operation and the traffic disruption cost caused by barrier closure.

[0014] In a preferred embodiment, the step of predicting the time domain rolling optimization comprises: At time point t, the water flow state and traffic state within N time steps in the future are predicted using a network fluid dynamics model; Based on the state prediction result, a finite time domain optimization problem is solved to obtain an optimal control sequence; Only the first control action of the optimal sequence is executed, and then the new system state is obtained at the next time point, and the prediction and optimization process is repeated.

[0015] In a preferred embodiment, the water level sensing based urban waterlogging emergency roadblock automatic control system is used to execute the water level sensing based urban waterlogging emergency roadblock automatic control method, comprising: A hydrological monitoring module for collecting water level sensor data, traffic flow data, and meteorological monitoring data in the city; A digital twin module for building and maintaining a city hydrological digital twin model; A network fluid dynamics calculation module for performing water flow simulation calculation based on graph structure; An influence propagation analysis module for analyzing the influence of roadblock control on water flow distribution; A prediction and optimization control module for calculating the optimal roadblock control strategy; A roadblock control execution module for sending control instructions to each roadblock control point; An emergency evacuation planning module for planning dynamically changing emergency evacuation channels.

[0016] The beneficial effects of the present application are: By constructing a city hydrological digital twin model and a graph structure network fluid dynamics solver, the computational complexity is reduced, so that complex fluid dynamics calculations can be performed in real time on edge devices, thereby achieving rapid response and active prevention of waterlogging.

[0017] Through the design of the roadblock control influence propagation model and the multi-step prediction and optimization controller, the present application realizes systematic and forward-looking control of urban waterlogging. Compared with the traditional single-point trigger control scheme, the present application can predict the water flow evolution trend in advance, coordinate the control strategies of multiple roadblock nodes, effectively avoid the chain reaction problem caused by local control, and improve the waterlogging prevention and control effect.

[0018] Based on adaptive grid optimization technology and parallel computing framework, the present application realizes efficient utilization of computing resources while ensuring calculation accuracy. By using differentiated processing strategies for fine calculation in high-risk areas and rough calculation in low-risk areas, the control accuracy of key areas is ensured, and the overall calculation load is reduced, making the system more practical.

[0019] The application establishes a dynamic optimization decision mechanism of hydrological traffic coupling by combining digital twinning technology with predictive control, can simultaneously consider flood control effect and traffic evacuation demand, provides an intelligent solution for urban flood control and disaster reduction, and has remarkable social benefits and practical application value. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flow chart of the urban waterlogging emergency roadblock automatic control method based on water level sensing of the application; Figure 2 is an area graph of water level change comparison of key areas in a rainstorm event of the application; Figure 3 is a line graph of traffic capacity change of main roads in a rainstorm event of the application; Figure 4 is a column graph of performance comparison of a graph structure fluid model and a traditional fluid model of the application; Figure 5 is a network graph of roadblock control influence propagation network graph of the application; Figure 6 is a radar chart of performance comparison of a predictive control system and a traditional control system of the application; Figure 7 is a Sankey chart of urban waterlogging flow path and roadblock control flow direction chart of the application. DETAILED DESCRIPTION

[0021] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is merely meant to provide a better understanding of the subject matter described herein and can include changes to the function and arrangement of the elements discussed without departing from the scope of the content of this specification. Various examples can omit, substitute, or add various procedures or components as desired. Additionally, features described in relation to one example can be combined in other examples.

[0022] An urban waterlogging emergency roadblock automatic control method based on water level sensing is disclosed in at least one embodiment of the application, as shown in Figure 1 includes the following steps: Step 1, build a city hydrological digital twinning model, integrate city geographic data and hydrological parameters to form a digital model; Specifically includes the following sub-steps: Step 1.1, city terrain data acquisition and processing; High-precision digital elevation model data is acquired, including city surface elevation, road geometry, drainage ditch location, etc. information, and the original data is converted into a regular grid representation through a geographic information processing algorithm to generate a terrain surface model suitable for fluid calculation.

[0023] Optionally, in some embodiments, laser radar scanning data can also be integrated to improve the accuracy of the terrain model, especially for the depiction of urban complex buildings and micro-topography.

[0024] Step 1.2, drainage system topology construction; Collect urban drainage network data, including pipe location, diameter, slope, pump station capacity and other parameters, construct drainage system topology structure, and establish connection relationship with surface flow model to form a complete drainage system model coupled with surface pipe network.

[0025] In some embodiments, when the drainage system data is incomplete, a parameterized method based on urban planning standards can be used to generate a theoretical drainage network, and local calibration can be performed through measured data.

