Multi-target particle swarm optimization scheduling system for flood control and waterlogging removal in polder area of plain river network

Abstract

The invention relates to the field of process control, in particular to a multi-target particle swarm optimization scheduling system for flood control and waterlogging removal in a plain river network polder area. The method comprises the specific implementation steps that a coupling hydrodynamic model module is constructed, the water flow exchange amount between a one-dimensional hydrodynamic model and a two-dimensional hydrodynamic model is calculated, multiple times of simulation are carried out to generate a training data set, and a neural network agent model is trained and generated; generating an initial population based on topological structure information of the plain river network and a heuristic rule by adopting a particle swarm optimization algorithm; according to external work condition information, dynamically adjusting feasible region constraints of an optimization algorithm, inputting candidate scheduling schemes conforming to the constraints into the neural network agent model, and outputting a predicted scheduling state; based on the predicted scheduling state and the weight coefficient, the fitness value of each candidate scheduling scheme is calculated, the optimal scheduling scheme is output through iterative optimization, and the decision-making efficiency, robustness and safety of the scheduling system in the face of complex working conditions and emergencies are improved.

Description

Technical Field

[0001] The present invention relates to the field of process control, and in particular to a multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in plain river network embankment areas. Background Art

[0002] In order to protect the lives and property of people in plain river network areas, flood control and drainage scheduling technology around water conservancy projects has made great progress. In terms of scheduling basis, existing technologies have formed a set of effective static scheduling procedures and scheduling diagrams based on characteristic values ​​such as warning water levels and guaranteed water levels in key sections; in terms of engineering systems, water conservancy hubs such as sluices, pumping stations, and regulating gates throughout the river network constitute a solid physical foundation for flood control and drainage; in terms of information acquisition, the hydrological telemetry station network and meteorological forecast system can provide relatively timely water and rainfall information, providing data input for scheduling decisions.

[0003] However, when faced with basin-wide, high-intensity extreme rainfall events, deep-seated contradictions in the existing scheduling system are increasingly exposed. These challenges are particularly prominent from a process control perspective. Some scheduling decisions exhibit typical decentralized characteristics. The start-stop logic of each scheduling unit is based solely on its local state variables, without considering the systemic impact of its scheduling behavior on downstream and global states. This uncoordinated concurrent discharge can easily lead to severe flood peak superposition in local areas, rapidly jacking up external river levels and transforming local drainage needs into regional flood control crises. Some scheduling decisions also suffer from significant lags. Traditional scheduling models are mostly passive responses based on historical experience and established rules, lacking the ability to accurately predict and dynamically respond to future water regime trends, making it difficult to seize precious pre-discharge scheduling windows. With the increasing number of water conservancy projects, the entire scheduling system has become a complex process control object with high dimensions, nonlinearity, and strong coupling. The rapid increase in decision variables has led to an exponential increase in the computational complexity of finding the global optimal solution, known as the curse of dimensionality, far exceeding the capabilities of traditional optimization algorithms and manual scheduling.

[0004] To this end, the present invention proposes a multi-objective particle swarm optimization scheduling system for flood control and drainage in plain river network embankment areas. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-objective particle swarm optimization scheduling system for flood control and drainage in plain river network embankment areas, enabling rapid, coordinated, and globally optimal intelligent scheduling of all water conservancy facilities in the area under complex and dynamic flood conditions. Specifically, this system is achieved by constructing a hierarchical decision-making framework that integrates physical simulation, agent learning, and intelligent optimization. This framework deeply integrates high-fidelity simulation based on one-dimensional and two-dimensional coupled hydrodynamics, rapid deduction of neural network agent models, and a multi-objective particle swarm optimization algorithm guided by topological rules and dynamic weights.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A multi-objective particle swarm optimization scheduling system for flood control and drainage in plain river network embankment areas, comprising:

[0008] Coupled hydrodynamic model module, which calculates the water flow exchange between the one-dimensional hydrodynamic model used to simulate underground pipe network water flow and the two-dimensional hydrodynamic model used to simulate surface flow;

[0009] An agent model generation module, calling the coupled hydrodynamic model module to perform multiple simulations to generate a training data set, training the training data set and generating a neural network agent model;

[0010] The multi-objective weight adaptive module dynamically determines a set of weight coefficients representing the relative importance of multiple scheduling objectives based on the received real-time warning level;

[0011] The particle swarm optimization scheduling module generates initial candidate scheduling schemes as the initial population of the particle swarm based on the topological structure information of the plain river network and preset heuristic rules. During the iterative optimization process, the feasible domain constraints of the optimization algorithm are dynamically adjusted according to the external working situation information received in real time. The candidate scheduling schemes are input into the neural network agent model and the predicted scheduling status is output. Based on the predicted scheduling status and weight coefficient, the fitness value of each candidate scheduling scheme is calculated, and the optimal scheduling scheme is output through the iterative optimization of the particle swarm algorithm.

[0012] Preferably, the coupled hydrodynamic model module calculates the water flow exchange between a one-dimensional hydrodynamic model for simulating underground pipe network water flow and a two-dimensional hydrodynamic model for simulating surface overflow, including: the one-dimensional hydrodynamic model is a storm flood management model solved based on the Saint-Venant equations; the two-dimensional hydrodynamic model is a flood overflow model solved based on the two-dimensional non-steady-state shallow water equations; the calculation method of the water flow exchange includes: calculation of the vertical exchange volume, calculating the overflow caused by the head difference between the underground pipe network node and the surface two-dimensional grid through the orifice flow formula and the weir flow formula; calculation of the lateral exchange volume, taking the river embankment of the one-dimensional hydrodynamic model as the internal boundary of the two-dimensional hydrodynamic model, and calculating the lateral overflow flow generated when the river water level exceeds the embankment elevation through the weir flow formula.

[0013] Preferably, the proxy model generation module includes: calling the coupled hydrodynamic model module to simulate the working conditions covering different rainfall scenarios and scheduling scheme combinations to generate a training data set; the neural network proxy model includes an input layer, a hidden layer and an output layer; the input layer receives a scheduling scheme vector composed of the opening degree of each sluice and the start and stop status of the pump station; the output layer outputs a multidimensional prediction scheduling state vector composed of the highest water level of the key section in the area and the inundation area.

