Campus lid facility rainfall runoff regulation system and method based on swmm

Abstract

This invention belongs to the field of urban stormwater management technology and discloses a campus LID facility rainfall runoff control system and method based on SWMM. The method includes: constructing a refined SWMM model to delineate catchment areas and pipe networks; deploying LID and grey facilities such as permeable paving, sunken green spaces, and storage ponds; integrating multi-objective genetic algorithms for optimized design of facility parameters and layout; integrating model predictive control algorithms to achieve real-time collaborative control among facilities; and employing deep reinforcement learning algorithms to enhance the system's adaptability and resilience under extreme rainfall. This algorithm-driven approach achieves intelligent management throughout the entire process from design optimization to operational control, significantly improving the efficiency of source reduction and process control of rainfall runoff, reducing pipe network renovation costs, and enhancing campus flood safety.

Description

Technical Field

[0001] This invention relates to the field of urban stormwater management technology, and in particular to a campus LID facility rainfall-runoff regulation system and method based on SWMM. Background Technology

[0002] With the acceleration of urbanization, the increase in impermeable surface area has led to a surge in rainwater runoff, exacerbating urban flooding problems. Traditional drainage systems rely on expanding pipe networks, but this is costly and inadequate for handling extreme rainfall. Low Impact Development (LID) facilities control runoff at its source through infiltration, retention, and storage, but their effectiveness requires quantitative analysis based on specific scenarios. Current technologies for LID facilities largely rely on empirical parameters, lacking refined simulations based on dynamic models, and fail to adequately consider synergistic effects with grey infrastructure (such as stormwater storage tanks), resulting in insufficient system resilience under extreme rainfall.

[0003] For example, existing campus stormwater management systems often suffer from an excessively high proportion of impermeable area (e.g., 79.6% without LID systems), leading to a significant increase in runoff coefficient with rainfall intensity, resulting in frequent pipe network overload and waterlogging. Furthermore, while LID systems are effective at controlling light rainfall, their peak-shaving capacity for heavy downpours is limited; for instance, the peak reduction rate is only 25.35% at P=20a, and they lack an adaptive optimization mechanism.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a rainfall-runoff regulation system and method for campus LID facilities based on SWMM, aiming to solve the technical problems of high cost of traditional drainage systems and the lack of refined simulation based on dynamic models and adaptive optimization mechanisms in the design of current LID facilities.

[0006] To achieve the above objectives, the present invention provides a campus LID facility rainfall-runoff control system based on SWMM. The campus LID facility rainfall-runoff control system based on SWMM is characterized in that it includes: a data acquisition module, a model building module, an algorithm integration module, an LID facility control module, and a feedback calibration module. The data acquisition module is used to collect basic data and real-time monitoring data of the campus area, and output the basic data and real-time monitoring data into a standardized dataset. The basic data includes topographic elevation, land use, pipeline topology, rainfall meteorology, and soil permeability coefficient. The real-time monitoring data includes rainfall, water level, and flow rate. The model building module is used to build an SWMM model based on the standardized dataset, divide the study area into several independent sub-catchments and several pipe network nodes according to the confluence path, and output the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipe network and the initial simulation data of the model. The sub-catchments are used to match the deployment range of LID facilities, and the pipe network nodes are used to monitor the hydraulic status. The algorithm integration module uses the SWMM model as the simulation core and integrates a multi-objective genetic algorithm, a model prediction control algorithm, and a deep reinforcement learning algorithm according to a preset three-level logic to output corresponding control commands. The LID facility control module is used to respond to the control commands output by the algorithm integration module and dynamically adjust the operating parameters of the intelligent LID facility. The spatial layout of the intelligent LID facility corresponds one-to-one with the sub-catchment areas output by the model construction module, and its initial capacity parameters are determined based on the impermeability and area in the catchment area runoff generation parameters. The feedback calibration module is used to collect the actual runoff data after it has been adjusted by the LID facility control module, compare the actual runoff data with the runoff data under the corresponding operating conditions output by the SWMM model simulation, and dynamically calibrate the runoff generation and collection parameters of the catchment area and the hydraulic characteristic parameters of the pipeline network when the error exceeds a preset threshold. The calibrated parameters are then sent back to the algorithm integration module for model and algorithm optimization to form a closed loop.

