A method and system for intelligent risk assessment of dam-break flood propagation of water network nodes

Abstract

This invention discloses an intelligent risk assessment method and system for dam-break flood propagation in water network nodes, comprising: collecting basic data of water network engineering and dam-break scenario parameters to generate a standardized dataset; constructing a directed topological structure model of water network nodes-channels that integrates engineering semantic constraints and hydraulic propagation direction constraints, and generating a water network adjacency matrix; using a neural network model based on water network structural constraints and the physical knowledge of mass and momentum conservation to predict the propagation process of dam-break floods in the water network; converting the flood propagation prediction results into flood impact indices, calculating the risk levels of engineering facilities and population units based on the flood impact indices, and outputting risk assessment results and decision support information. This invention provides decision support information for water network scheduling and emergency management.

Description

Technical Field

[0001] This invention belongs to the field of water disaster risk assessment technology, specifically relating to an intelligent risk assessment method and system for the propagation of dam-break floods at water network nodes. Background Technology

[0002] With the large-scale construction of water conservancy projects, the water network system composed of reservoirs, dams, rivers, and water diversion projects is becoming increasingly complex, and the hydraulic connections between nodes are becoming closer. As a type of water disaster with strong suddenness and great destructive power, once a dam-break flood occurs, the flood will spread rapidly along the water network channels, posing a serious threat to engineering facilities and the safety of people's lives and property along the route. Therefore, accurately assessing the propagation process and risk level of dam-break floods in the water network is the key to improving the emergency response capabilities for water disasters and ensuring the safety of water conservancy projects and the safety of people's lives and property.

[0003] Currently, methods for assessing the propagation and risk of dam-break floods are mainly divided into two categories: traditional hydraulic modeling and data-driven modeling. Traditional hydraulic modeling is based on fluid dynamics equations and can reflect the physical mechanism of flood propagation, but it suffers from problems such as high computational complexity, poor convergence, and weak adaptability to complex water network topologies. In particular, in complex water networks with multiple nodes and channels, it is difficult to quickly and accurately predict the flood propagation process and cannot meet the timeliness requirements of emergency decision-making.

[0004] Data-driven modeling has been widely used in flood forecasting due to its powerful fitting and prediction capabilities. However, most existing data-driven models rely solely on historical data for training, lacking constraints on the semantics of water network engineering and the direction of hydraulic propagation. They also fail to fully integrate core physical knowledge such as mass conservation and momentum conservation, leading to model prediction results that easily deviate from actual hydraulic laws. This results in problems such as insufficient prediction accuracy and poor generalization ability, making it difficult to accurately reflect the propagation characteristics of dam-break floods in complex water networks.

[0005] In addition, existing risk assessment methods mostly target only a single engineering facility or a local area, failing to achieve coordinated risk assessment of engineering facilities and population units within the water network system. Furthermore, the risk level classification is not detailed enough, making it impossible to provide accurate key control targets and decision support information for emergency response, and thus failing to meet the actual needs of emergency management of dam-break floods in complex water network systems.

[0006] This invention proposes an intelligent risk assessment method and system for dam-break flood propagation at water network nodes, which solves the above-mentioned problems. It integrates engineering semantics, hydraulic constraints and physical knowledge to achieve accurate prediction of dam-break flood propagation, while taking into account the collaborative risk assessment of engineering facilities and population units. Summary of the Invention

[0007] Objective: To provide an intelligent risk assessment method for the propagation of dam-break floods at water network nodes, thereby addressing the aforementioned problems in existing technologies. Furthermore, to provide an intelligent risk assessment system for the propagation of dam-break floods at water network nodes.

[0008] Technical solution: A smart risk assessment method for dam-break flood propagation at water network nodes, comprising the following steps:

[0009] Step S1: Collect basic data on water network engineering and dam failure scenario parameters, and generate a standardized dataset;

[0010] Step S2: Construct a directed topological model of water network nodes and channels that integrates engineering semantic constraints and hydraulic propagation direction constraints, and generate a water network adjacency matrix;

[0011] Step S3: Use a neural network model based on the constraints of the water network structure and the physical knowledge of mass conservation and momentum conservation to predict the propagation process of dam-break flood in the water network;

[0012] Step S4: Convert the flood propagation prediction results into flood impact indicators, calculate the risk level of engineering facilities and population units based on the flood impact indicators, and output the risk assessment results and decision support information.

[0013] According to one aspect of this application, step S1 is further comprising:

[0014] Step S11: Collect basic operational data and dam failure scenario parameters of reservoirs, dams, rivers and water diversion projects in the water network system. The basic operational data includes the initial water level of the reservoir, reservoir capacity, outflow and river hydraulic parameters. The dam failure scenario parameters include the outflow process and boundary condition information of the breach.

[0015] Step S12: Unify the coordinate system and time step of the basic operation data and dam failure scenario parameters of the collected water network system, including reservoir dams, rivers and water diversion projects, and unify and normalize the dimensions of water level, outflow and reservoir capacity data to generate a standardized dataset.

[0016] According to one aspect of this application, step S2 further comprises:

[0017] Step S21: Construct a directed topological model of water network nodes and channels based on engineering semantic constraints and hydraulic propagation direction constraints;

[0018] Step S22: Input the standardized dataset into the directed topology model of water network nodes-channels to generate the water network adjacency matrix.

[0019] According to one aspect of this application, step S3 further comprises:

[0020] Step S31: Construct a neural network model based on water network structure constraints and the conservation of mass and momentum;

[0021] Step S32: Extract the prediction time, water network adjacency matrix, reservoir initial state, hydraulic attribute matrix, external boundary conditions, and breach discharge process as inputs to the neural network model. Under the constraints of water network structure and physical knowledge, perform time-series prediction of the propagation process of dam-break flood in the water network, and obtain the water level, water depth, and flow velocity of each node in the water network at any prediction time.