[0014] Preferably, the multi-objective weight adaptive module includes: a scenario-strategy mapping rule base is stored inside the multi-objective weight adaptive module; wherein each scenario is represented by a warning level defined by real-time rainfall intensity and external river tide data; each strategy corresponds to a set of weight coefficient combinations preset for the two goals of flood control and drainage; real-time external incoming real-time rainfall intensity and real-time external river tide data representing the current warning level are received in real time; the received real-time data is matched one by one with multiple preset scenarios in the rule base to determine the interval defined by multiple closest preset scenarios into which the current real-time data falls; when the real-time data accurately matches a preset scenario, the module directly outputs the preset strategy weight corresponding to the scenario; when the real-time data falls between two preset scenarios, the module generates and outputs a set of dynamic weight coefficients applicable to the current scenario by performing linear interpolation calculation on the preset strategy weights corresponding to the two preset scenarios.

[0015] Preferably, the particle swarm optimization scheduling module generates initial candidate scheduling schemes as the initial population of the particle swarm based on the topological structure information of the plain river network and preset heuristic rules, including: obtaining topological structure information of a plain river network represented by nodes and edges, which includes the upstream and downstream relationships of each water conservancy facility; applying preset heuristic rules to perform path traversal and priority sorting on the topological structure; the heuristic rules include spatial priority rules, which determine the spatial execution order of scheduling operations based on the upstream and downstream relationships in the topological structure; the heuristic rules also include temporal priority rules, which determine the temporal sequence of scheduling operations based on the division of main streams and tributaries in the topological structure; based on the determined spatial execution order and temporal sequence, a group of candidate scheduling schemes with random perturbations in specific operation parameters and that overall conform to the heuristic rules are generated as the initial population.

[0016] Preferably, in the iterative optimization process, the feasible domain constraints of the optimization algorithm are dynamically adjusted according to the external working condition information received in real time, the candidate scheduling schemes are input into the neural network proxy model and the predicted scheduling state is output, including: receiving external working condition information including real-time fault alarm information of water conservancy project facilities, dynamically adjusting the value range of the corresponding decision variables in the optimization algorithm as the feasible domain constraints; checking the candidate scheduling schemes newly generated by the particle swarm algorithm, and performing boundary repair processing on the schemes that exceed the feasible domain constraints; using the neural network proxy model as the fitness function evaluation engine of the particle swarm algorithm, and transmitting the candidate scheduling schemes generated by the particle swarm algorithm in each iteration as the input vector to the input layer of the neural network proxy model; the neural network proxy model performs a forward propagation calculation instead of calling the coupled hydrodynamic model to perform a complete and time-consuming physical simulation; and the output layer of the neural network proxy model outputs a predicted scheduling state vector that can characterize the multi-dimensional physical state of the highest water level of the key section in the region and the inundated area under the candidate scheduling scheme.

[0017] Preferably, based on the predicted scheduling state and weight coefficient, the fitness value of each candidate scheduling scheme is calculated, and the optimal scheduling scheme is output through the iterative optimization of the particle swarm algorithm, including: performing weighted summation on the weight coefficient and the predicted scheduling state vector to calculate the scalar fitness value of each candidate scheduling scheme; each candidate scheduling scheme is regarded as a particle, and the speed and position of the particle in the solution space are updated according to the scheme with the optimal fitness value corresponding to the particle's own historical iteration and the scheme with the optimal fitness value corresponding to the historical iteration of the entire particle population, to generate a new candidate scheduling scheme; after the preset termination condition is met, the scheme with the optimal fitness value corresponding to the current entire particle population historical iteration is output as the optimal scheduling scheme.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. The present invention constructs a coupled hydrodynamic model as a high-fidelity physical world simulation engine, and adopts a neural network proxy model to learn and replace the coupled hydrodynamic model, thereby realizing a fast and efficient approximate deduction of the complex working conditions of the plain river network embankment area. It can not only comprehensively and dynamically simulate the complex water flow exchange process between surface flow and underground pipe network, but also compress the physical simulation time that originally required several hours to seconds, ensuring that subsequent optimization decisions have a solid scientific basis that reflects the real physical world; at the same time, the present invention uses the proxy model as a fast fitness function evaluation engine of the optimization algorithm, which improves the system's real-time response capability and forward-looking decision-making capability under rapidly changing flood conditions, and meets the current flood control and drainage scheduling requirements for high timeliness.

[0020] 2. The present invention introduces heuristic rules based on the topological structure information of the entire river network and hydraulic common sense during the initialization phase of the particle swarm optimization algorithm, thereby achieving intelligent guidance of the optimization algorithm's search starting point. This not only avoids the blind random search of traditional optimization algorithms in a huge solution space, but also concentrates computing resources on areas where the optimal solution is more likely to exist physically from the first step of the algorithm startup. This topology-aware initialization mechanism significantly improves the algorithm's convergence speed and the success rate of finding the global optimal solution, effectively solving the "dimensionality curse" problem in high-dimensional and complex scheduling problems.

[0021] 3. The present invention realizes the intelligent balance of multiple and conflicting scheduling objectives by constructing an adaptive module that can dynamically adjust the weights of multiple objectives according to the real-time warning level, and uses this dynamic weight to guide the iteration direction of the particle swarm optimization algorithm. It not only enables the system to automatically switch the core management objectives at different flood stages, but also can convert a difficult-to-solve multi-objective optimization problem into a single-objective problem with a clear optimization direction in real time. This scenario-adaptive decision-making paradigm ensures that the final output scheduling plan can maximize the comprehensive disaster reduction efficiency of the entire region under the current specific working conditions, thereby improving the intelligence level and practical applicability of scheduling decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Schematic diagram of the overall architecture of a multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in a plain river network embankment area according to an embodiment of the present invention;

[0023] Figure 2 Schematic diagram of the coupling mechanism of the coupled hydrodynamic model according to an embodiment of the present invention;

[0024] Figure 3 Schematic diagram of the workflow of the particle swarm optimization scheduling module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose of the present invention, technical solutions and the beneficial effects produced more fully and clearly understood, the technical solutions of the present invention will be described in detail below in conjunction with the specific embodiments of the present invention. It should be pointed out that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of protection of the present invention. Those of ordinary skill in the art can make various equivalent modifications or substitutions based on these embodiments without departing from the concept and scope of the present invention, and these modifications and substitutions should all be deemed to fall within the scope of protection of the present invention.