[0007] In one embodiment, the multi-objective genetic algorithm receives the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipeline network, and the initial simulation data of the model output by the model building module. Using the area and impermeability of each sub-catchment area in the runoff generation and confluence parameters, it determines the upper limit of the LID facility deployment area and the minimum proportion constraint of permeable paving. Using the design water depth of each node in the pipeline network hydraulic characteristic parameters, it determines the pipeline network carrying capacity threshold constraint. Furthermore, using the total runoff and peak flow rate in the initial simulation data of the model as optimization benchmarks, it solves for the target static configuration scheme of the LID facilities and outputs the Pareto target facility configuration scheme as initial control parameters. The Pareto target facility configuration scheme includes the LID facility type, deployment location, and initial parameters. The model predictive control algorithm receives real-time monitoring data from the data acquisition module, rolling simulation data from the SWMM model, and initial control parameters output by the multi-objective genetic algorithm. It outputs dynamic control commands to the LID facility regulation module and simultaneously feeds back regulation data to the training environment of the deep reinforcement learning algorithm. The deep reinforcement learning algorithm receives extreme rainfall simulation data output by the model building module, regulation data from the model prediction control algorithm, and error data output by the subsequent feedback calibration module, and trains to generate adaptive enhanced control commands under extreme conditions.

[0008] In one embodiment, the multi-objective genetic algorithm, the model prediction control algorithm, and the deep reinforcement learning algorithm are called sequentially in runtime order, and the data interaction relationships at different stages include: When in the offline pre-optimization stage, the multi-objective genetic algorithm is invoked, with a single optimization cycle of 24-72 hours. The output of the multi-objective genetic algorithm is used as the initial control parameters of the model prediction control algorithm. When in the online operation phase, the model prediction control algorithm is called cyclically, with a control cycle of 5-15 minutes per cycle. The output of the model prediction control algorithm is a dynamic control command, and the running data is synchronized to the training set of the deep reinforcement learning algorithm. When in an extreme response phase, the deep reinforcement learning algorithm is triggered, wherein the adaptive enhanced control command output by the deep reinforcement learning algorithm is used to replace or supplement the dynamic control command output by the model predictive control algorithm.

[0009] In one embodiment, the multi-objective genetic algorithm uses four objective functions—minimizing peak runoff, minimizing total runoff, minimizing facility construction cost, and maximizing system resilience index—to solve for the target static configuration scheme of LID facilities under constraints, including: The coverage rate of LID facilities in the sub-catchment area is ≤80% of the sub-catchment area; The total area of ​​variable porosity permeable pavement is ≥ 30% of the hardened area of ​​the sub-catchment area; Total volume of the regulating reservoir ≤ catchment area of ​​the sub-catchment area × 50mm; Construction cost ≤ preset budget threshold.

[0010] In one embodiment, the model predictive control algorithm uses the SWMM model as the internal predictive model, and the corresponding control logic is that the prediction time domain is set to 1-6 hours, and the rolling optimization cycle is 5-15 minutes; the input data includes real-time monitoring data from the data acquisition module and simulation data of the SWMM model at the current moment; the maximum water level of the pipeline node is controlled at 70%-80% of the design water level, and the total runoff reduction rate is ≥60%; the output control commands include variable pore opening, height of the sunken green space water storage layer, opening of the overflow valve of the regulating reservoir, and speed of the drainage pump.

[0011] In one embodiment, the intelligent LID facility includes a variable porosity permeable pavement, an adjustable water storage layer sunken green space, and a controllable overflow storage tank. The variable porosity permeable pavement adopts a double-layer structure, with the upper layer being porous concrete with a porosity of 15%-25% and the lower layer being a variable porosity adjustment layer. The dynamic adjustment of the operating parameters is achieved by controlling the pore opening from 0-100% through an electric push rod to achieve a dynamic adjustment of the infiltration rate from 0.5-5m / h. The adjustable water storage layer sunken green space is equipped with a liftable water-resistant plate. The dynamic adjustment of the operating parameters is achieved by controlling the thickness of the water storage layer from 50 to 300 mm through a stepper motor, and by using a soil moisture sensor for feedback adjustment, with the error being less than the preset error. The controllable overflow storage tank is equipped with an electric overflow valve and a variable frequency drainage pump. The dynamic adjustment of the operating parameters is achieved by the PLC controller receiving instructions from the algorithm integration module, responding and adjusting the overflow threshold (0.5-2.0m) and pump speed (0-100m) within a preset time. 3 / h.

[0012] In one embodiment, the feedback calibration module is used to collect actual rainfall, water level, and flow data after LID facility regulation, and to calculate the actual total runoff Qactual and peak flow Qp. The SWMM model was used to simulate the total runoff Q and peak flow Qp under the same operating conditions. The least squares method is used to calculate the error values ​​E1=|Qactual-Qsimulated| / Qactual and E2=|Qpactual-Qpsimulated| / Qpactual. When E1>10% or E2>15%, calibration is initiated. The calibrated model parameters are synchronously updated to the constraints of the multi-objective genetic algorithm to optimize the next generation of Pareto solutions; The calibrated error data is used as a prediction bias compensation term for the model predictive control algorithm to correct the control command; The calibrated extreme condition data are added to the training sample library of the deep reinforcement learning algorithm to improve the model's generalization ability.