[0022] According to one aspect of this application, step S4 further comprises:

[0023] Step S41: Set the dangerous water depth threshold, calculate the time index of the first arrival of the flood to trigger the risk conditions, and convert the flood propagation prediction results into flood impact indicators;

[0024] Step S42: Calculate the risk level of engineering facilities and population units based on flood impact indicators;

[0025] Step S43: Output risk assessment results and decision support information.

[0026] According to one aspect of this application, step S41 further comprises:

[0027] Step S41a: Set dangerous water depth thresholds based on water network node types and regional water conservancy disaster prevention specifications; retrieve flood time series prediction data based on dangerous water depth thresholds; calculate the time index of the risk triggering condition for the first arrival of flood; and extract the hydraulic parameters at that moment to construct the first impact index vector.

[0028] Step S41b: Based on the initial impact index vector, extract and quantify the arrival response index, extreme impact index and cumulative effect index respectively. After the three types of indexes are processed by interval standardization, introduce risk contribution weight to uniformly map them into the node comprehensive impact intensity index.

[0029] According to one aspect of this application, step S42 further comprises:

[0030] Step S42a: For different types of projects within the water network structure, construct vulnerability functions corresponding to the facility types, and calculate the probability of facility failure under different flood impact conditions;

[0031] Step S42b: Construct a population exposure risk model based on population size and calculate the exposure risk of population units.

[0032] According to one aspect of this application, step S43 further comprises:

[0033] Step S43a: Combine the failure risk of engineering facilities with the exposure risk of the population to construct a comprehensive risk index;

[0034] Step S43b: Based on the magnitude of the comprehensive risk value, classify the engineering facilities and population units into risk levels of extremely high risk, high risk, medium risk and low risk, and output the spatial distribution map of high-risk objects, the list of engineering facilities and population units to be prioritized and monitored, and obtain risk ranking and control recommendations for emergency response.

[0035] According to another aspect of this application, a smart risk assessment system for dam-break flood propagation at a water network node is provided, comprising:

[0036] At least one processor; and

[0037] A memory communicatively connected to at least one of the processors; wherein,

[0038] The memory stores instructions that can be executed by the processor to implement the intelligent risk assessment method for dam-break flood propagation at water network nodes as described in any of the above technical solutions.

[0039] Beneficial effects: The adoption of an intelligent risk assessment method for the propagation of dam-break floods at water network nodes provides accurate and actionable decision support information for emergency response, effectively improving the scientific nature and pertinence of emergency management of dam-break floods in complex water network systems. Attached Figure Description

[0040] Figure 1 This is a flowchart of the present invention.

[0041] Figure 2 This is a flowchart of step S1 of the present invention.

[0042] Figure 3 This is a flowchart of step S2 of the present invention.

[0043] Figure 4 This is a flowchart of step S3 of the present invention.

[0044] Figure 5 This is a flowchart of step S4 of the present invention. Detailed Implementation

[0045] like Figure 1 As shown, the following technical solution is proposed. According to one aspect of this application, a method for intelligent risk assessment of dam-break flood propagation at water network nodes is provided, characterized by comprising the following steps:

[0046] Step S1: Collect basic data on water network engineering and dam failure scenario parameters, and generate a standardized dataset;

[0047] Step S2: Construct a directed topological model of water network nodes and channels that integrates engineering semantic constraints and hydraulic propagation direction constraints, and generate a water network adjacency matrix;

[0048] Step S3: Use a neural network model based on the constraints of the water network structure and the physical knowledge of mass conservation and momentum conservation to predict the propagation process of dam-break flood in the water network;

[0049] Step S4: Convert the flood propagation prediction results into flood impact indicators, calculate the risk level of engineering facilities and population units based on the flood impact indicators, and output the risk assessment results and decision support information.

[0050] like Figure 2 As shown, according to one aspect of this application, step S1 further comprises:

[0051] Step S11: Collect basic operational data and dam failure scenario parameters of reservoirs, dams, rivers and water diversion projects in the water network system. The basic operational data includes the initial water level of the reservoir, reservoir capacity, outflow and river hydraulic parameters. The dam failure scenario parameters include the outflow process and boundary condition information of the breach.

[0052] Acquire basic data for water network projects, including: reservoir dam data (dam site coordinates, reservoir capacity-water level curve, normal storage level, current water level); river channel and water diversion project data (connectivity, direction, length, gradient, roughness, cross-sectional width); and boundary and initial value data (upstream inflow historical process line, downstream control water level, and reservoir capacity).

[0053] Step S12: Unify the coordinate system and time step of the basic operation data and dam failure scenario parameters of the collected water network system, including reservoir dams, rivers and water diversion projects, and unify and normalize the dimensions of water level, outflow and reservoir capacity data to generate a standardized dataset.

[0054] The above data is processed to achieve a unified spatiotemporal reference: the coordinate system is unified, the time step is unified, and the dimensions of water level, flow rate, reservoir capacity and other quantities are unified and normalized to generate a standardized dataset that can be used as input to the model.

[0055] like Figure 3 As shown, according to one aspect of this application, step S2 further comprises:

[0056] Step S21: Construct a directed topological model of water network nodes and channels based on engineering semantic constraints and hydraulic propagation direction constraints;

[0057] Step S22: Input the standardized dataset into the directed topology model of water network nodes-channels to generate the water network adjacency matrix.