[0026] The present invention provides a multi-objective particle swarm optimization scheduling system for flood control and drainage in plain river network areas. The core of the system is to achieve accurate and efficient management of complex water conservancy systems by building a hierarchical and modular intelligent decision-making framework. Figure 1,The system mainly includes the following four closely cooperating ,functional modules: coupled hydrodynamic model module, proxy model generation ,module, multi-objective weight adaptation module, and particle swarm optimization ,scheduling module.

[0027] Example 1

[0028] A multi-objective particle swarm optimization scheduling system for flood control and drainage in plain river network embankment areas, comprising:

[0029] Coupled hydrodynamic model module, which calculates the water flow exchange between the one-dimensional hydrodynamic model used to simulate underground pipe network water flow and the two-dimensional hydrodynamic model used to simulate surface flow;

[0030] An agent model generation module, calling the coupled hydrodynamic model module to perform multiple simulations to generate a training data set, training the training data set and generating a neural network agent model;

[0031] The multi-objective weight adaptive module dynamically determines a set of weight coefficients representing the relative importance of multiple scheduling objectives based on the received real-time warning level;

[0032] The particle swarm optimization scheduling module generates initial candidate scheduling schemes as the initial population of the particle swarm based on the topological structure information of the plain river network and preset heuristic rules. During the iterative optimization process, the feasible domain constraints of the optimization algorithm are dynamically adjusted according to the external working situation information received in real time. The candidate scheduling schemes are input into the neural network agent model and the predicted scheduling status is output. Based on the predicted scheduling status and weight coefficient, the fitness value of each candidate scheduling scheme is calculated, and the optimal scheduling scheme is output through the iterative optimization of the particle swarm algorithm.

[0033] Further, refer to Figure 2 The coupled hydrodynamic model module calculates the water flow exchange between the one-dimensional hydrodynamic model for simulating underground pipe network water flow and the two-dimensional hydrodynamic model for simulating surface flow, including:

[0034] The one-dimensional hydrodynamic model is a Stormwater Flood Management Model (SWMM) that solves the Saint-Venant equations. The calculation of water flow in pipes and rivers is based on solving the complete one-dimensional Saint-Venant equations, including the continuity equation and the momentum equation.

[0035] The continuity equation describes the law of conservation of mass, which states that the amount of water flowing into and out of a control volume must be balanced.

[0036] The momentum equation characterizes the relationship between forces acting on a body of water (such as gravity, pressure, and friction) and the rate of change of that water's momentum. It accurately simulates complex hydraulic phenomena such as water storage, backwater support, and pressure flow. Frictional losses in the flow are calculated based on the Manning equation, with their magnitude proportional to the square of the Manning roughness coefficient and the square of the flow velocity, and inversely proportional to the hydraulic radius (four-thirds power). By numerically solving the Saint-Venant equations, the one-dimensional hydrodynamic model can output water level and flow at any section in pipes and rivers.

[0037] The two-dimensional hydrodynamic model is a flood and overland flow model based on the two-dimensional unsteady shallow water equation. This model discretizes the surface area into a series of square computational grids and solves the water exchange between each grid and its four adjacent grids. The core calculations of the model are also based on the continuity equation (to ensure water balance within each grid) and the momentum equation. In the two-dimensional model, the exchange flow between adjacent grids is also calculated based on the principles of Manning's equation, and its magnitude depends on the head difference between the two grids, the cross-sectional area of ​​the water flow, and the roughness coefficient. By solving the two-dimensional unsteady shallow water equation, the two-dimensional hydrodynamic model can simulate the two-dimensional overland flow, evolution, and retreat of flood water on the surface.

[0038] The coupled hydrodynamic model module also implements a two-way dynamic interaction method with mutual boundary conditions. Within each coupled calculation time step, the system performs the following data exchange and calculations in parallel:

[0039] Calculating and outputting the weighted average water level value of all two-dimensional grids at the downstream outlet connection section of the one-dimensional river model using the two-dimensional hydrodynamic model;

[0040] The one-dimensional hydrodynamic model receives the water level value and uses it as a dynamically updated fixed water level lower boundary condition for calculation in the next time step.

[0041] Calculate and output the total flow rate out of the downstream outlet using the one-dimensional hydrodynamic model;

[0042] The two-dimensional hydrodynamic model receives this flow value and uses it as an upper boundary condition of a flow source term, and distributes it to multiple surface grids at the connecting sections as the initial inflow for the surface flow calculation in the next time step.

[0043] This embodiment uses a real-time two-way boundary interaction mechanism to simulate the complex backwater jacking effect that cannot be accurately captured by one-way coupling or simplified connection methods. It improves the fundamental technical problem of traditional methods that overestimate the system drainage capacity and thus cause decision-making errors due to the inability to simulate such two-way feedback. This is also the key basis for the present invention to achieve high-fidelity simulation.

[0044] The calculation method of the water exchange amount includes:

[0045] The vertical exchange volume is calculated between the nodes such as the inspection wells in the SWMM model and the corresponding surface grids in the two-dimensional model. When the water level at the network node is higher than the surface grid water level, the overflow from the network to the surface is calculated using the orifice outflow formula. The flow rate is proportional to the square root of the head difference. When the surface water level is higher than the network node water level, the return flow from the surface to the network is calculated using the weir flow formula. The flow rate is proportional to the cube of the head difference.

[0046] For the calculation of lateral exchange, when the water level calculated by the one-dimensional river channel model exceeds the preset embankment elevation, the river channel embankment is regarded as a broad-crowned weir, and the resulting lateral overland flow is calculated using the weir flow formula and input into the two-dimensional overland flow model as a source term.