[0013] Furthermore, to achieve the above objectives, a campus LID facility rainfall-runoff regulation method based on SWMM is provided. This method applies the SWMM-based campus LID facility rainfall-runoff regulation system described above, and includes the following steps: The data acquisition module collects basic data and real-time monitoring data of the campus area, and outputs the basic data and real-time monitoring data into a standardized dataset. The basic data includes topographic elevation, land use, pipeline topology, rainfall meteorology, and soil permeability coefficient, and the real-time monitoring data includes rainfall, water level, and flow rate. The model building module constructs an SWMM model based on the standardized dataset, divides the study area into several independent sub-catchments and several pipe network nodes according to the confluence path, and outputs the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipe network, and the initial simulation data of the model. The sub-catchments are used to match the deployment range of LID facilities, and the pipe network nodes are used to monitor the hydraulic status. The algorithm integration module uses the SWMM model as the simulation core and integrates multi-objective genetic algorithm, model prediction control algorithm and deep reinforcement learning algorithm according to the preset three-level logic to output corresponding control commands; The LID facility control module responds to the control commands output by the algorithm integration module and dynamically adjusts the operating parameters of the intelligent LID facility. The spatial layout of the intelligent LID facility corresponds one-to-one with the sub-catchment areas output by the model construction module, and its initial capacity parameters are determined based on the impermeability and area in the catchment area's runoff generation parameters. The feedback calibration module collects the actual runoff data after it has been adjusted by the LID facility control module. It compares the actual runoff data with the runoff data under the corresponding operating conditions output by the SWMM model simulation. When the error exceeds the preset threshold, it dynamically calibrates the runoff generation and collection parameters of the catchment area and the hydraulic characteristic parameters of the pipeline network. The calibrated parameters are then sent back to the algorithm integration module for model and algorithm optimization to form a closed loop. In one embodiment, the multi-objective genetic algorithm receives the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipeline network, and the initial simulation data of the model output by the model building module. Using the area and impermeability of each sub-catchment area in the runoff generation and confluence parameters, it determines the upper limit of the LID facility deployment area and the minimum proportion constraint of permeable paving. Using the design water depth of each node in the pipeline network hydraulic characteristic parameters, it determines the pipeline network carrying capacity threshold constraint. Furthermore, using the total runoff and peak flow rate in the initial simulation data of the model as optimization benchmarks, it solves for the target static configuration scheme of the LID facilities and outputs the Pareto target facility configuration scheme as initial control parameters. The Pareto target facility configuration scheme includes the LID facility type, deployment location, and initial parameters. The model predictive control algorithm receives real-time monitoring data from the data acquisition module, rolling simulation data from the SWMM model, and initial control parameters output by the multi-objective genetic algorithm. It outputs dynamic control commands to the LID facility regulation module and simultaneously feeds back regulation data to the training environment of the deep reinforcement learning algorithm. The deep reinforcement learning algorithm receives extreme rainfall simulation data output by the model building module, regulation data from the model prediction control algorithm, and error data output by the subsequent feedback calibration module, and trains to generate adaptive enhanced control commands under extreme conditions.

[0014] In one embodiment, the multi-objective genetic algorithm, the model prediction control algorithm, and the deep reinforcement learning algorithm are called sequentially in runtime order, and the data interaction relationships at different stages include: When in the offline pre-optimization stage, the multi-objective genetic algorithm is invoked, with a single optimization cycle of 24-72 hours. The output of the multi-objective genetic algorithm is used as the initial control parameters of the model prediction control algorithm. When in the online operation phase, the model prediction control algorithm is called cyclically, with a control cycle of 5-15 minutes per cycle. The output of the model prediction control algorithm is a dynamic control command, and the running data is synchronized to the deep reinforcement learning algorithm training set. When in an extreme response phase, the deep reinforcement learning algorithm is triggered, wherein the adaptive enhanced control command output by the deep reinforcement learning algorithm is used to replace or supplement the dynamic control command output by the model predictive control algorithm.

[0015] This invention constructs a refined SWMM model to delineate catchment areas and pipe networks; lays out permeable paving, sunken green spaces, and water storage ponds, among other LID (Limited Area Development) and grey area facilities; integrates a multi-objective genetic algorithm for optimizing facility parameters and layout; integrates a model predictive control algorithm to achieve real-time coordinated regulation among facilities; and employs a deep reinforcement learning algorithm to enhance the system's adaptability and resilience under extreme rainfall. This algorithm-driven approach achieves intelligent management throughout the entire process, from design optimization to operational control, significantly improving the efficiency of source reduction and process control of rainfall runoff, reducing pipe network renovation costs, and enhancing campus flood safety. Attached Figure Description

[0016] Figure 1 This is a structural block diagram of the first embodiment of the campus LID facility rainfall-runoff regulation system based on SWMM of the present invention; Figure 2 This is a flowchart illustrating the first embodiment of the campus LID facility rainfall-runoff regulation method based on SWMM of the present invention.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0019] This invention provides a campus LID facility rainfall-runoff regulation system based on SWMM, referring to... Figure 1 , Figure 1 This is a structural block diagram of a first embodiment of a campus LID facility rainfall-runoff regulation system based on SWMM according to the present invention.