[0026] As shown in Figure 2 , it shows the influence of water level change in key areas under the same rainstorm intensity between traditional control method and predictive control method proposed in this patent. From the chart, it can be seen that the predictive control system can control the water level peak at a lower level, and the overall water level curve is more gentle, indicating that the system has a significant predictive control effect, which can predict the water accumulation trend 15 to 30 minutes in advance and take preventive measures.

[0027] Step 1.3, roadblock control node labeling; Label all controllable roadblock node positions and their control parameters in the digital model, including roadblock type, switch state influence range, operation response time, etc., and establish the mapping relationship between roadblock control nodes and road network.

[0028] Roadblock control nodes can include various types, such as liftable roadblock barriers, electronic warning signs, controllable traffic lights, etc., and the method of the present application is applicable to different types of roadblock control devices.

[0029] Step 1.4, historical data fusion calibration; Collect historical waterlogging event data, including rainfall, water accumulation depth, water accumulation range, etc. records, use parameter optimization algorithm to calibrate model parameters, make model output match historical observation data to the greatest extent, and improve model prediction accuracy.

[0030] In addition, in the model calibration process, Bayesian optimization method can be used to automatically adjust key parameters such as surface roughness and infiltration rate to minimize the error between model prediction and historical observation.

[0031] Step 2, based on the digital twin model, construct a graph-structured network fluid dynamics solver, abstract the urban road network and water flow channel into a flow network with node-edge structure, and use the improved finite volume method to convert the shallow water equation set into a discrete form based on the graph structure; Specifically, the following sub-steps are included: Step 2.1, node-edge flow network construction; Abstract the urban road network and water flow channel into a flow network with node-edge structure, where nodes represent road intersections or key topographic feature points, and edges represent road segments or water flow channels, to generate a topological structure reflecting the coupling relationship between urban water flow and traffic.

[0032] According to one embodiment of the present application, a multi-resolution grid structure can be used, with more detailed node division in key areas to improve calculation accuracy.

[0033] Step 2.2, discrete shallow water equation based on graph structure; The improved finite volume method is used to convert the traditional two-dimensional shallow water equation set into a discrete form based on the graph structure. The traditional two-dimensional shallow water equation set describes the relationship between the rate of change of the water flow state vector over time and the traffic state vector and barrier control vector, where the water flow state vector represents the water level height at the nodes of the graph and the flow rate at the edges of the graph, the traffic state vector represents the road capacity, and the barrier control vector represents the blocking state of the nodes.

[0034] In this graph structure discretization process, the specific implementation of the water flow state change rate function and the traffic state change rate function is as follows: The water flow state change rate function describes the rate of change of the water flow state, taking into account the water level gradient, gravity, ground friction, and the influence of barrier blocking on water flow.

[0035] On the edge between node i and node j, the water flow rate is calculated using the improved Manning formula: first determine the water flow direction by comparing the positive and negative signs of the water level height difference between node i and node j; then calculate the actual flow size, multiply the effective cross-sectional area of the edge by the square of two-thirds of the hydraulic radius , multiply by the square root of the absolute value of the water level height difference, and divide by the product of the Manning roughness coefficient and the square root of the edge length; finally, multiply the result by the complement of the barrier blocking factor (1 minus the barrier blocking factor), where a barrier blocking factor of 0 indicates complete blocking and a barrier blocking factor of 1 indicates complete opening.

[0036] The traffic state change rate function describes the change rate of the traffic state, mainly considering the influence of water level on traffic capacity. The traffic capacity at node i is dynamically adjusted as follows: first, the maximum traffic capacity value under non-flooding conditions is obtained, then the ratio of the current water level height to the critical water level height is calculated, the ratio is subtracted from 1 to obtain a traffic capacity attenuation coefficient, and the attenuation coefficient is ensured to be not less than 0, and finally the maximum traffic capacity is multiplied by the attenuation coefficient to obtain the current actual traffic capacity. The critical water level height is usually set to 25 cm, that is, the water limit of ordinary vehicles.

[0037] In some embodiments, the above equation can be adjusted according to specific application scenarios. For example, an additional correction term can be introduced to consider the influence of rainfall intensity on surface roughness, or to consider the difference in water crossing ability of different vehicle types.

[0038] Step 2.3, adaptive grid optimization technology implementation; For areas at high risk of waterlogging or complex water flow areas, dynamically adjust the calculation precision, use fine-grained calculation in key areas and coarse-grained calculation in non-key areas, form a calculation resource adaptive allocation mechanism, and reduce the overall calculation load while ensuring accuracy.

[0039] The specific implementation of the adaptive grid optimization technology includes: Risk assessment function definition: calculate the risk index of each region according to the terrain characteristics, historical waterlogging records and current water flow state .