[0047] This embodiment couples and solves the Saint-Venant equations of a one-dimensional hydrodynamic model and the shallow water equations of a two-dimensional hydrodynamic model, and clearly defines the vertical and lateral exchange mechanisms based on the physical formulas of orifice and weir flow. It constructs a simulation engine that can reproduce the real physical world with high fidelity, solving the fundamental problem that a single-dimensional model cannot accurately simulate the multi-scale and cross-domain characteristics of urban flooding.

[0048] Furthermore, the proxy model generation module includes:

[0049] The coupled hydrodynamic model module is called to simulate operating conditions covering different rainfall scenarios and scheduling scheme combinations to generate a training data set; the neural network agent model includes an input layer, a hidden layer and an output layer; the input layer receives a scheduling scheme vector consisting of the opening degree of each sluice and the start and stop status of the pump station; the output layer outputs a multidimensional predicted scheduling state vector consisting of the highest water level of the key section in the area and the inundated area.

[0050] The proxy model generation module calls the coupled hydrodynamic model module to perform offline simulations of operating conditions covering different combinations of rainfall scenarios and scheduling schemes to generate a large-scale training data set. The rainfall scenarios cover typical rainstorm events of different intensities, durations, and spatial distributions. The scheduling scheme combinations include multiple opening settings for the sluice gates (for example, discrete gears or continuous percentages from fully closed to fully open) and the start and stop states and operating power configurations of the pump stations (for example, binary start and stop states or multi-level power output). Sufficiently diverse and representative operating conditions are generated through systematic sampling methods (such as Latin hypercube sampling). The input (rainfall scenario, scheduling scheme) and output (highest water level at key sections in the area, inundated area, etc.) of each simulation constitute a training sample. These data will be used for subsequent supervised learning of the neural network.

[0051] The constructed neural network proxy model includes an input layer, a hidden layer, and an output layer. In this embodiment, a multilayer perceptron is preferably used as the proxy model. The input layer receives a scheduling plan vector consisting of the opening degrees of each sluice gate (e.g., a normalized continuous value between 0 and 1 or a discrete code) and the start and stop status of the pump station (e.g., a binary variable of 0 / 1 or a discrete code representing multiple power output levels). The dimension of the input vector is equal to the total number of scheduling variables for all sluice gates and pump stations. The output layer outputs a multidimensional predictive scheduling state vector consisting of the highest water level at key sections within the region and the inundation area. These output values ​​serve as key indicators for the particle swarm optimization algorithm to evaluate the fitness of the scheduling plan. The dimension of the output vector is equal to the number of key sections to be predicted plus the inundation area. The model includes multiple hidden layers to learn the nonlinear relationship between the input scheduling plan and the complex hydrodynamic response. The specific number of layers and the number of neurons in each layer can be adjusted based on model complexity and data volume. For example, two or three hidden layers can be used with a decreasing number of neurons (e.g., 64 neurons in the first hidden layer and 32 neurons in the second hidden layer). These hidden layers use rectified linear units as nonlinear activation functions to learn and fit the deep nonlinear mapping relationship between the scheduling scheme and the complex hydrodynamic response.

[0052] The training process uses supervised learning, using the generated training dataset to train the neural network. The Adam adaptive learning rate optimizer is used as the optimizer, with a learning rate set to 0.001. The mean squared error loss function is used to minimize the deviation between the surrogate model's predictions and the actual simulation results of the coupled hydrodynamic model. Training continues until the model's performance on the validation set reaches convergence or the maximum number of training cycles is reached.

[0053] The proxy model generation module also adopts a phased training method when training the neural network proxy model:

[0054] In the first stage, preliminary training is performed: from the training data set, operating condition data containing only single water conservancy facility scheduling or simple dual-facility linkage is selected to form a first training subset; using the first training subset, the randomly initialized neural network agent model is trained to learn and fit the basic causal relationship between scheduling behavior and hydrodynamic response;

[0055] The second phase involves in-depth training: After the initial training is complete, the trained neural network proxy model is further fine-tuned and optimized using a complete training dataset containing all complex multi-facility joint scheduling scenarios.

[0056] The method may also include a cross-regional parameter migration step for constructing a proxy model for the new region, using part or all of the weight parameters of a neural network proxy model that has been trained in the source region as the initial weight parameters for training the new regional proxy model to be constructed; using a small amount of simulated data from the new region to fine-tune the new regional proxy model loaded with the initial weight parameters.

[0057] This embodiment decomposes the learning process of complex nonlinear mapping relationships by adopting this staged training method. Through the cross-regional parameter migration and fine-tuning mechanism, it greatly reduces the repeated simulation and training calculations required when building proxy models for new regions. It improves the technical bottleneck of needing to conduct time-consuming, from-scratch model training for each new research area in actual engineering applications, and significantly improves the scalability and engineering practicality of the entire solution.

[0058] This embodiment constructs a neural network proxy model to achieve rapid and efficient approximate deduction of complex working conditions in plain river network embankment areas, compressing the physical simulation time that originally required several hours to seconds. This significantly improves the system's real-time response capability and forward-looking decision-making capabilities under rapidly changing flood conditions, and meets the current demand for high timeliness in flood control and drainage scheduling.

[0059] Furthermore, the multi-objective weight adaptive module includes:

[0060] The multi-objective weight adaptive module internally stores a scenario-strategy mapping rule base, which encodes historical experience and expert knowledge in a structured manner. Each scenario is characterized by a warning level defined by real-time rainfall intensity (for example, in millimeters per hour) and external river tide data (for example, in meters). These warning levels can be scientifically divided according to regional flood control and drainage standards, historical hydrological data and expert experience. For example, they are divided into "blue warning" (light rainfall, low tide, that is, the real-time rainfall intensity is less than 5 mm per hour and the external river tide level is 0.5 meters below the base level), "yellow warning" (moderate rainfall, medium tide, that is, the real-time rainfall intensity is between 5 mm per hour and 15 mm per hour, or the external river tide level is 0.5 meters below the base level), "yellow warning" (moderate rainfall, medium tide, that is, the real-time rainfall intensity is between 5 mm per hour and 15 mm per hour, or the external river tide level is 0.5 meters below the base level), "red ... The strategies include "orange alert" (heavy rainfall, high tide level, i.e., real-time rainfall intensity between 15 mm and 30 mm per hour, or external river tide level between 1.5 and 2.5 meters above datum), and "red alert" (extremely heavy rainfall, super-high tide level, i.e., real-time rainfall intensity exceeding 30 mm per hour, or external river tide level exceeding 2.5 meters above datum). Each strategy corresponds to a combination of pre-set weighting coefficients for the two main objectives of flood control and drainage, and may also include water resource utilization objectives, where the sum of all weighting coefficients is 1. For example, during light rainfall and low tide, the weight of water resource utilization may be higher (e.g., 0.4), while the weights of flood control and drainage may be lower (0.3 each). During heavy rainfall and high tide, the weights of flood control and drainage will be significantly increased (e.g., 0.8 for flood control and 0.15 for drainage), while the weight of water resource utilization will be reduced (e.g., 0.05).