[0058] The water network system is abstracted as a weighted directed graph model that integrates engineering semantic constraints and hydraulic propagation direction discrimination mechanisms. This model is used to uniformly describe the hydraulic connectivity between reservoirs, dams, rivers, and water diversion projects, as well as the potential propagation direction, propagation capacity, and propagation differences of dam-break floods within the water network structure. Its mathematical expression is as follows:

[0059] G = (V, E, A);

[0060] In the formula: V = {V1, V2, ..., V} N} represents the set of nodes, where each node represents an actual existing reservoir dam; N is the total number of reservoir dams in the water network; E = {e ij Let} be the set of connected edges. When reservoir i and reservoir j are connected by a stable hydraulic link through a natural river or artificial water diversion channel, and the water flow can propagate from node i to node j under the combined action of hydraulic and engineering constraints during engineering operation and extreme conditions, a directed edge e is established from the upstream node to the downstream node. ij The direction is from the upstream dam to the downstream dam; wherein, the direction of the directed connection not only reflects the physical direction of water flow driven by topographic water level difference, but also comprehensively considers the longitudinal slope of the river channel, the control method of hydraulic structures and the operation rules of the project, forming a hydraulic propagation direction constraint that satisfies the semantic consistency of the project.

[0061] Based on this, in order to enhance the ability of the water network structure diagram to express the differences in engineering attributes and flood propagation, the directed edges, while describing the physical direction of water flow, further introduce engineering semantic weight parameters related to the engineering characteristics of the river or channel, so as to reflect the differences in the carrying and transport capacity of different rivers or channels under dam-break flood conditions.

[0062] A = [a] ij Let be the adjacency matrix of the water network, where matrix element a ij Used to characterize the directed connection relationship between node i and node j and its engineering semantic features, when there is a directed edge e ij At that time, a ij The parameters are jointly determined by the flood control capacity, channel-scale parameters, and engineering control characteristics associated with the connected path, and are used to comprehensively reflect the effective conduction capacity of the path during the propagation of dam-break floods; when there is no propagation path between nodes i and j that satisfies both engineering and hydraulic constraints, let a ij=0, thus constructing an asymmetric weighted adjacency matrix. This structure can characterize the potential propagation paths, directional constraints, and differences in propagation capabilities of dam-break floods within a unified mathematical framework, providing a structured input basis for subsequent flood propagation prediction, risk evolution analysis, and key impact path identification based on the water network structure. Through the above definition, the complex water network engineering system can be simplified into a structured network model composed of "reservoir nodes—engineering semantic edges," thereby clarifying the possible propagation directions, differences in propagation capabilities, and key impact paths of dam-break floods within the water network under a unified mathematical framework.

[0063] Furthermore, by constructing the directed graph model G=(V,E,A) of the water network structure that integrates engineering semantic constraints and hydraulic propagation discrimination mechanism, the structural topological characteristics and engineering constraints of the water network engineering system can be described simultaneously under a unified mathematical framework. This forms an engineering semantic constraint spatiotemporal graph model for the propagation problem of dam-break floods, providing a standardized, computable, and scalable input data foundation for subsequent simulation and risk assessment of the propagation of dam-break floods in the water network.

[0064] In one embodiment, specifically:

[0065] A mountainous watershed water network system comprises three operational reservoirs: upstream reservoir A, midstream reservoir B, and downstream reservoir C. The core parameters of each reservoir and its connecting channels are as follows, forming the basis of a standardized dataset:

[0066] Node V1 (Upstream Reservoir A): Dam height 25m, total reservoir capacity 8 million m³ 3 Controlled drainage area of ​​120 km² 2 The normal water level is 186.5m.

[0067] Node V2 (Midstream Reservoir B): Dam height 32m, total reservoir capacity 12 million m³ 3 Controlled drainage area of ​​280 km² 2 The normal water level is 172.3m.

[0068] Node V3 (downstream reservoir C): Dam height 18m, total reservoir capacity 5 million m³ 3 Controlled drainage area of ​​85 km² 2 The normal water level is 158.7m.

[0069] Connection channels: V1 and V2 are connected by a natural river channel, 18km long, 25m wide at the bottom, with a longitudinal slope of 0.6‰, and no artificial control structures; V2 and V3 are connected by a natural river channel and a control gate, 12km long, 20m wide at the bottom, with a longitudinal slope of 0.4‰, and a maximum discharge capacity of 350m³. 3 / s; There is no natural river channel or artificial water diversion channel between V1 and V3, and they do not have stable hydraulic connection conditions.

[0070] Taking the three reservoir dams in this water network as core nodes, i.e., V={V1, V2, V3}, where the total number of reservoir dams is n=3, and each node uniquely corresponds to an actual reservoir dam;

[0071] Combining hydraulic propagation direction constraints and engineering semantic constraints, the hydraulic connectivity and propagation direction between nodes are determined. The hydraulic propagation direction constraints comprehensively consider topographic water level difference, river longitudinal slope, hydraulic structure control methods and engineering operation rules, while the engineering semantic constraints focus on the engineering characteristics and connectivity stability of the channel.

[0072] Based on the assessment, the normal water level of V1 (186.5m) is higher than that of V2 (172.3m). The natural riverbed slope between the two is 0.6‰, and there are no artificial control structures. The water flow can stably propagate from V1 to V2 driven by the topographic water level difference, satisfying the engineering semantics and hydraulic constraints. Therefore, a directed connection E is established. 12 The direction is V1→V2; the normal water level of V2 (172.3m) is higher than that of V3 (158.7m), and the longitudinal slope of the river between them is 0.4‰. A control gate (engineering control characteristic) is provided. According to the engineering operation rules, the control gate can be opened normally under both normal and extreme conditions, allowing water flow to propagate from V2 to V3, satisfying the constraints. Therefore, a directed connection E is established. 23 The direction is V2→V3; there is no connecting channel between V1 and V3, and the conditions for hydraulic propagation are not met, so no edge is established. Ultimately, the set of directed edges is E={E...} 12 E 23};

[0073] To characterize the differences in flood carrying and transport capacity among different channels, a weight parameter is assigned to each directed connection. This weight parameter is jointly determined by flood control capacity parameters, channel-scale parameters, and engineering control characteristic parameters, where E... 12 Without artificial control structures, flood control capacity is primarily determined by the river channel size. 23 Because a control gate is installed, the weight parameters need to be modified in combination with the control characteristics of the control gate to obtain the directed topological structure model of water network node-channel G=(V,E,A).