[0040] The calculation method of the risk index is as follows: first, the current water level is multiplied by the first weight coefficient , then the water level change rate is multiplied by the second weight coefficient , the historical highest water level normalized value is multiplied by the third weight coefficient , and finally the terrain elevation normalized value (low-lying area value close to 0) is negated (i.e. subtracted from the value) and multiplied by the fourth weight coefficient , the sum of the four results is the final regional risk index.

[0041] This calculation method comprehensively considers current waterlogging conditions, waterlogging development trend, historical waterlogging degree and terrain characteristics, etc. It can comprehensively evaluate the waterlogging risk in the region.

[0042] Grid division strategy: Based on the risk index, the urban area is divided into high, medium, and low precision regions. High-risk areas use fine-grained grids (e.g., 5-meter intervals), medium-risk areas use medium-precision grids (e.g., 20-meter intervals), and low-risk areas use coarse-grained grids (e.g., 50-meter intervals).

[0043] In the case of sufficient computing resources, the threshold of grid precision can be dynamically adjusted to further improve the calculation accuracy of critical areas.

[0044] Grid boundary processing: At the junction of different precision grids, interpolation algorithms are used to ensure the continuity of the calculation results and avoid numerical instability caused by sudden changes in grid precision.

[0045] Step 2.4, parallel computing framework deployment; Based on the characteristics of edge computing hardware, a parallel computing framework suitable for graph structure network flow calculation is constructed. Large-scale fluid calculation tasks are decomposed into parallel processing subtasks, making full use of multi-core processing capabilities to achieve near real-time water flow state updates.

[0046] The implementation of the parallel computing framework includes: Region decomposition strategy: The entire calculation area is divided into multiple relatively independent sub-regions using the geographical partitioning method, and each sub-region is assigned to a computing core.

[0047] Boundary data exchange mechanism: Adjacent sub-regions exchange boundary node data through shared memory to minimize communication overhead.

[0048] In addition, in some embodiments, an asynchronous communication mechanism can be used to allow each sub-region to calculate at different step lengths, further improving calculation efficiency.

[0049] Load balancing algorithm: According to the calculation complexity of each sub-region (related to the number of grids and water flow activity), dynamically adjust task allocation to ensure balanced load of each computing core.

[0050] Real-time guarantee mechanism: Introduce calculation timeout control. When the calculation complexity of some areas suddenly increases, causing the real-time requirement to be unable to be met, automatically reduce the calculation precision of the area to ensure that the response time of the overall system does not exceed the preset threshold (usually 1 second).

[0051] In practical applications, the network fluid dynamics solver can run on an edge computing device equipped with a multi-core processor, handle water flow simulation tasks covering an area of about 10 square kilometers of urban area, and have a calculation update frequency of up to one per second, meeting the real-time requirements of waterlogging early warning and roadblock control. For example, in the rainstorm waterlogging scene test of a certain urban low-lying area, the solver can perform real-time simulation on a city water flow network containing 1000 nodes and 2500 edges on a general industrial computer (8-core CPU, 16GB memory), and the calculation result has an error of less than 5% compared with the traditional two-dimensional hydrodynamic model, and the calculation speed is increased by about 20 times.

[0052] As Figure 3 shown, the effects of the traditional control system and the predictive control system on the traffic capacity of the main road during the rainstorm event are compared. The traffic capacity of the traditional control system decreases to 35% after 60 minutes, while the predictive control system can still maintain 75% of the traffic capacity, proving that the system of the present patent can better maintain the traffic capacity of important roads while ensuring water safety and reducing the interference with traffic.

[0053] Step 3, using the calculation results of the network fluid dynamics solver, a roadblock control influence propagation model is constructed to quantitatively analyze the influence of roadblock control operations on water flow distribution, and a mathematical correlation between roadblock state and water flow change is established; Specifically, the following sub-steps are included: Step 3.1, single-point roadblock control influence analysis; For each roadblock control point, by simulating the water flow changes under different control states (open / close), the influence coefficients of its adjacent area water level, flow velocity and other parameters are calculated, and a local influence model of single-point roadblock control is formed.

[0054] The specific construction method of the single-point influence model includes: Control experiment design: for each roadblock control point , simulate the water flow distribution under the fully open state and the fully closed state in the network fluid dynamics model, keeping all other parameters unchanged.

[0055] In some embodiments, the intermediate state (such as the half-open state) of the roadblock can also be simulated to obtain more detailed control response characteristics.

[0056] Influence range determination: by comparing the water level changes of each node under the two states , the effective influence range of the roadblock control point is determined , defined as the set of all nodes whose water level changes exceed the threshold (usually 1 centimeter): ; wherein, denotes the effective influence range set of the roadblock control point ; denotes the affected node number; denotes the water level change amount of the node caused by the state change of the roadblock control point ; denotes the threshold value of water level change.