[0061] When the system is running in real time, the dynamic determination process of the multi-objective weight adaptive module includes: receiving the real-time rainfall intensity and real-time external river tide data that represent the current warning level in real time; matching the received real-time data with multiple preset scenarios in the rule base one by one to determine the interval defined by multiple closest preset scenarios into which the current real-time data falls; when the real-time data accurately matches a preset scenario, the module directly outputs the preset strategy weight corresponding to the scenario; when the real-time data falls between two preset scenarios (for example, "yellow warning" and "orange warning"), the module performs linear interpolation calculations on the preset strategy weights corresponding to the two preset scenarios to generate and output a set of dynamic weight coefficients applicable to the current scenario. Specifically, the weights are obtained as follows:

[0062] When the system is running in real time, the module first receives external real-time rainfall intensity and real-time external river tide data that represent the current warning level, and then matches this set of real-time data with multiple preset scenarios in the rule base one by one.

[0063] When the real-time data accurately matches the definition of a preset scenario (for example, the current rainfall intensity is 20 mm / hour, which falls exactly within the range of "orange warning"), the module will directly output the preset strategy weight corresponding to the scenario (for example, directly returning a flood control weight of 0.8 and a drainage weight of 0.15).

[0064] When real-time data falls between the quantized intervals of two pre-set scenarios (for example, the current rainfall intensity is 14.5 mm / hour, on the boundary between "Yellow Alert" and "Orange Alert"), the module initiates a linear interpolation calculation mechanism to ensure a smooth transition of weight switching and avoid sudden changes in the scheduling policy. This mechanism first obtains the preset policy weights corresponding to the two adjacent pre-set scenarios ("Yellow Alert" and "Orange Alert"), then performs linear interpolation calculations based on the relative position of the current real-time data between the two scenario boundaries, generating and outputting a set of dynamic weight coefficients that are most appropriate for the current specific scenario.

[0065] This embodiment achieves intelligent trade-offs between multiple, conflicting scheduling objectives by constructing an adaptive module that dynamically adjusts multi-objective weights based on real-time warning levels. This dynamic weighting is then used to guide the iteration direction of the particle swarm optimization algorithm. This enables the system to automatically switch core management objectives at different flood stages, transforming a difficult multi-objective optimization problem into a single-objective problem with a clear optimization direction in real time. This scenario-adaptive decision-making paradigm ensures that the final output scheduling solution maximizes the comprehensive disaster reduction efficiency of the entire region under current specific operating conditions, improving the intelligence and practical applicability of scheduling decisions.

[0066] Further, refer to Figure 3 The particle swarm optimization scheduling module generates an initial candidate scheduling scheme as the initial population of the particle swarm based on the topological structure information of the plain river network and the preset heuristic rules, including:

[0067] The particle swarm optimization scheduling module implements a dual intelligent guidance optimization method:

[0068] In the optimization startup phase, an initialization bootstrapping step based on topological rules is executed. This step generates a set of initial candidate scheduling solutions that are biased towards the physically feasible solution region in the solution space based on the topological structure information of the plain river network and the preset heuristic rules, which serve as the initial population of the particle swarm.

[0069] In the optimization iteration stage, a fitness guidance step based on dynamic weighting is executed. This step determines the optimization direction of each iteration by weighted summing the weight coefficient output by the multi-objective weight adaptation module and the prediction effect output by the neural network agent model.

[0070] This embodiment adopts this dual intelligent guidance mechanism to deeply integrate domain expert knowledge and real-time scenario judgment into the inner and outer loops of the optimization algorithm. Initialization guidance improves the defects of traditional particle swarm optimization algorithms such as slow convergence and waste of computing resources caused by blind random initialization when facing high-dimensional complex problems by constraining the starting point of the search; fitness guidance dynamically adjusts the optimization goal, improving the technical bottleneck of the algorithm easily falling into local optimality due to vague evaluation criteria and unclear search direction in multi-objective conflict scenarios.

[0071] Obtain the topological structure information of a plain river network represented by nodes (representing water conservancy facilities or key river sections) and edges (representing river channels or pipeline connections), which includes the upstream and downstream relationships of various water conservancy facilities (such as sluices, pumping stations, and regulating gates), including the geographical location, connection relationship, control range, etc. of each facility. This information is usually stored in the form of a graph database or adjacency matrix.

[0072] Applying preset heuristic rules, path traversal and priority sorting are performed on the topological structure to generate an initial population with physical rationality, wherein the path traversal can be achieved by performing a graph traversal algorithm (such as breadth-first search or depth-first search) on the topological graph of the water conservancy facilities of the plain river network to identify the upstream and downstream connection relationships between the facilities. The heuristic rules include spatial priority rules, which determine the spatial execution order of scheduling operations based on the upstream and downstream relationships in the topological structure. For example, in the flood prevention stage, upstream facilities are usually prioritized to control inflow, or downstream facilities are prioritized to ensure smooth outflow; in the drainage stage, drainage facilities in low-lying areas may be opened first; the heuristic rules also include time priority rules, which determine the time sequence of scheduling operations based on the division of main streams and tributaries in the topological structure. For example, before the flood comes, tributaries may be prioritized for pre-discharge to empty the reservoir capacity; during the flood, the tributaries may be prioritized for pre-discharge to empty the reservoir capacity; during the flood, the tributaries may be prioritized for pre-discharge to empty the reservoir capacity. When the peak passes, the main stream facilities are dispatched in priority to ensure the smooth flow of the main drainage channel; based on the determined spatial execution order and time sequence, a group of candidate scheduling schemes with random perturbations on specific operating parameters (such as sluice opening, pump station start and stop time, and operating power) that generally conform to the heuristic rules are generated as the initial population, wherein the random perturbation refers to the random adjustment of specific operating parameters such as sluice opening and pump station start and stop status within their designed operating range on the basis of the initial scheme generated by the heuristic rules, so as to increase the diversity of the initial population.