[0074] The parameters of each node and channel in the above implementation scenario are standardized to remove redundant information and unify the parameter units, forming a standardized dataset. Specifically, it includes the basic engineering parameters of each node, the flood control capacity parameters (such as river discharge) corresponding to each directed connection, the river scale parameters (bottom width, length, longitudinal slope) and the engineering control characteristic parameters (whether there are control structures and the operating parameters of control structures).

[0075] The standardized dataset is input into the directed topology model of water network nodes-channels. An asymmetric weighted adjacency matrix A is generated according to the definition of an adjacency matrix. The matrix element aᵢⱼ represents the directed connection relationship between node i and node j and its engineering semantic features. The specific rules are as follows:

[0076] When there is a directed connection between node i and node j that satisfies engineering and hydraulic constraints, aᵢⱼ is determined by weighted calculation of the flood control capacity parameter, channel scale parameter and engineering control characteristic parameter corresponding to the connection.

[0077] When there is no effective propagation path between node i and node j, let a ij =0.

[0078] Using the standardized dataset of this embodiment, the adjacency matrix A is calculated.

[0079] The specific meaning of this adjacency matrix is: a 12 =0.72, representing the directed connection from V1 to V2, with a weight of 0.72, reflecting that the channel is uncontrolled, has a large river scale, and a strong ability to propagate floods due to dam break; a 23 =0.58, representing the directed connection from V2 to V3, with a weight of 0.58. Because this channel has a control gate, it restricts the propagation of floodwater to a certain extent, so its propagation capacity is slightly lower than E. 12 The remaining matrix elements are all 0, indicating that there is no hydraulic propagation path that satisfies the constraints between the corresponding nodes.

[0080] In existing technologies, the parameters of the main engineering components of water network systems, such as reservoirs, dams, and rivers, are diverse in type and inconsistent in format, making it difficult to perform unified modeling and analysis. This method abstracts the water network system into a weighted directed graph model G=(V,E,A) that integrates engineering semantic constraints and hydraulic propagation direction discrimination mechanisms. Reservoirs and dams are used as nodes, and channels that satisfy the constraints are used as directed edges. Multi-source engineering parameters and hydraulic parameters are uniformly integrated into the graph model and adjacency matrix, realizing a structured and standardized description of the water network engineering system. This solves the technical pain point that multi-source heterogeneous data cannot be effectively integrated. For example, in this embodiment, reservoir parameters, river parameters, and gate control parameters are uniformly integrated into the model to form a standardized adjacency matrix, providing a unified data foundation for subsequent analysis.

[0081] This method, when constructing the model, not only considers the physical direction of water flow driven by topographic water level difference, but also integrates engineering semantic constraints such as river longitudinal slope, hydraulic structure control methods, and engineering operation rules. This ensures that the direction and weight of directed edges are consistent with the actual engineering operation scenario, avoiding the problem of existing pure physical flow direction modeling ignoring engineering constraints and being out of touch with reality. For example, in this embodiment, the edge weight between V2 and V3 is modified in combination with the control characteristics of the control gate, accurately reflecting the impact of the control gate on flood propagation. This enables the model to truly represent the actual propagation law of dam-break floods, improving the reliability and practicality of the model.

[0082] In existing technologies, the adjacency matrix of water network topology models is mostly a 0-1 matrix, which can only represent whether there is a connectivity relationship between nodes and cannot reflect the differences in flood propagation capacity of different channels. This method introduces engineering semantic weight parameters to directed edges, and the weights are jointly determined by parameters such as flood carrying capacity, river scale, and engineering control characteristics, so that the elements of the adjacency matrix can quantify the flood conduction capacity of different channels. For example, in this embodiment, a 12 =0.72、a 23 =0.58, clearly distinguishing the difference in propagation capacity between the two channels, providing accurate quantitative basis for subsequent identification of dam-break flood propagation paths and assessment of key channels;

[0083] The asymmetric weighted adjacency matrix generated by this method is standardized, computable, and scalable structured data. It can be directly used as input data for subsequent tasks such as dam-break flood propagation prediction, risk evolution analysis, and key impact path identification without additional data conversion and processing, thus reducing the complexity of subsequent analysis. This method simplifies the complex water network engineering system into a structured network model composed of "reservoir nodes - engineering semantic edges", which greatly simplifies the modeling process and reduces the complexity and computational load of the model.

[0084] like Figure 4 As shown, according to one aspect of this application, step S3 further comprises:

[0085] Step S31: Construct a neural network model based on water network structure constraints and the conservation of mass and momentum;

[0086] Step S32: Extract the prediction time, water network adjacency matrix, reservoir initial state, hydraulic attribute matrix, external boundary conditions, and breach discharge process as inputs to the neural network model. Under the constraints of water network structure and physical knowledge, perform time-series prediction of the propagation process of dam-break flood in the water network, and obtain the water level, water depth, and flow velocity of each node in the water network at any prediction time.