[0057] Influence coefficient calculation: for each node in the influence range, calculate the influence coefficient , which denotes the water level change rate of the node caused by the state change of the roadblock control point ; ; wherein, denotes the influence coefficient of the roadblock control point on the node ; denotes the water level change amount of the node caused by the state change of the roadblock control point ; denotes the control state change amount of the roadblock control point , with a value range of [0, 1], where 0 represents complete closing and 1 represents complete opening, and the change amount is 1 in the case of complete on-off switching.

[0058] It should be noted that the above influence coefficient calculation can further consider the time factor to form a time-varying influence coefficient to depict the time delay effect of roadblock control.

[0059] Step 3.2, modeling of multi-point roadblock synergistic effect; Based on the single-point influence model, analyze the synergistic effect generated by simultaneous or sequential operation of multiple roadblock control points, identify mutually enhancing or offsetting control combinations, and build a global influence model of roadblock control.

[0060] The specific implementation method of synergistic effect modeling includes: Roadblock control combination experiment: select different control combinations of key roadblock points for simulation to obtain water flow distribution results under combined control.

[0061] In addition, in some embodiments, orthogonal experimental design method can be used to reduce the number of simulations and improve modeling efficiency.

[0062] Synergistic effect quantification: compare the difference between the combined control effect and the superposition of single-point control effects, and calculate the synergistic coefficient : ; in, Indicates a roadblock control point and The synergy coefficient between It's a roadblock and When operating simultaneously, nodes water level changes, and Roadblock Control Points and Node when operating independently water level changes.

[0063] when When , it indicates a synergistic enhancement effect, that is, the water level change caused by the simultaneous operation of the two roadblocks is greater than the simple addition of their individual operation effects; when When , it indicates a synergistic offsetting effect, that is, the water level change caused by the simultaneous operation of the two roadblocks is less than the simple addition of their individual operating effects.

[0064] Collaboration matrix construction: organize the coordination coefficients of all roadblock point pairs into a coordination matrix , which is used to guide the formulation of collaborative control strategies for multi-point roadblocks.

[0065] Step 3.3, sensitivity matrix generation; Calculate the roadblock control vector using the water flow simulation results The change of water flow state vector The sensitivity of the impact is formed into a sensitivity matrix , where the matrix elements It indicates the degree of influence of the change of the j-th roadblock control parameter on the i-th water flow state parameter.

[0066] The methods for generating the sensitivity matrix include: Small perturbation analysis method: for each roadblock control parameter Apply a small perturbation , calculate the flow state parameters caused by change , and take the partial derivative as the sensitivity: ; in, Indicates the sensitivity matrix Rank The elements of the column reflect the The roadblock control parameters are The degree of influence of each water flow state parameter; Indicates the a water flow state parameter, such as water level, flow rate, etc. a control parameter representing the roadblock, with a value range of [0, 1]; a partial derivative of the water flow state parameter with respect to the roadblock control parameter . a small change in the water flow state parameter . a small change in the roadblock control parameter .

[0067] Time evolution sensitivity: analyze the change characteristics of sensitivity over time, construct a dynamic sensitivity matrix , used to predict the time delay effect of roadblock control.

[0068] Sensitivity threshold screening: set a sensitivity threshold , identify high sensitivity control relationships, and simplify the complexity of the control model.

[0069] In practical applications, the sensitivity matrix may be a high-dimensional sparse matrix, so sparse matrix storage and calculation techniques can be used to improve computational efficiency.

[0070] Step 3.4, control propagation path optimization; Based on the sensitivity matrix, identify key control points and optimal control sequences, and use the minimum spanning tree algorithm in graph theory to determine the combination of roadblock operations that can achieve the target water flow regulation effect with the minimum control cost.

[0071] The specific algorithm implementation of control propagation path optimization includes: Control graph construction: taking roadblock nodes as vertices and sensitivity as edge weights, construct a control influence propagation graph .

[0072] Key path identification: use an improved Prim algorithm to calculate the minimum spanning tree , which covers all target control areas while minimizing the total control cost (e.g., the number of roadblock operations).

[0073] In some embodiments, multi-objective optimization algorithms can be introduced, considering multiple objectives such as control cost and response time.

[0074] Control sequence optimization: based on the structural characteristics of the tree, determine the optimal execution order of the roadblock control, and preferentially control roadblock nodes with large influence range and low execution cost.

[0075] Feedback adjustment mechanism: combine real-time water flow monitoring data to dynamically adjust the sensitivity matrix and control strategy, adapting to real-time changes in water flow state.