[0020] It should be noted that with the acceleration of urbanization, the increase in impervious surface area has led to a surge in rainwater runoff, exacerbating urban flooding problems. Traditional drainage systems rely on expanding pipe networks, but this is costly and inadequate for handling extreme rainfall. Low Impact Development (LID) facilities control runoff at its source through infiltration, retention, and storage, but their effectiveness requires quantitative analysis based on specific scenarios. Current technologies often rely on empirical parameters for LID facility design, lacking refined simulations based on dynamic models and failing to adequately consider synergistic effects with grey infrastructure (such as stormwater storage tanks), resulting in insufficient system resilience under extreme rainfall. For example, existing campus stormwater management systems often suffer from excessively high impervious surface areas (e.g., 79.6% without LID), leading to a significant increase in runoff coefficient with rainfall intensity, resulting in frequent pipe network overload and waterlogging. Furthermore, while LID facilities are effective in controlling small rainfall events, their peak reduction capacity for heavy rainstorms is limited; for example, the peak reduction rate is only 25.35% at P=20 years, and they lack adaptive optimization mechanisms.

[0021] To address the aforementioned technical challenges, this embodiment constructs a refined SWMM model to delineate catchment areas and pipe networks; lays out permeable paving, sunken green spaces, and water storage ponds, among other LID (Limited Area Development) and grey area facilities; integrates a multi-objective genetic algorithm for optimizing facility parameters and layout; integrates a model predictive control algorithm to achieve real-time collaborative regulation among facilities; and employs a deep reinforcement learning algorithm to enhance the system's adaptability and resilience under extreme rainfall. This algorithm-driven approach achieves intelligent management throughout the entire process, from design optimization to operational control, significantly improving the efficiency of source reduction and process control of rainfall runoff, reducing pipe network renovation costs, and enhancing campus flood safety. Specifically, it can be implemented as follows.

[0022] In this embodiment, the campus LID facility rainfall-runoff control system based on SWMM includes a data acquisition module 10, a model building module 20, an algorithm integration module 30, an LID facility control module 40, and a feedback calibration module 50. The system construction includes model building: dividing the campus area into 61 sub-catchments and 62 pipe network nodes, constructing a refined SWMM model, and using dynamic wave simulation to model the hydraulic process. Low Impact Development (LID) facility deployment: based on land use and vertical design, permeable pavement, sunken green spaces, vegetated swales, permeable asphalt pavement, grass-covered permeable pavement, and rainwater storage tanks (at least three combinations) are deployed. Rainfall simulation: using the Chicago rainfall pattern to generate rainfall data with different return periods (0.802a-20a), the total simulation duration is 240 minutes, with 1-minute time intervals. Monitoring sensor network: deploying ultrasonic water level sensors, flow meters, and video monitoring devices to collect real-time water level, flow rate, and water accumulation image data at key nodes. Then, parameter optimization was performed, including LID facility parameters: permeable pavement surface layer thickness 30mm, porosity 0.1; sunken green space soil layer thickness 304.8mm, hydraulic conductivity 12.7mm / h. Infiltration model: Horton model was adopted, with a maximum infiltration rate of 78.1mm / h, a minimum infiltration rate of 4.83mm / h, and an attenuation coefficient of 2h. -1 Coordinated configuration: By combining the volume of the regulating tank with the drainage capacity of the submersible sewage pump, the storage and discharge ratio of LID facilities and gray facilities is dynamically adjusted.

[0023] In its implementation, the data acquisition module 10 is used to collect basic data and real-time monitoring data of the campus area, and output the basic data and real-time monitoring data as a standardized dataset. The basic data includes topographic elevation, land use, pipeline topology, rainfall meteorology, and soil permeability coefficient, while the real-time monitoring data includes rainfall, water level, and flow rate.

[0024] Furthermore, the model building module 20 is used to build an SWMM model based on the standardized dataset, divide the study area into several independent sub-catchments and several pipe network nodes according to the confluence path, and output the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipe network, and the initial simulation data of the model. For example, the campus area is divided into 61 sub-catchments and 62 pipe network nodes to build a refined SWMM model. The sub-catchments are used to match the deployment range of LID facilities, and the pipe network nodes are used to monitor the hydraulic status.

[0025] Furthermore, the algorithm integration module 30, using the SWMM model as the simulation core, integrates a multi-objective genetic algorithm, a model prediction control algorithm, and a deep reinforcement learning algorithm according to a preset three-level logic to output corresponding control commands.