[0073] This embodiment introduces heuristic rules based on the entire river network's topological structure and common hydraulic knowledge during the initialization phase of the particle swarm optimization algorithm. This intelligently guides the optimization algorithm's search starting point, avoiding the blind, random search of the vast solution space often encountered by traditional optimization algorithms. From the very first step of the algorithm's initiation, computing resources are focused on areas where the optimal solution is physically more likely to exist. This topology-aware initialization mechanism significantly improves the algorithm's convergence speed and the success rate of finding the global optimal solution, effectively addressing the "curse of dimensionality" problem in high-dimensional, complex scheduling problems.

[0074] Furthermore, during the iterative optimization process, the feasible domain constraints of the optimization algorithm are dynamically adjusted according to the external working situation information received in real time, the candidate scheduling solutions are input into the neural network agent model and the predicted scheduling status is output, including:

[0075] The system first receives real-time external operational status information from an external monitoring system or a higher-level scheduling platform. This information can include a real-time fault alarm signal from a pump due to power supply problems or a status feedback indicating that a sluice gate cannot reach its maximum opening due to a mechanical failure. Upon receiving this information, the particle swarm optimization scheduling module dynamically adjusts the internally defined feasible domain used to constrain decision variables. Specifically, if a fault shutdown signal is received for "Main Drain Pump Station 2," the algorithmically enforces the range of the decision variable corresponding to the operating status of Pump Station 2 in the particle's position vector from its original allowable operating range to a fixed value, for example, [0, 0], indicating its complete shutdown. After completing the dynamic update of the feasible domain, the system examines each newly generated candidate scheduling solution. If any dimension is found to be outside the boundaries of the updated feasible domain, a pre-defined recovery strategy, such as boundary absorption, is employed to force the value to the nearest valid boundary.

[0076] After completing the above-mentioned real-time constraint processing, the verified and repaired physically feasible candidate scheduling scheme is transmitted as an input vector to the input layer of the neural network proxy model. The neural network proxy model serves as the fitness function evaluation engine of the particle swarm algorithm, and transmits the candidate scheduling scheme generated by the particle swarm algorithm in each iteration, which represents a set of specific gate and pump operation instructions (that is, the position of the particle in the solution space represents a set of gate and pump operation instructions), as an input vector to the input layer of the neural network proxy model; the neural network proxy model performs a forward propagation calculation instead of calling the coupled hydrodynamic model to perform a complete physical simulation based on solving partial differential equations; and the output layer of the neural network proxy model outputs a predicted scheduling state vector that can characterize the multi-dimensional physical state of the highest water level of key sections in the region and the inundated area under the candidate scheduling scheme.

[0077] This embodiment uses a neural network proxy model as the fitness function evaluation engine of the particle swarm optimization algorithm, converting a physical simulation process with extremely high computational cost into a lightweight neural network forward propagation process. This solves the core technical bottleneck that complex hydrodynamic models cannot be directly used for real-time optimization scheduling due to the huge computational time (usually taking several hours). The system can complete the evaluation and optimization of tens of thousands of solutions within minutes, thereby improving the timeliness and practical usability of optimization decisions.

[0078] Furthermore, based on the predicted scheduling state and weight coefficient, the fitness value of each candidate scheduling solution is calculated, and the optimal scheduling solution is output through iterative optimization of the particle swarm algorithm, including:

[0079] After obtaining the predicted scheduling status, the system performs a weighted summation on the weight coefficients output by the multi-objective weight adaptation module and the multi-dimensional predicted scheduling state vector output by the neural network agent model to calculate the scalar fitness value of each candidate scheduling scheme. This scalar value reflects the pros and cons of the scheme's comprehensive objectives under the current scenario. Generally, the smaller the objective function value (or the larger, depending on the optimization direction), the better the fitness.

[0080] Each candidate scheduling solution is treated as a particle. Based on the solution with the optimal fitness value corresponding to the particle's own historical iteration (individual optimality) and the solution with the optimal fitness value corresponding to the entire particle swarm's historical iteration (global optimality), the particle's speed and position in the solution space are updated to generate a new candidate scheduling solution. In practice, the operating parameters of the particle swarm algorithm can be adjusted based on actual conditions. For example, the particle swarm size is typically set between 20 and 100 particles, and the inertia weight used to update particle speed is typically set between 0.4 and 0.9. In addition, the individual learning factor (cognitive coefficient) and the social learning factor (social coefficient) are typically set between 1.5 and 2.5, controlling the degree to which the particle approaches its own historical optimal position and the group's historical optimal position, respectively.

[0081] After the preset termination conditions are met, the system outputs the solution with the optimal fitness value corresponding to the current historical iteration of the entire particle population as the optimal scheduling solution. The termination conditions may include reaching the maximum number of iterations (for example, 1000 times), the fitness value converging to a certain threshold (for example, the improvement rate is less than 0.001 for 20 consecutive iterations), or the calculation time reaching the upper limit (for example, 5 minutes), to ensure that a high-quality solution is obtained under real-time requirements.

[0082] This embodiment calculates the scalar fitness value by performing weighted summation of dynamically changing weight coefficients and high-dimensional prediction effect vectors, thereby converting a multi-objective optimization problem that is difficult to directly handle mathematically and contains multiple conflicting objectives into a single-objective optimization problem with clear evaluation criteria. Through the particle swarm algorithm's inherent evolutionary mechanism that takes into account both individual exploration and global optimization, it ensures that the algorithm can search efficiently and robustly in a huge, nonlinear solution space, improving the technical bottleneck of traditional optimization algorithms that are prone to falling into local optimality and being unable to find the global optimal solution due to unclear search directions and vague evaluation criteria when facing high-dimensional, multi-objective conflicting problems, thereby significantly improving the quality and reliability of decision-making.