[0076] In practical applications, the roadblock control influence propagation model can reveal the complex interaction between water flow and roadblocks in urban waterlogging. For example, in a certain city's rainstorm waterlogging emergency drill, the model successfully identified a key roadblock point located in a higher terrain but with significant control effect. By prioritizing the control of this point, the water flow distribution of the downstream five areas can be influenced. The model also found strong synergistic effects between certain roadblock combinations, and through coordinated control, temporary "drainage channels" can be formed, effectively reducing the risk of waterlogging in key areas.

[0077] As Figure 4 shown, the performance differences of the graph structure fluid model proposed in this patent and the traditional fluid model in three key indicators of calculation time, memory occupation, and calculation accuracy error are compared. The graph structure fluid model only takes 5% of the calculation time of the traditional model, and the memory occupation is only 12% of the traditional model. The calculation accuracy error only increases by 2 percentage points (from 3% to 5%), verifying that the system in this patent reduces the computational complexity by converting traditional fluid mechanics calculations into network flow calculations of graph structures.

[0078] Step 4, based on the roadblock control influence propagation model, a multi-step prediction and optimization controller is implemented to calculate the optimal roadblock control strategy using model predictive control theory, achieving active prevention and dynamic regulation of waterlogging. Specifically, the following sub-steps are included: Step 4.1, multi-objective optimization function construction; An integrated optimization function is established that includes water safety objectives and traffic evacuation objectives. This function represents the system loss integral from the initial time to the termination time , where the loss function considers the waterlogging risk, traffic disruption cost, and control operation cost to quantify the overall system performance.

[0079] The specific definition of the loss function is: the total system loss is equal to the water safety loss multiplied by the first weight coefficient , plus the traffic disruption loss multiplied by the second weight coefficient , and plus the control operation loss multiplied by the third weight coefficient . These three weight coefficients are used to balance the importance of different objectives. The definitions of each sub-loss function are as follows: Water safety loss function : First, determine the set of key monitoring nodes , and for each node in the set, calculate its current water level ​Exceeding the safety water level threshold The square value of the part exceeding the safety threshold, multiplied by the risk weight of the node (relevant to population density, distribution of important facilities, etc.), and finally summing the weighted over-limit water level square values of all nodes to get the total water safety loss. This calculation method imposes a quadratic penalty on water levels exceeding the safety threshold, effectively guiding the system to prioritize controlling waterlogging in high-risk areas.

[0080] The expression is: ; Where, represents the total value of the water safety loss function; represents the set of key monitoring nodes; represents a certain node in the set; represents the risk weight of node , related to factors such as population density, distribution of important facilities, etc. represents the current water level of node ; represents the safety water level threshold of node ; represents the part of the current water level exceeding the safety threshold, taking 0 if not exceeding.

[0081] Traffic disruption loss function : First determine the set of key road edges , for each road edge in the set, calculate the ratio of its current traffic capacity to the maximum traffic capacity , subtract 1 from the ratio to get the traffic capacity loss rate, multiply it by the traffic importance weight of the road, and finally sum the weighted traffic capacity loss rates of all roads to get the total traffic disruption loss. This calculation method can guide the system to maintain the traffic capacity of important roads as much as possible while ensuring water safety.

[0082] The expression is: ; Where, represents the total value of the traffic disruption loss function; represents the set of key road edges; represents a certain road edge in the set; represents the traffic importance weight of road e, related to factors such as road grade, traffic volume, etc. represents the current traffic capacity of road e; represents the maximum traffic capacity of road e; The loss rate of the road e.

[0083] Control operation loss function The function consists of two parts. The first part represents the switching cost of control operation (penalty term for frequent switching), which is calculated by calculating the absolute value of the difference between the current state and the state at the last time for each barrier control point (total number ), multiplying the operation switching weight , and then summing up. The second part represents the traffic disruption cost caused by barrier closure, which is calculated by subtracting the opening degree of each barrier control point from 1 , multiplying the state weight , and then summing up. This calculation method can suppress frequent switching of barrier states while maintaining barrier opening as much as possible under the premise of meeting water safety goals.

[0084] The expression is: ; Where, represents the total value of the control operation loss function; represents the total number of barrier control points; represents the i-th barrier control point; represents the operation switching weight of the i-th barrier control point, which is used to punish frequent state switching; represents the opening degree of the i-th barrier control point at time t, 0 represents complete closure, and 1 represents complete opening; represents the opening degree of the i-th barrier control point at time t-1; represents the absolute value of barrier state change; represents the state weight of the i-th barrier control point, which is used to measure the traffic impact caused by barrier closure; represents the closing degree of the barrier.