[0026] It should be noted that this embodiment integrates three core algorithms to achieve system design optimization and real-time collaborative control: Multi-objective genetic optimization algorithm: used for initial optimization of LID facility parameters and layout. With the objectives of minimizing peak runoff, total runoff, construction cost, and maximizing system resilience, and using permeable pavement area ratio, sunken green space water storage layer thickness, and storage tank volume as decision variables, the NSGA-II algorithm is used for multi-objective optimization, outputting a Pareto optimal solution set to provide a scientific basis for facility design. Model predictive control collaborative control algorithm: used for real-time collaborative operation of LID facilities and gray facilities. Using SWMM as the internal prediction model, combined with real-time rainfall forecasts and system status, it continuously solves for the optimal sequence of control commands such as storage tank drainage pump start / stop and overflow outlet opening in future periods, achieving minimization of pipeline overload risk and balanced storage and discharge. State prediction model: using SWMM as the internal prediction model, it takes real-time rainfall sequences and facility status as input; outputs water depth at each node, pipeline flow, and storage tank water level in future periods. Optimized control objectives: Minimize the risk of pipeline overload; balance the storage and discharge strategies of the stormwater storage tank to avoid overflow; ensure no water accumulation in key areas. Control variables: Start-up and shutdown sequence of stormwater storage tank drainage pumps; overflow outlet opening of vegetated swales and sunken green spaces. Rolling optimization mechanism: Update rainfall forecasts and system status every 5 minutes; solve for the optimal control sequence within the next 30 minutes; execute the initial control command and iterate continuously. Deep reinforcement learning: Used to improve the system's adaptability under extreme rainfall. Construct a reinforcement learning environment with rainfall intensity, facility status, and pipeline load as the state space, and dynamic adjustment of facility parameters and multi-level pump speed control as the action space. Use DDPG or PPO algorithms to learn the optimal control strategy to achieve continuous improvement in system resilience. State space: Real-time rainfall intensity, duration, and previous soil moisture; water storage status of each LID facility, stormwater storage tank water level, and pipeline load rate. Action space: Adjust the infiltration rate of permeable pavement (through variable pore structure design); dynamically adjust the release rate of the water storage layer in sunken green spaces; multi-level speed control of stormwater storage tank drainage pumps. Reward Function: Positive rewards: reducing peak runoff, minimizing waterlogging points, and extending the safe operating time of the pipeline network; Negative rewards: facility overflow, excessive energy consumption, and delayed control response. Training and Deployment: Offline training is performed using historical rainfall sequences coupled with an SWMM environment; the optimal control strategy is learned using DDPG or PPO algorithms; during online deployment, the strategy is fine-tuned using real-time sensor data.

[0027] Specifically, the multi-objective genetic algorithm, the model predictive control algorithm, and the deep reinforcement learning algorithm are called sequentially according to the runtime order. The data interaction relationship at different stages includes: when in the offline pre-optimization stage, the multi-objective genetic algorithm is called, with a single optimization cycle of 24-72 hours, wherein the output of the multi-objective genetic algorithm is used as the initial control parameters of the model predictive control algorithm; when in the online running stage, the model predictive control algorithm is called cyclically, with a control cycle of 5-15 minutes / time, wherein the output of the model predictive control algorithm is a dynamic control instruction, and the running data is synchronized to the training set of the deep reinforcement learning algorithm; when in the extreme response stage, the deep reinforcement learning algorithm is triggered, wherein the adaptive enhanced control instruction output by the deep reinforcement learning algorithm is used to replace or supplement the dynamic control instruction output by the model predictive control algorithm.

[0028] Furthermore, the multi-objective genetic algorithm uses four objective functions—minimizing peak runoff, minimizing total runoff, minimizing facility construction cost, and maximizing system resilience index—to solve for the target static configuration scheme of LID facilities under constraints. These constraints include: LID facility coverage in sub-catchment areas ≤ 80% of the sub-catchment area; total area of ​​variable porosity permeable paving ≥ 30% of the hardened area of ​​the sub-catchment; total volume of regulating reservoirs ≤ sub-catchment catchment area × 50 mm; and construction cost ≤ a preset budget threshold. The specific algorithm flow is as follows: initialize the population and randomly generate facility parameter combinations; call the SWMM model for hydrological and hydraulic simulation; calculate the values ​​of each objective function; use NSGA-II for non-dominated sorting and crowding calculation; perform selection, crossover, and mutation operations to generate a new generation of population; iterate until convergence or the maximum number of generations is reached, and output the Pareto optimal solution set.

[0029] In one embodiment, the model predictive control algorithm uses the SWMM model as the internal predictive model, and the corresponding control logic is that the prediction time domain is set to 1-6 hours, and the rolling optimization cycle is 5-15 minutes; the input data includes real-time monitoring data from the data acquisition module and simulation data of the SWMM model at the current moment; the maximum water level of the pipeline node is controlled at 70%-80% of the design water level, and the total runoff reduction rate is ≥60%; the output control commands include variable pore opening, height of the sunken green space water storage layer, opening of the overflow valve of the regulating tank, and speed of the drainage pump.