[0083] This embodiment constructs a hierarchical intelligent decision-making system that integrates four core modules: coupled hydrodynamic simulation, proxy model generation, multi-objective weight adaptation, and particle swarm optimization scheduling. The coupled hydrodynamic model module simulates physical processes with high fidelity; the proxy model generation module learns and generates a neural network proxy model that can rapidly replace physical simulations; the multi-objective weight adaptation module dynamically determines the optimal decision-making objective weights based on real-time warning levels; and the particle swarm optimization scheduling module, guided by topological rules, utilizes the proxy model for rapid evaluation and, guided by the dynamic weights, outputs the optimal scheduling solution through iterative optimization. By organically integrating and synergizing high-fidelity physical simulation, machine learning fast reasoning, and multi-objective intelligent optimization algorithms within a unified, end-to-end framework, the system achieves complementary advantages and synergistic efficiency across multiple advanced technologies, thereby resolving the fundamental issue of existing scheduling technologies, which, due to inherent flaws in models, algorithms, and decision logic, cannot guarantee scientific, real-time, global, and adaptive decision-making.

[0084] Example 2

[0085] This embodiment fully deploys the above-mentioned plain river network flood control and dispatching system in a plain river network embankment area to realize intelligent and refined management of regional flood control and drainage. The plain river network embankment area has low terrain, dense river networks, complex water systems, and huge pressure on flood control and drainage. Traditional dispatching methods mainly rely on manual experience and are difficult to cope with the risk of urban flooding under sudden heavy rainfall and complex tidal combinations. Therefore, there is an urgent need for a flood control and dispatching system that can respond quickly, accurately predict, and intelligently optimize.

[0086] When facing an impending rainstorm event, the system will receive and integrate multi-source data in real time, including real-time rainfall forecast data (e.g., hourly rainfall in the next 24 hours, in millimeters / hour), river network water level data (e.g., real-time water level at key sections, in meters), pipe network drainage data (e.g., flow at major drainage outlets, in cubic meters / second), and external river tide data (e.g., real-time tide level of external rivers, in meters).

[0087] The simulation results of the coupled hydrodynamic model module will serve as the basic data for the gate and pump scheduling optimization model, providing high-precision predictions of the physical world state. The model uses the opening time and opening degree (for example, the gate opening angle, in degrees or the opening height, in meters) of the sluice gates in the area, as well as the start-up time and operating power (for example, the number of operating pump stations or pump power, in kilowatts) of the water pumps in the area as decision variables that need to be optimized.

[0088] The multi-objective weight adaptive module identifies the current water situation based on real-time rainfall intensity and external river tide data, and dynamically adjusts the weight coefficients of flood control and drainage objectives. For example, when the high tide level of the external river and heavy rainfall in the internal river overlap, the flood control safety weight will be significantly increased, and the system will give priority to controlling the water level of the internal river to prevent the backflow of external river water. Specifically, the system comprehensively considers multiple important goals and assigns corresponding weights: flood control goal, minimize the risk of water levels in rivers and pipelines exceeding warning levels to ensure regional flood control safety; drainage goal, reduce the depth and duration of water accumulation in the region to avoid waterlogging disasters; water resource utilization goal, rationally store and utilize water resources to improve water resource utilization efficiency. The formulation of scheduling plans must strictly comply with various physical and operational constraints, including gate and pump physical constraints (the opening angle of the sluice gate, the operating power of the water pump, etc. must not exceed their design range and upper and lower limits), water balance constraints (the amount of water flowing in and out of the entire area must be balanced, and changes in the water storage capacity of the water storage facilities must be taken into account) and water level constraints (the water level of the river and pipeline must be controlled within a safe range, neither too high to cause flood risks nor too low to affect normal functions).

[0089] Based on the predicted scheduling state and dynamic weight coefficients, the system calculates the scalar fitness value of each physically feasible candidate scheduling solution through a weighted summation method. This fitness value comprehensively evaluates the solution's pros and cons in the current scenario. The algorithm then enters the particle update phase. Each candidate scheduling solution, as a particle, updates its speed and position in the solution space based on the solution with the optimal fitness value corresponding to its own historical iteration (individual optimality) and the solution with the optimal fitness value corresponding to the historical iteration of the entire particle population (global optimality), thereby generating a new generation of more optimal candidate scheduling solutions. Moreover, in each iteration, it can dynamically adjust the feasible domain constraints of its search based on real-time external working information (such as pump failures) and inspect and repair newly generated particles (candidate solutions).

[0090] Finally, after meeting the preset termination criteria (e.g., reaching the maximum number of iterations or fitness convergence), the system initiates a final physical verification phase, submitting key candidate solutions to a high-fidelity physical model for verification. This results in a reliable, efficiently found, and accurately verified optimal scheduling solution that minimizes inundation and maintains water levels at key sections within safe limits. Throughout the optimization process, additional constraints, such as pump station operating energy consumption and sluice gate operation frequency, can be incorporated into the fitness function or used as screening criteria for feasible solutions, ensuring the comprehensiveness and feasibility of the final solution.

[0091] After the optimal scheduling plan is generated, the system will push detailed scheduling instructions to the flood control command center in real time. The command center personnel can make decisions based on the system's suggestions and the actual situation on site, and execute the scheduling instructions through the remote control system. At the same time, the system provides a visual interface to display real-time water conditions, predicted inundation maps, and comparative effects of different scheduling plans to assist decision makers in making judgments.