[0085] In different application scenarios, the form and weight coefficients of the above loss function can be adjusted according to specific needs. For example, in densely populated areas, the weight of the water safety loss function can be increased, while in traffic hub areas, the weight of the traffic disruption loss function can be increased.

[0086] As shown in Figure 5 , the influence relationship and its strength between different regional barrier control points are shown (the numerical value on the edge represents the influence coefficient). Through this diagram, the global influence path of barrier control can be clearly seen, for example, the influence coefficient of A zone barrier 1 on B zone barrier 1 is as high as 0.85, indicating that through accurate control of a small number of key nodes, effective regulation of the overall water flow network can be achieved, reducing the number of barrier control points activated by more than 50%.

[0087] Step 4.2, Predictive horizon rolling optimization; Based on the current flow state and weather forecast, predict the flow evolution trend in the prediction horizon (usually 30 minutes to 2 hours in the future), calculate the optimal control sequence for multiple time steps, and update the control decision in real time using the rolling horizon strategy.

[0088] The specific implementation process of rolling optimization control includes: State prediction: At time point t, predict the flow state and traffic state in the next N time steps using the network fluid dynamics model.

[0089] Specifically, for each future time step k (from 1 to N), first, based on the predicted flow state , predicted traffic state and barrier control state at the previous time, use the flow state dynamic model function to calculate the predicted flow state at the current time; then based on the predicted flow state at the current time, predicted traffic state and barrier control state at the previous time, use the traffic state dynamic model function to calculate the predicted traffic state at the current time.

[0090] In this way, the system state evolution trajectory in the entire prediction horizon can be obtained.

[0091] Control sequence optimization: based on the above state prediction results, solve the finite horizon optimization problem to obtain the optimal control sequence.

[0092] Specifically, find the barrier control sequence from the current time t to t+N-1, so that the sum of the loss function values at each time in the prediction horizon is minimized, that is, solve: ; Where, t represents the current time; N represents the prediction horizon length; k represents the time step index in the prediction horizon, from 0 to N-1; xt+k represents the predicted flow state at time t+k at time t; xt+k represents the predicted traffic state at time t+k at time t; uk represents the barrier control vector at time t+k; L represents the loss function, which is used to evaluate the goodness or badness of the system state; uk represents the parameter value that minimizes the objective function.

[0093] The optimization problem needs to consider multiple constraints, such as roadblock control constraints (e.g., operation frequency limits), water safety constraints (e.g., ensuring water levels in critical areas do not exceed safety thresholds), and necessary traffic passage guarantee constraints (e.g., ensuring emergency evacuation passages are unobstructed).

[0094] Control execution and rolling update: only the first control action of the optimal sequence is executed Then, at the next time point, the new system state is obtained, and the prediction and optimization process is repeated.

[0095] In some embodiments, the prediction horizon length can be dynamically adjusted according to the prediction uncertainty, using a longer prediction horizon when the prediction reliability is high, to improve the forward-looking nature of the control.

[0096] As shown in Figure 6 , the performance of the predictive control system is compared with that of the traditional control system from five dimensions: prediction response time, computational efficiency, collaborative optimization ability, dynamic adaptability, and resource allocation efficiency. The predictive control system significantly outperforms the traditional system in all dimensions, especially in collaborative optimization ability (95 points vs. 30 points) and resource allocation efficiency (92 points vs. 38 points), verifying the global collaborative optimization and resource optimization allocation capabilities of the system.

[0097] Step 4.3, hierarchical decomposition algorithm implementation; The global optimization problem is decomposed into multiple local sub-problems, and a distributed optimization algorithm is used to solve it, first determining control target allocation at the regional level, and then optimizing specific control parameters at the single-point level, effectively reducing the complexity of solving.

[0098] The specific implementation of the hierarchical decomposition algorithm includes: Regional division: based on hydrological characteristics and road network structure, the city is divided into multiple relatively independent control regions , where , , represent the first , , relatively independent control regions; represents the total number of regions.

[0099] Regional-level optimization: in the upper-level optimization, the control targets of each region are determined , where , , represent the control targets of the first , , region, The total number of regions, such as the maximum allowed water accumulation, the minimum guaranteed traffic capacity, etc. The upper-level optimization considers the water flow interaction between regions and global objectives.

[0100] Node-level optimization: For each region Under the condition of known control objectives Optimize the specific control parameters of the roadblock nodes within the region to make the local optimization results meet the regional control objectives.

[0101] Coordination mechanism: The coupling constraints between regions are handled by the Lagrange relaxation method to achieve coordinated and consistent control between regions. In addition, in some embodiments, advanced distributed optimization algorithms such as the alternating direction multiplier method can be used to further improve the solution efficiency.