[0087] In this embodiment, an adaptive physical knowledge-guided neural network model with propagation stage perception capability is constructed for water network structures to predict the spatiotemporal evolution of dam-break floods within the water network engineering system. This neural network not only utilizes historical and structural data for fitting and prediction but also introduces hydrodynamic physical constraints and a propagation stage discrimination mechanism to achieve physical consistency constraints and dynamic control of the flood propagation process. The model's inputs include time variables, water network structure information, the initial state of each reservoir, river channel and water diversion channel engineering attributes, breach discharge process lines, and external boundary conditions, specifically:

[0088] (t, X(t0), Z(t0), A, Q) breach (t));

[0089] In the formula: t is the prediction time variable, representing the time step after the dam breach, in seconds or hours, used to characterize the temporal position of the flood evolution; X(t0) represents the initial state matrix of all reservoir nodes in the water network at the time t0 of the dam breach, specifically:

[0090] X(t0)=[[H1(t0), h1(t0), Q1(t0)],

[0091] ⋮

[0092] [H N (t0), h N (t0), Q N (t0)]];

[0093] In the formula: H i (t0) represents the initial water level of the i-th reservoir before the dam failure; h i (t0) represents the average water depth of the i-th reservoir before the dam failure; Q i Z(t0) represents the outflow from the i-th reservoir before the dam failure; Z(t0) represents the hydraulic attribute matrix of the river channels and water diversion channels in the water network, specifically:

[0094] Z = [[L1, S1, n1],

[0095] ⋮

[0096] [L M S M n M ]];

[0097] In the formula: L j S is the length of the j-th river channel or passage; j n represents the longitudinal slope of the corresponding channel. jQ represents the roughness coefficient of the corresponding channel; A is the water network adjacency matrix, used to describe the connection relationship between reservoir nodes and river channels. If there is a hydraulic connection between different reservoir nodes, then A=1; otherwise, A=0; breach (t) is a function of the flow rate downstream of the breach after the dam breaks, which is used to characterize the external hydrodynamic input of the dam break process to the water network system. This flow rate process is pre-calculated by the breach parameter calculation model and is input into the neural network model as a known time series.

[0098] The model outputs the predicted state of each reservoir node in the water network at any given time:

[0099] y*_i(t)=[H*_i(t), h*_i(t), v*_i(t)];

[0100] In the formula: H*_i(t), h*_i(t), and v*_i(t) represent the water level, water depth, and flow velocity of the i-th reservoir at time t, respectively. That is, the model can directly give how high the water rises, how deep the water is, and how fast the water flows near each reservoir at any time after the dam breaks.

[0101] In this embodiment, to ensure that the prediction results conform to the basic laws of hydrodynamics at the water network scale, mass conservation and momentum conservation are introduced as physical constraints during the neural network training and deduction process. The mass conservation residual is defined as:

[0102] R_ij^(m)(x,t)=Ξh*_ij / Ξt+Ξ(h*_ijv*_ij) / Ξx-r*_ij(x,t);

[0103] This indicates whether the predicted changes in water depth and flow rate satisfy the "conservation of water volume", where Ξ represents the partial derivative and r*_ij(x,t) is a dimensionless source-sink term function that characterizes the increase or decrease in water volume per unit time and unit space.

[0104] The residual of momentum conservation is defined as:

[0105] R_ij^(p)(x,t)=Ξv*_ij / Ξt+v*_ijΞv*_ij / Ξx+gΞh*_ij / Ξx+g(S_f,ij-S_0,ij);

[0106] In the formula, g is the gravitational acceleration, S_0,ij is the river slope, which is used to reflect the influence of river resistance on water flow. The physical residual is not introduced as a static constraint term, but is dynamically related to the propagation stage of flood in the water network, and is used to characterize the difference in the importance of physical consistency constraints under different propagation stages.

[0107] During model training and prediction, the data fitting error and the aforementioned physical residuals are combined to form the total loss function, specifically:

[0108] L=L_data+λ_m(t)L_mass+λ_p(t)L_mom;

[0109] In the formula:

[0110] L_mass=(1 / |Ω|)Σ(R^(m)) 2 , L_mom=(1 / |Ω|)Σ(R^(p))²;

[0111] Where L_mass and L_mom represent the mean square values ​​of the residuals of mass conservation and momentum conservation, respectively.

[0112] To adapt to the characteristics of different stages of flood "rise-propagation-attenuation", an adaptive weighting mechanism is introduced, and a flood change intensity index is defined:

[0113] k(t)=(∇_tH*(t)_2) / (||H*(t)||_2+ε);

[0114] Where k(t) is the flood change intensity index, ∇_t is the gradient operator with respect to time t, representing the rate of change of H*(t) with time, H*(t) is the water level of the reservoir at time t, and ε is a very small regularization constant to prevent the denominator from being 0 and to ensure the stability of numerical calculation;

[0115] And update the physical knowledge constraint weights accordingly:

[0116] λ_m(t)=λ_m^min+(λ_m^max-λ_m^min)*σ(αk(t)-k_0);

[0117] Where λ_m(t) is the adaptive weight of the mass conservation constraint, λ_m^max and λ_m^min are the minimum / maximum boundaries of the weight, respectively, to prevent the weight from being too small or too large, σ(⋅) is the activation function, which maps the input to the (0,1) interval, α is the scaling factor, which is used to adjust the influence intensity of k(t), and k_0 is the threshold parameter, which controls the critical water level change level at which the weight begins to increase significantly;

[0118] When flood propagation is rapid and system state changes drastically, the physical consistency constraint strength is automatically increased to suppress non-physical oscillations; when the flood enters a slower evolution phase, the physical constraint strength is appropriately weakened to improve prediction flexibility and local accuracy. Through the above input, output, and physical knowledge constraint mechanisms, the model can predict the propagation process of floods in the water network after a dam breach in a relatively short time, providing a reliable data foundation for subsequent flood arrival time, inundation depth, and risk assessment.

[0119] In one embodiment, specifically:

[0120] Based on the directed topological structure model of water network nodes-channels and the water network adjacency matrix, and incorporating water network structural constraints, while using mass conservation and momentum conservation as core physical knowledge constraints, and combining an adaptive weighting mechanism to adapt to the different stages of flood "rise-propagation-attenuation", a neural network model is constructed to predict the spatiotemporal evolution of dam-break floods in the water network engineering system.

[0121] The prediction time, water network adjacency matrix, reservoir initial state, hydraulic property matrix, external boundary conditions, and breach discharge process are extracted, standardized, and then input into the neural network model.