[0092] It should be emphasized that the specific embodiments described herein are merely examples illustrating the core concepts of the present invention and are not intended to limit its scope. Those skilled in the art will appreciate that any modification, combination, or equivalent substitution of the above embodiments, without departing from the core principles and spirit of the present invention, falls within the scope of protection claimed by the present invention. The ultimate scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A multi-objective particle swarm optimization scheduling system for flood control and drainage in plain river network areas, characterized by: include: Coupled hydrodynamic model module, which calculates the water flow exchange between the one-dimensional hydrodynamic model used to simulate underground pipe network water flow and the two-dimensional hydrodynamic model used to simulate surface flow; An agent model generation module, calling the coupled hydrodynamic model module to perform multiple simulations to generate a training data set, training the training data set and generating a neural network agent model; The multi-objective weight adaptive module dynamically determines a set of weight coefficients representing the relative importance of multiple scheduling objectives based on the received real-time warning level; The particle swarm optimization scheduling module generates initial candidate scheduling schemes as the initial population of the particle swarm based on the topological structure information of the plain river network and preset heuristic rules. During the iterative optimization process, the feasible domain constraints of the optimization algorithm are dynamically adjusted according to the external working situation information received in real time. The candidate scheduling schemes are input into the neural network agent model and the predicted scheduling status is output. Based on the predicted scheduling status and weight coefficient, the fitness value of each candidate scheduling scheme is calculated, and the optimal scheduling scheme is output through the iterative optimization of the particle swarm algorithm.

2. A multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in plain river network polder areas according to claim 1, characterized in that: The coupled hydrodynamic model module calculates the water flow exchange between a one-dimensional hydrodynamic model for simulating underground pipe network water flow and a two-dimensional hydrodynamic model for simulating surface overland flow, including: the one-dimensional hydrodynamic model is a storm flood management model solved based on the Saint-Venant equations; the two-dimensional hydrodynamic model is a flood overland flow model solved based on the two-dimensional unsteady shallow water equations; the calculation method of the water flow exchange includes: calculation of the vertical exchange volume, using the orifice flow formula and the weir flow formula to calculate the overflow caused by the head difference between the underground pipe network node and the two-dimensional surface grid; calculation of the lateral exchange volume, using the river embankment of the one-dimensional hydrodynamic model as the internal boundary of the two-dimensional hydrodynamic model, and calculating the lateral overland flow generated when the river water level exceeds the embankment elevation through the weir flow formula.

3. The multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in plain river network polder areas according to claim 1 is characterized in that: The proxy model generation module includes: calling a coupled hydrodynamic model module to simulate operating conditions covering different rainfall scenarios and scheduling scheme combinations to generate a training data set; the neural network proxy model includes an input layer, a hidden layer, and an output layer; the input layer receives a scheduling scheme vector composed of the opening degrees of each sluice gate and the start and stop status of the pump station; the output layer outputs a multidimensional predicted scheduling state vector composed of the highest water level of key sections in the area and the inundated area.

4. The multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in plain river network polder areas according to claim 1 is characterized in that: The multi-objective weight adaptive module includes: a scenario-strategy mapping rule base stored in the multi-objective weight adaptive module; wherein each scenario is represented by a warning level defined by real-time rainfall intensity and external river tide data; each strategy corresponds to a set of weight coefficient combinations preset for the two goals of flood control and drainage; real-time external incoming real-time rainfall intensity and real-time external river tide data representing the current warning level are received in real time; the received real-time data is matched one by one with multiple preset scenarios in the rule base to determine the interval defined by multiple closest preset scenarios into which the current real-time data falls; when the real-time data accurately matches a preset scenario, the module directly outputs the preset strategy weight corresponding to the scenario; when the real-time data falls between two preset scenarios, the module generates and outputs a set of dynamic weight coefficients applicable to the current scenario by performing linear interpolation calculation on the preset strategy weights corresponding to the two preset scenarios.

5. The multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in plain river network embankment areas according to claim 1 is characterized in that: The particle swarm optimization scheduling module generates initial candidate scheduling schemes as the initial population of the particle swarm based on the topological structure information of the plain river network and preset heuristic rules, including: obtaining topological structure information of the plain river network represented by nodes and edges, which includes the upstream and downstream relationships of various water conservancy facilities; applying preset heuristic rules to traverse paths and prioritize the topological structure; the heuristic rules include spatial priority rules, which determine the spatial execution order of scheduling operations based on the upstream and downstream relationships in the topological structure; the heuristic rules also include temporal priority rules, which determine the temporal sequence of scheduling operations based on the division of main streams and tributaries in the topological structure; based on the determined spatial execution order and temporal sequence, a group of candidate scheduling schemes with random perturbations in specific operation parameters and that generally comply with the heuristic rules are generated as the initial population.

6. The multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in plain river network embankment areas according to claim 1 is characterized in that: During the iterative optimization process, the feasible domain constraints of the optimization algorithm are dynamically adjusted according to the external working condition information received in real time, the candidate scheduling schemes are input into the neural network proxy model, and the predicted scheduling state is output, including: receiving external working condition information including real-time fault alarm information of water conservancy project facilities, dynamically adjusting the value range of the corresponding decision variables in the optimization algorithm as the feasible domain constraints; checking the candidate scheduling schemes newly generated by the particle swarm algorithm, and performing boundary repair processing on the schemes that exceed the feasible domain constraints; using the neural network proxy model as the fitness function evaluation engine of the particle swarm algorithm, and transmitting the candidate scheduling schemes generated by the particle swarm algorithm in each iteration as input vectors to the input layer of the neural network proxy model; the neural network proxy model performs a forward propagation calculation instead of calling the coupled hydrodynamic model to perform a complete and time-consuming physical simulation; and the output layer of the neural network proxy model outputs a predicted scheduling state vector that can characterize the multi-dimensional physical state of the highest water level of key sections in the region and the inundated area under the candidate scheduling scheme.

7. The multi-objective particle swarm optimization scheduling system for flood control and waterlogging removal in plain river network polder areas according to claim 1 is characterized in that: Based on the predicted scheduling state and weight coefficient, the fitness value of each candidate scheduling scheme is calculated, and the optimal scheduling scheme is output through iterative optimization of the particle swarm algorithm, including: performing weighted summation on the weight coefficient and the predicted scheduling state vector to calculate the scalar fitness value of each candidate scheduling scheme; each candidate scheduling scheme is regarded as a particle, and the speed and position of the particle in the solution space are updated according to the scheme with the optimal fitness value corresponding to the particle's own historical iteration and the scheme with the optimal fitness value corresponding to the historical iteration of the entire particle population, thereby generating a new candidate scheduling scheme; after a preset termination condition is met, the scheme with the optimal fitness value corresponding to the current entire particle population historical iteration is output as the optimal scheduling scheme.

Images