[0030] Furthermore, the LID facility control module 40 is used to respond to the control commands output by the algorithm integration module and dynamically adjust the operating parameters of the intelligent LID facility. The spatial layout of the intelligent LID facility corresponds one-to-one with the sub-catchment areas output by the model construction module, and its initial capacity parameters are determined based on the impermeability and area in the catchment area runoff generation parameters.

[0031] In its specific implementation, the intelligent LID facility includes variable porosity permeable paving, adjustable water storage layer sunken green space, and controllable overflow storage tank. The variable porosity permeable paving adopts a double-layer structure: the upper layer is porous concrete with a porosity of 15%-25%, and the lower layer is a variable porosity adjustment layer. Dynamic adjustment of operating parameters is achieved by controlling the pore opening from 0-100% via an electric actuator, thus dynamically adjusting the infiltration rate from 0.5-5 m / h. The adjustable water storage layer sunken green space is equipped with a liftable water-resistant plate. Dynamic adjustment of operating parameters is achieved by controlling the water storage layer thickness from 50-300 mm via a stepper motor, with feedback adjustment via a soil moisture sensor, and the error is less than a preset error. The controllable overflow storage tank is equipped with an electric overflow valve and a variable frequency drainage pump. Dynamic adjustment of operating parameters is achieved by receiving instructions from the algorithm integration module via a PLC controller, responding within a preset time to adjust the overflow threshold from 0.5-2.0 m and the pump speed from 0-100 m / s. 3 / h.

[0032] Furthermore, the feedback calibration module 50 is used to collect the actual runoff data after it has been adjusted by the LID facility control module, compare the actual runoff data with the runoff data under the corresponding operating conditions output by the SWMM model simulation, and dynamically calibrate the runoff generation and collection parameters of the catchment area and the hydraulic characteristic parameters of the pipeline network when the error exceeds a preset threshold. The calibrated parameters are then sent back to the algorithm integration module for model and algorithm optimization to form a closed loop.

[0033] In one embodiment, actual rainfall, water level, and flow data after LID facility regulation are collected to calculate the actual total runoff Qactual and peak flow Qp; the SWMM model is called to simulate the total runoff Qsimulated and peak flow Qp under the same operating conditions; the least squares method is used to calculate the error values ​​E1 = |Qactual - Qsimulated| / Qactual and E2 = |Qp - Qp| / Qp. When E1 > 10% or E2 > 15%, calibration is initiated; the calibrated model parameters are synchronously updated to the constraints of the multi-objective genetic algorithm to optimize the next generation Pareto solution; the calibrated error data is used as the prediction bias compensation term of the model predictive control algorithm to correct the control command; the calibrated extreme operating condition data is added to the training sample library of the deep reinforcement learning algorithm to improve the model's generalization ability.

[0034] Furthermore, this embodiment also involves an IoT-based waterlogging identification and early warning algorithm. It employs a YOLOv5 deep learning model to identify waterlogged areas in monitoring videos; combines ultrasonic water level sensor data for multi-source fusion to improve identification accuracy; and uses a spatiotemporal sequence prediction model to provide early warning of waterlogging risks 10-30 minutes in advance. An adaptive parameter calibration algorithm uses a Bayesian optimization algorithm to dynamically calibrate key SWMM parameters (such as Manning coefficient and infiltration parameters); it updates model parameters using real-time monitoring data (flow rate, water level) to improve simulation accuracy; and it automatically performs parameter calibration after each rainfall event, forming a continuously learning model system.

[0035] After runoff control using the above methods and algorithm optimization, the LID facility achieved a peak runoff reduction rate of 82.34% for light rainfall (P=0.802a) and a peak reduction capacity of 32.7% for extreme rainfall (P=20a). System resilience: Through algorithmic coordinated regulation, the storage tank receives excess rainwater, reducing the overload duration of the pipe network by 50%-70%, reducing the number of waterlogging points by more than 60%, and shortening the system recovery time by approximately 25%.