[0122] The prediction time variable t represents the time step after the dam break, in hours, and is used to characterize the time position of the flood evolution. In this embodiment, the time step is set to 0.5 hours, and the prediction duration is 72 hours, covering the complete process of flood rise, propagation, and decay.

[0123] The initial state matrix S_0 of the reservoir: represents the initial state matrix of all reservoir nodes in the water network at the time t=0 of the dam failure, specifically as follows:

[0124] S0=[[h 10 d 10 q 10 ],

[0125] [h 20 d 20 q 20 ],

[0126] [h 30 d 30 q 30 ]];

[0127] Where: h i0 d represents the initial water level of the i-th reservoir before the dam failure (in meters); i0 Let q be the average water depth (in meters) of the i-th reservoir before the dam failure; i0 The outflow from the i-th reservoir before the dam failure (unit: m³) 3 / s);

[0128] In this embodiment, the initial state parameters of each reservoir are as follows:

[0129] V1(h) 10 =186.5m, d 10 =12.3m, q 10 =25m 3 / s); V2(h 20 =172.3m, d20 =15.7m, q 20 =38m 3 / s); V3 (h 30 =158.7m, d 30 =9.8m, q 30 =19m 3 / s);

[0130] Hydraulic attribute matrix Z: Represents the hydraulic attribute matrix of rivers and water diversion channels in the water network, specifically:

[0131] Z = [[L1, J1, n1],

[0132] [L2, J2, n2]];

[0133] In the formula: L j J represents the length of the j-th river or channel (in km); j n represents the longitudinal slope of the corresponding passage (unit: ‰); j To correspond to the roughness coefficient of the channel, in this embodiment, the hydraulic property parameters of the two channels are as follows: E 12 (L1=18km, J1=0.6‰, n1=0.025); E 23 (L2=12km, J2=0.4‰, n2=0.028);

[0134] Water network adjacency matrix A: The asymmetric weighted adjacency matrix generated in step S22 is used to describe the connection relationship and propagation capacity differences between reservoir nodes and river channels;

[0135] The breach discharge flow rate Q(t) is a function of the change in the breach discharge flow rate over time after the dam breach occurs. It is used to characterize the external hydrodynamic input of the dam breach process to the water network system. In this embodiment, it is assumed that the dam V1 breaches. This flow rate process is pre-calculated by the breach parameter calculation model and used as a known time series input to the neural network model. Its core parameter is: the initial breach discharge rate of 1200 m³ / s. 3 / s, decaying to 800m after 1 hour. 3 / s, decaying to 400m after 3 hours. 3 / s, and stabilized after 6 hours (200m). 3 / s);

[0136] External boundary conditions: The external boundary conditions are mainly the watershed rainfall (set as a constant in this embodiment, with a rainfall of 5 mm / h) and the river outlet water level constraint (the downstream river outlet water level of V3 is fixed at 156.2 m).

[0137] Accurate predictions were made of the flood propagation process in the water network after a dam breach. Specific prediction results are as follows (excerpts of key moments):

[0138] At t=6 hours (flood propagation stage): V1 water level drops to 178.2m, water depth 10.1m, flow velocity 1.8m / s; V2 water level rises to 175.6m, water depth 18.9m, flow velocity 1.2m / s; V3 water level rises to 160.3m, water depth 11.5m, flow velocity 0.9m / s.

[0139] At t=24 hours (early stage of flood decay): the water level of V1 stabilizes (175.8m), with a depth of 9.6m and a flow velocity of 0.7m / s; the water level of V2 drops to 173.2m, with a depth of 16.4m and a flow velocity of 0.8m / s; the water level of V3 rises to 162.1m, with a depth of 13.3m and a flow velocity of 0.6m / s.

[0140] t=72 hours (flood stabilization stage): The water levels of all reservoirs tend to stabilize, with V1 water level at 175.5m, V2 water level at 172.5m, and V3 water level at 161.8m, and the flow velocity all dropping below 0.3m / s.

[0141] like Figure 5 As shown, according to one aspect of this application, step S4 further comprises:

[0142] Step S41: Set the dangerous water depth threshold, calculate the time index of the first arrival of the flood to trigger the risk conditions, and convert the flood propagation prediction results into flood impact indicators;

[0143] Step S42: Calculate the risk level of engineering facilities and population units based on flood impact indicators;

[0144] Step S43: Output risk assessment results and decision support information.

[0145] In this embodiment, the continuous time-series results of water level, water depth, and flow velocity predicted by the aforementioned neural network based on adaptive physics knowledge are further transformed into quantitative indicators that can characterize the risk response features of different nodes, engineering facilities, and population units in the water network. This provides a unified and computable indicator input for subsequent risk classification assessment and decision analysis. The aforementioned model can output the predicted water level, water depth, and flow velocity of each node in the water network at any time after a dam break. Since the above variables have multiple dimensions, long time spans, and complex spatial distributions, they are difficult to use directly for risk analysis. Therefore, it is necessary to construct an indicator compression and transformation mechanism for the propagation behavior of dam-break floods, mapping the continuous propagation process into risk response indicators with clear engineering semantics.

[0146] According to one aspect of this application, step S41 further comprises:

[0147] Step S41a: Set dangerous water depth thresholds based on water network node types and regional water conservancy disaster prevention specifications; retrieve flood time series prediction data based on dangerous water depth thresholds; calculate the time index of the risk triggering condition for the first arrival of flood; and extract the hydraulic parameters at that moment to construct the first impact index vector.

[0148] Step S41b: Based on the initial impact index vector, extract and quantify the arrival response index, extreme impact index and cumulative effect index respectively. After the three types of indexes are processed by interval standardization, introduce risk contribution weight to uniformly map them into the node comprehensive impact intensity index.