[0102] Step 4.4, dynamic emergency channel planning; Based on the water flow prediction results and traffic demand, plan the dynamic change of emergency evacuation channels, optimize the traffic evacuation efficiency under the premise of ensuring safety, and dynamically adjust the control strategy through real-time feedback mechanism.

[0103] The specific implementation method of dynamic emergency channel planning includes: Traffic demand assessment: Based on urban traffic models and real-time monitoring data, assess the evacuation demand and importance of each region.

[0104] Safe channel identification: Combine the water flow prediction results to identify the subset of road network that will not be flooded in the future period.

[0105] Optimal path planning: On the safe channel network, solve the multi-source and multi-sink network flow optimization problem to determine the path allocation scheme that can maximize the evacuation efficiency. It should be noted that this optimization problem can consider various constraints such as the traffic capacity, congestion status, and future water accumulation risk of different road segments.

[0106] Roadblock collaborative control: Adjust the roadblock control strategy to ensure the smoothness of key evacuation channels, and if necessary, take temporary traffic control measures such as alternating opening and closing.

[0107] In practical applications, this multi-step prediction and optimization controller can predict the development trend of waterlogging in advance according to rainfall forecasts and real-time monitoring data, and develop forward-looking roadblock control strategies. For example, in the rainstorm emergency response of a certain city, the controller successfully predicted the water accumulation risk of a low-lying area and started the upstream roadblock control 15 minutes before the water actually formed, guiding traffic to avoid the risk area while leaving enough evacuation channels for vehicles within the region to safely evacuate.

[0108] The system also dynamically adjusts the original control plan according to the real-time feedback of the water level change rate, and shifts the control focus from the initially predicted A area to the B area with a faster actual water accumulation speed, showing the adaptability to complex changes.

[0109] The most significant advantage of the control system is its global collaborative control capability. By calculating the overall effect of different roadblock control schemes, the system can find a global scheme that is better than local control. In an actual waterlogging response, the system found that by temporarily closing a roadblock on a secondary road (forming a temporary water storage area), it could significantly alleviate the waterlogging pressure on the important traffic hub downstream. Ultimately, it successfully ensured the continuity of the main road traffic and controlled the overall water accumulation risk within a safe range.

[0110] As shown in Figure 7 , it directly shows the complete path of water flow from the beginning of the rain to the final drainage in a rainstorm event, and the impact of the roadblock control system on water flow distribution. The figure shows how the coordinated control of roadblock control points A, B, C, and D can guide part of the water flow to a temporary water storage area (90 units of flow), reduce the pressure on the downstream drainage system (80 units of flow), and at the same time ensure that the safe area is not affected (30 units of flow), verifying that the system can identify the key nodes of waterlogging development and guide the priority deployment of emergency resources.

[0111] The above describes the embodiments of the present application, but the embodiments are not limited to the specific implementation described above, which is only illustrative and not limiting. Those skilled in the art can make more forms of equivalent embodiments under the inspiration of the embodiments, which are all within the protection scope of the embodiments.

Claims

1. An automatic control method for urban flood emergency roadblocks based on water level sensing, characterized in that: The following steps are involved: Build a digital twin model of urban hydrology, integrating urban geographic data with hydrological parameters to form a digital model; Based on the digital twin model, a graph-structured network fluid dynamics solver was constructed. The urban road network and water flow channels were abstracted into a node-edge flow network. An improved finite volume method was used to transform the shallow water equations into a graph-structured discrete form. Using the calculation results of the network fluid dynamics solver, a roadblock control impact propagation model was constructed to quantitatively analyze the impact of roadblock control operations on water flow distribution and establish a mathematical relationship between roadblock status and water flow changes. Based on the roadblock control impact propagation model, a multi-step predictive optimization controller is implemented, and the model predictive control theory is used to calculate the optimal roadblock control strategy to achieve active prevention and dynamic regulation of urban waterlogging.

2. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 1 is characterized in that: The steps of constructing the urban hydrological digital twin model include: Obtain high-precision digital elevation model data, including urban surface elevation, road geometry, and drainage ditch locations; Collect urban drainage network data, build drainage system topology structure, and establish connection with surface flow model; Mark the locations of all controllable roadblock nodes and their control parameters in the digital model; Historical waterlogging event data were collected and the model parameters were calibrated using parameter optimization algorithms.

3. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 1 is characterized in that: The step of constructing a graph-structured network fluid dynamics solver comprises: Abstract the urban road network and water flow channels into a flow network with node-edge structure; The improved finite volume method is used to transform the traditional two-dimensional shallow water equations into a discrete form based on graph structure. Dynamically adjust the calculation accuracy for high-risk areas of waterlogging or complex water flow areas, using fine-grained calculations in key areas and coarse-grained calculations in non-key areas; Build a parallel computing framework suitable for graph-structured network flow computing, and decompose large-scale fluid computing tasks into subtasks for parallel processing.

4. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 3 is characterized in that: The flow state change rate function in the network fluid dynamics solver describes the rate of change of the flow state, taking into account the effects of water level gradient, gravity, ground friction, and roadblocks on the water flow. The water flow rate is calculated using the improved Manning formula, which multiplies the effective cross-sectional area of ​​the edge by the second power of the hydraulic radius, then multiplies it by the square root of the absolute value of the water level difference, divided by the product of the Manning roughness coefficient and the square root of the edge length, and finally multiplies the calculated result by the complementary value of the roadblock factor.

5. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 1 is characterized in that: The steps of constructing the roadblock control impact propagation model include: For each roadblock control point, by simulating the water flow changes under different control states, calculate its influence coefficient on the water level, flow velocity and other parameters of the adjacent area; Analyze the synergistic effects of simultaneous or sequential operation of multiple roadblock control points and identify control combinations that enhance or offset each other; Calculate the sensitivity of the impact of the change of the roadblock control vector on the water flow state vector to form a sensitivity matrix; Based on the sensitivity matrix, the key control points and the optimal control sequence are identified, and the minimum spanning tree algorithm in graph theory is used to determine the combination of roadblock operations that can achieve the target water flow regulation effect with the minimum control cost.

6. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 5 is characterized in that: The methods for generating the sensitivity matrix include: Apply a small perturbation to each roadblock control parameter, calculate the resulting change in the flow state parameters, and obtain the partial derivative as the sensitivity; Analyze the time-varying characteristics of sensitivity and construct a dynamic sensitivity matrix to predict the time delay effect of roadblock control; Set sensitivity thresholds, identify highly sensitive control relationships, and simplify the complexity of control models.

7. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 1 is characterized in that: The steps of implementing the multi-step predictive optimization controller include: Establish a comprehensive optimization function that includes water safety goals and traffic evacuation goals. The comprehensive optimization function represents the integral of system losses from the initial moment to the final moment. The loss function comprehensively considers the risk of waterlogging, traffic disruption costs, and control operation costs. Based on the current water flow state and weather forecast, the water flow evolution trend is predicted within the forecast time domain and the optimal control sequence for multiple time steps is calculated; Decompose the global optimization problem into multiple local sub-problems and solve them using a distributed optimization algorithm; Based on water flow prediction results and traffic demand, dynamically changing emergency evacuation channels are planned to optimize traffic evacuation efficiency while ensuring safety.

8. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 7 is characterized in that: The specific definition of the loss function is: The total system loss is equal to the water safety loss multiplied by the first weight coefficient, plus the traffic blockage loss multiplied by the second weight coefficient, plus the control operation loss multiplied by the third weight coefficient, where: The water safety loss function calculates the square of the portion of the current water level of each node in the set of key monitoring nodes that exceeds the safe water level threshold, multiplying it by the risk weight of the node, and then summing them; The traffic blockage loss function calculates the ratio of the current traffic capacity to the maximum traffic capacity of each road edge in the critical road edge set, subtracts the ratio from 1 to get the capacity loss rate, multiplies it by the traffic importance weight of the road, and then sums the results; The control operation loss function includes the switching cost of the control operation and the traffic blocking cost caused by the closure of the roadblock.

9. The automatic control method for urban flood emergency roadblocks based on water level sensing according to claim 7 is characterized in that: The steps of forecast horizon rolling optimization include: At time point t, the network fluid dynamics model is used to predict the water flow and traffic conditions in the next N time steps; Based on the state prediction results, solve the finite time domain optimization problem and obtain the optimal control sequence; Only the first control action of the optimal sequence is executed, and then the new system state is obtained at the next time point, and the prediction and optimization process is repeated.

10. An automatic control system for urban flood emergency roadblocks based on water level sensing, used to implement the automatic control method for urban flood emergency roadblocks based on water level sensing according to any one of claims 1 to 9, characterized in that: include: Hydrological monitoring module, used to collect water level sensor data, traffic flow data, and meteorological monitoring data from various parts of the city; Digital twin module, used to build and maintain urban hydrological digital twin models; Network fluid dynamics calculation module, used to perform water flow simulation calculations based on graph structure; Impact propagation analysis module, used to analyze the impact of roadblock control on water flow distribution; Predictive optimization control module, used to calculate the optimal roadblock control strategy; Roadblock control execution module, used to send control instructions to each roadblock control point; Emergency evacuation planning module, used to plan dynamically changing emergency evacuation channels.

Images