[0036] Furthermore, this embodiment provides an example: (1) Study area: A newly built campus in Wuhan City, with an area of ​​9.91 hectares, divided into 61 sub-catchments, with impermeable area accounting for ≤80%. (2) LID facility layout: Permeable paving (22623.74m²) 2 Surface layer 30mm (porosity 0.1), water storage layer 50mm (porosity 0.45). Sunken green space (4498.035m) 2 Soil layer: 304.8 mm (hydraulic conductivity 12.7 mm / h); water storage layer: 150 mm. Regulation tank: Volume designed according to the lake's water storage capacity; equipped with submersible sewage pumps (drainage capacity ≥ 100 m³ / h). 3 / h), and incorporates the MPC algorithm for real-time scheduling. (3) Algorithm implementation process: Design phase: Run the multi-objective genetic algorithm, input land use constraints and cost parameters, and obtain the Pareto optimal facility configuration scheme. Operation phase: a. The MPC algorithm receives real-time sensor data and short-term rainfall forecasts every 5 minutes, solves the problem on a rolling basis, and executes the optimal control command. b. The DRL algorithm continuously learns in the background and updates the control strategy model after each rainfall. c. The Bayesian optimization algorithm automatically calibrates the Manning coefficient, infiltration parameters, etc. of SWMM based on monitoring data to maintain the model prediction accuracy. Early warning phase: The YOLOv5 water accumulation identification model is combined with the LSTM prediction model to provide water accumulation risk warnings 10-30 minutes in advance. (4) Simulation parameters: Chicago rain type peak coefficient r=0.4-0.5, rainfall duration 120 minutes. Manning coefficient of permeable surface 0.25, Manning coefficient of impermeable surface 0.015. (5) Simulation results: When P=0.802a, the runoff coefficient decreased from 0.751 to 0.15, and the peak flow was reduced by 82.34%. When P=20a, the number of overloaded pipe network sections decreased from 55 to 20, and the water storage capacity of the regulating reservoir reached 35% of the total runoff. (6) System optimization and expansion: The proportion of permeable pavement was dynamically adjusted to 30%, and the thickness of the water storage layer in the sunken green space was adaptively adjusted according to the optimization results. The central control platform integrates all algorithm modules to realize closed-loop management of "monitoring-simulation-optimization-control". The system supports access to the meteorological early warning platform through API to realize cross-system linkage control.

[0037] In this embodiment, a refined SWMM model is constructed to delineate the catchment area and pipe network; permeable paving, sunken green spaces, and water storage ponds are laid out as LID (Limited Area Development) and gray infrastructure; a multi-objective genetic algorithm is integrated to optimize the design of facility parameters and layout; a model predictive control algorithm is integrated to achieve real-time coordinated regulation among facilities; and a deep reinforcement learning algorithm is used to enhance the system's adaptability and resilience under extreme rainfall. This algorithm-driven approach achieves intelligent management of the entire process from design optimization to operational regulation, significantly improving the efficiency of source reduction and process control of rainfall runoff, reducing pipe network renovation costs, and enhancing campus flood safety.

[0038] Reference Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the campus LID facility rainfall-runoff regulation method based on SWMM of the present invention.

[0039] like Figure 2 As shown in the embodiments of the present invention, the campus LID facility rainfall-runoff regulation method based on SWMM includes: Step S10: The data acquisition module collects basic data and real-time monitoring data of the campus area, and outputs the basic data and real-time monitoring data as a standardized dataset.

[0040] Step S20: The model building module constructs an SWMM model based on the standardized dataset, divides the study area into several independent sub-catchment areas and several pipe network nodes according to the confluence path, and outputs the confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipe network, and the initial simulation data of the model.

[0041] Step S30: The algorithm integration module uses the SWMM model as the simulation core and integrates the multi-objective genetic algorithm, model prediction control algorithm and deep reinforcement learning algorithm according to the preset three-level logic to output the corresponding control commands.

[0042] Step S40: The LID facility control module responds to the control command output by the algorithm integration module and dynamically adjusts the operating parameters of the intelligent LID facility.

[0043] Step S50: The feedback calibration module collects the actual runoff data after it has been adjusted by the LID facility control module. It compares the actual runoff data with the runoff data under the corresponding operating conditions output by the SWMM model simulation. When the error exceeds the preset threshold, it dynamically calibrates the runoff generation and collection parameters of the catchment area and the hydraulic characteristic parameters of the pipeline network. The calibrated parameters are then sent back to the algorithm integration module for model and algorithm optimization to form a closed loop.

[0044] In this embodiment, a refined SWMM model is constructed to delineate the catchment area and pipe network; permeable paving, sunken green spaces, and water storage ponds are laid out as LID (Limited Area Development) and gray infrastructure; a multi-objective genetic algorithm is integrated to optimize the design of facility parameters and layout; a model predictive control algorithm is integrated to achieve real-time coordinated regulation among facilities; and a deep reinforcement learning algorithm is used to enhance the system's adaptability and resilience under extreme rainfall. This algorithm-driven approach achieves intelligent management of the entire process from design optimization to operational regulation, significantly improving the efficiency of source reduction and process control of rainfall runoff, reducing pipe network renovation costs, and enhancing campus flood safety.

[0045] This application also provides a campus LID facility rainfall-runoff control device based on SWMM, including a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other through the communication bus. The memory is used to store the campus LID facility rainfall-runoff control program based on SWMM. When the processor executes the program stored in the memory, it implements the above-mentioned campus LID facility rainfall-runoff control method based on SWMM.

[0046] The communication bus mentioned in the above-mentioned campus LID facility rainfall-runoff control equipment based on SWMM can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc.

[0047] The communication interface is used for communication between the aforementioned SWMM-based campus LID facility rainfall-runoff regulation equipment and other equipment.