[0149] In this embodiment, a dangerous water depth threshold h_th is introduced for the i-th water network node to characterize the triggering condition for a flood to have a substantial risk impact on the node. The time index T_i and the impact index vector I_i for the first arrival of the flood to trigger the risk are defined as follows:

[0150] I_i=[T_i, H_i^max, V_i^max, D_i];

[0151] T_i=min{t|h*_i(t)≥h_th};

[0152] The above formula represents the time when the flood first reaches the dangerous water depth threshold h_th at node i, where: t is the time variable; h*_i(t) is the predicted water depth of node i at time t; h_th is the dangerous water depth threshold, given by specifications or experience; T_i is the time when the flood reaches node i, reflecting the length of the warning and evacuation time.

[0153] To reflect the instantaneous impact intensity of flood on nodes during its propagation, the maximum water depth and maximum flow velocity that may occur at the nodes throughout the entire evolution process are defined as follows:

[0154] H_i^max=max_h*_i(t);

[0155] V_i^max=max_v*_i(t);

[0156] Let v*_i(t) represent the maximum water depth and maximum flow velocity of node i during the entire flood process, respectively. In the formula: v*_i(t) is the predicted flow velocity of node i at time t; H_i^max reflects the degree of inundation; and V_i^max reflects the scouring and destructive capacity.

[0157] Considering that flood risk is related not only to the peak value but also to the duration of the hazardous state, a threshold cumulative effect time index D_i is defined for node 𝑖:

[0158] D_i=∫I(h*_i(t)≥h_th)dt;

[0159] D_i represents the cumulative duration for which the water depth at node i exceeds the danger threshold. In the formula, I(⋅) is an indicator function, which takes the value 1 if the condition is met and 0 otherwise. This index is used to characterize the cumulative duration for which the flood is in a dangerous state at the node, thereby reflecting the cumulative risk effects of long-term inundation, structural softening, and personnel retention.

[0160] Based on this, the arrival response index, extreme impact index, and cumulative effect index are uniformly mapped to the nodal comprehensive impact intensity index S_i:

[0161] S_i=φ(I_i)=φ(T_i, H_i^max, V_i^max, D_i);

[0162] In the formula: S_i is the comprehensive impact intensity of node i; φ(⋅) is the impact index combination function, which is used to reflect that the deeper the water, the greater the flow velocity, the longer the duration, and the faster the arrival, that is, the stronger the impact of the flood, so that the obtained comprehensive impact intensity index has the ability to express time sensitivity, dynamic sensitivity and continuous risk.

[0163] Through the above-mentioned indicator conversion process, the complex propagation process of dam-break floods in the water network can be compressed into a set of risk response indicators with clear engineering semantics and risk orientation, providing a unified quantitative basis for risk comparison, hierarchical assessment and decision optimization among different nodes, engineering facilities and population units.

[0164] According to one aspect of this application, step S42 further comprises:

[0165] Step S42a: For different types of projects within the water network structure, construct vulnerability functions corresponding to the facility types, and calculate the probability of facility failure under different flood impact conditions;

[0166] Step S42b: Construct a population exposure risk model based on population size and calculate the exposure risk of population units.

[0167] The obtained impact intensity indicators of each node are further transformed into results of engineering facility damage risk, population safety risk and comprehensive risk level, realizing quantitative analysis from flood propagation intensity to risk management decision information.

[0168] For different types of projects (such as sluice gates, pumping stations, bridges, roads, and power facilities) within the water network structure, a vulnerability function F_k(⋅) corresponding to the facility type is constructed to describe the probability of facility failure under different flood impact conditions. Unlike the traditional fixed vulnerability model, the vulnerability response function is not only related to the flood intensity, but also explicitly introduces the characteristics of the flood propagation process as a modulation factor, thereby reflecting the differentiated impact of the speed, intensity, and duration of the flood on the probability of facility failure.

[0169] The calculation process for the failure probability of facility k is as follows:

[0170] P_f, k=F_k(H_k^max, V_k^max, D_k);

[0171] In the formula: P_f,k is the failure probability of facility k; H_k^max is the maximum water depth in the area where the facility is located; V_k^max is the corresponding maximum flow velocity; D_k is the duration of exceeding the dangerous water depth; F_k(⋅) is a vulnerability function pre-determined according to the facility type, construction standards and flood control capacity. It is used to characterize the differentiated vulnerability response of different engineering objects to flood propagation characteristics.

[0172] For population units, considering population size, a population exposure risk model is constructed by introducing the coupled effects of flood arrival time and water depth intensity.

[0173] R_pop, k=P_kψ(H_k^max, T_k);

[0174] In the formula: R_pop,k is the exposure risk of population unit k; P_k is the population size in the population unit; H_k^max is the maximum possible water depth; T_k is the time it takes for the flood to reach the unit; ψ(⋅) is the population exposure risk function, which reflects that the deeper the water and the faster it arrives, the higher the risk.

[0175] According to one aspect of this application, step S43 further comprises:

[0176] Step S43a: Combine the failure risk of engineering facilities with the exposure risk of the population to construct a comprehensive risk index;

[0177] Step S43b: Based on the magnitude of the comprehensive risk value, classify the engineering facilities and population units into risk levels of extremely high risk, high risk, medium risk and low risk, and output the spatial distribution map of high-risk objects, the list of engineering facilities and population units to be prioritized and monitored, and obtain risk ranking and control recommendations for emergency response.

[0178] By combining the risk of engineering facility failure with the risk of population exposure, a comprehensive risk indicator can be constructed.

[0179] R_k = w1P_f, k + w2R_pop, k;

[0180] In the formula: R_k is the comprehensive risk value of object k; P_f, k is the probability of facility failure; R_pop, k is the population exposure risk; w1, w2 are weighting coefficients, reflecting the differences in the importance of risks.

[0181] Unlike traditional risk-weighted methods, this comprehensive risk index retains the interpretability of risk sources and can distinguish whether the risk mainly originates from engineering failure, population exposure, or a combination of both, thus providing a basis for differentiated risk management and resource allocation.

[0182] Based on the magnitude of the comprehensive risk value R_k, engineering facilities and population units are classified into risk levels: extremely high risk, high risk, medium risk, and low risk. The output includes: a spatial distribution map of high-risk objects; a list of engineering facilities and population units that require priority handling and key monitoring; and risk ranking and control recommendations for emergency response.

[0183] Through the above steps, the propagation results of dam-break floods are transformed from "physical prediction" to a systematic approach of "risk identification, risk ranking, and risk management," providing direct support for emergency decision-making and risk management.

[0184] According to another aspect of this application, a smart risk assessment system for dam-break flood propagation at water network nodes is provided, characterized in that it includes:

[0185] At least one processor; and

[0186] A memory communicatively connected to at least one of the processors; wherein,

[0187] The memory stores instructions that can be executed by the processor to implement the intelligent risk assessment method for dam-break flood propagation at water network nodes as described above.

[0188] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for intelligent risk assessment of dam-break flood propagation at water network nodes, characterized in that, Includes the following steps: Step S1: Collect basic data on water network engineering and dam failure scenario parameters, and generate a standardized dataset; Step S2: Construct a directed topological model of water network nodes and channels that integrates engineering semantic constraints and hydraulic propagation direction constraints, and generate a water network adjacency matrix; Step S3: Use a neural network model based on the constraints of the water network structure and the physical knowledge of mass conservation and momentum conservation to predict the propagation process of dam-break flood in the water network; Step S4: Convert the flood propagation prediction results into flood impact indicators, calculate the risk level of engineering facilities and population units based on the flood impact indicators, and output the risk assessment results and decision support information.

2. The intelligent risk assessment method for dam-break flood propagation at water network nodes as described in claim 1, characterized in that, Step S1 further comprises: Step S11: Collect basic operational data and dam failure scenario parameters of reservoirs, dams, rivers and water diversion projects in the water network system. The basic operational data includes the initial water level of the reservoir, reservoir capacity, outflow and river hydraulic parameters. The dam failure scenario parameters include the outflow process and boundary condition information of the breach. Step S12: Unify the coordinate system and time step of the basic operation data and dam failure scenario parameters of the collected water network system, including reservoir dams, rivers and water diversion projects, and unify and normalize the dimensions of water level, outflow and reservoir capacity data to generate a standardized dataset.

3. The intelligent risk assessment method for dam-break flood propagation at water network nodes as described in claim 1, characterized in that, Step S2 further comprises: Step S21: Construct a directed topological model of water network nodes and channels based on engineering semantic constraints and hydraulic propagation direction constraints; Step S22: Input the standardized dataset into the directed topology model of water network nodes-channels to generate the water network adjacency matrix.

4. The intelligent risk assessment method for dam-break flood propagation at water network nodes as described in claim 1, characterized in that, Step S3 further comprises: Step S31: Construct a neural network model based on water network structure constraints and the conservation of mass and momentum; Step S32: Extract the prediction time, water network adjacency matrix, reservoir initial state, hydraulic attribute matrix, external boundary conditions, and breach discharge process as inputs to the neural network model. Under the constraints of water network structure and physical knowledge, perform time-series prediction of the propagation process of dam-break flood in the water network, and obtain the water level, water depth, and flow velocity of each node in the water network at any prediction time.

5. The intelligent risk assessment method for dam-break flood propagation at water network nodes as described in claim 1, characterized in that, Step S4 further comprises: Step S41: Set the dangerous water depth threshold, calculate the time index of the first arrival of the flood to trigger the risk conditions, and convert the flood propagation prediction results into flood impact indicators; Step S42: Calculate the risk level of engineering facilities and population units based on flood impact indicators; Step S43: Output risk assessment results and decision support information.

6. The intelligent risk assessment method for dam-break flood propagation at water network nodes as described in claim 5, characterized in that, Step S41 further comprises: Step S41a: Set dangerous water depth thresholds based on water network node types and regional water conservancy disaster prevention specifications; retrieve flood time series prediction data based on dangerous water depth thresholds; calculate the time index of the risk triggering condition for the first arrival of flood; and extract the hydraulic parameters at that moment to construct the first impact index vector. Step S41b: Based on the initial impact index vector, extract and quantify the arrival response index, extreme impact index and cumulative effect index respectively. After the three types of indexes are processed by interval standardization, introduce risk contribution weight to uniformly map them into the node comprehensive impact intensity index.

7. The intelligent risk assessment method for dam-break flood propagation at water network nodes as described in claim 5, characterized in that, Step S42 further comprises: Step S42a: For different types of projects within the water network structure, construct vulnerability functions corresponding to the facility types, and calculate the probability of facility failure under different flood impact conditions; Step S42b: Construct a population exposure risk model based on population size and calculate the exposure risk of population units.

8. The intelligent risk assessment method for dam-break flood propagation at water network nodes as described in claim 5, characterized in that, Step S43 further comprises: Step S43a: Combine the failure risk of engineering facilities with the exposure risk of the population to construct a comprehensive risk index; Step S43b: Based on the magnitude of the comprehensive risk value, classify the engineering facilities and population units into risk levels of extremely high risk, high risk, medium risk and low risk, and output the spatial distribution map of high-risk objects, the list of engineering facilities and population units to be prioritized and monitored, and obtain risk ranking and control recommendations for emergency response.

9. A smart risk assessment system for dam-break flood propagation at water network nodes, characterized in that, include: At least one processor; as well as A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor to implement the intelligent risk assessment method for dam-break flood propagation at water network nodes as described in any one of claims 1 to 8.

Images