[0048] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0049] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0050] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0052] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0053] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0054] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0055] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0056] In addition, for technical details not described in detail in this embodiment, please refer to the campus LID facility rainfall-runoff regulation method based on SWMM provided in any embodiment of the present invention, which will not be repeated here.

[0057] Furthermore, it should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0058] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0059] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0060] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

[0061] It is understood that the system provided in the embodiments of the present invention corresponds to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

Claims

1. A campus LID facility rainfall-runoff regulation system based on SWMM, characterized in that, The campus LID facility rainfall-runoff regulation system based on SWMM includes: a data acquisition module, a model building module, an algorithm integration module, an LID facility regulation module, and a feedback calibration module; The data acquisition module is used to collect basic data and real-time monitoring data of the campus area, and output the basic data and real-time monitoring data into a standardized dataset. The basic data includes topographic elevation, land use, pipeline topology, rainfall meteorology, and soil permeability coefficient. The real-time monitoring data includes rainfall, water level, and flow rate. The model building module is used to build an SWMM model based on the standardized dataset, divide the study area into several independent sub-catchments and several pipe network nodes according to the confluence path, and output the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipe network and the initial simulation data of the model. The sub-catchments are used to match the deployment range of LID facilities, and the pipe network nodes are used to monitor the hydraulic status. The algorithm integration module uses the SWMM model as the simulation core and integrates a multi-objective genetic algorithm, a model prediction control algorithm, and a deep reinforcement learning algorithm according to a preset three-level logic to output corresponding control commands. The LID facility control module is used to respond to the control commands output by the algorithm integration module and dynamically adjust the operating parameters of the intelligent LID facility. The spatial layout of the intelligent LID facility corresponds one-to-one with the sub-catchment areas output by the model construction module, and its initial capacity parameters are determined based on the impermeability and area in the catchment area runoff generation parameters. The feedback calibration module is used to collect the actual runoff data after it has been adjusted by the LID facility control module, compare the actual runoff data with the runoff data under the corresponding operating conditions output by the SWMM model simulation, and dynamically calibrate the runoff generation and collection parameters of the catchment area and the hydraulic characteristic parameters of the pipeline network when the error exceeds a preset threshold. The calibrated parameters are then sent back to the algorithm integration module for model and algorithm optimization to form a closed loop.

2. The campus LID facility rainfall-runoff regulation system based on SWMM as described in claim 1, characterized in that, The multi-objective genetic algorithm receives the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipeline network, and the initial simulation data of the model output by the model building module. It uses the area and impermeability of each sub-catchment area in the runoff generation and confluence parameters of the catchment area to determine the upper limit of the LID facility layout area and the minimum proportion constraint of permeable pavement. It uses the design water depth of each node in the hydraulic characteristic parameters of the pipeline network to determine the pipeline network carrying capacity threshold constraint. It uses the total runoff and peak flow in the initial simulation data of the model as optimization benchmarks to solve the target static configuration scheme of the LID facilities and outputs the Pareto target facility configuration scheme as the initial control parameters. The Pareto target facility configuration scheme includes LID facility type, layout location, and initial parameters. The model predictive control algorithm receives real-time monitoring data from the data acquisition module, rolling simulation data from the SWMM model, and initial control parameters output by the multi-objective genetic algorithm. It outputs dynamic control commands to the LID facility regulation module and simultaneously feeds back regulation data to the training environment of the deep reinforcement learning algorithm. The deep reinforcement learning algorithm receives extreme rainfall simulation data output by the model building module, regulation data from the model prediction control algorithm, and error data output by the subsequent feedback calibration module, and trains to generate adaptive enhanced control commands under extreme conditions.

3. The campus LID facility rainfall-runoff regulation system based on SWMM as described in claim 2, characterized in that, The multi-objective genetic algorithm, the model prediction and control algorithm, and the deep reinforcement learning algorithm are called sequentially in runtime order, and the data interaction relationships at different stages include: When in the offline pre-optimization stage, the multi-objective genetic algorithm is invoked, with a single optimization cycle of 24-72 hours. The output of the multi-objective genetic algorithm is used as the initial control parameters of the model prediction control algorithm. When in the online operation phase, the model prediction control algorithm is called cyclically, with a control cycle of 5-15 minutes per cycle. The output of the model prediction control algorithm is a dynamic control command, and the running data is synchronized to the training set of the deep reinforcement learning algorithm. When in an extreme response phase, the deep reinforcement learning algorithm is triggered, wherein the adaptive enhanced control command output by the deep reinforcement learning algorithm is used to replace or supplement the dynamic control command output by the model predictive control algorithm.

4. The campus LID facility rainfall-runoff regulation system based on SWMM as described in claim 2, characterized in that, The multi-objective genetic algorithm uses four objective functions—minimizing peak runoff, minimizing total runoff, minimizing facility construction cost, and maximizing system resilience index—to solve for the target static configuration scheme of LID facilities under constraints, including: The coverage rate of LID facilities in the sub-catchment area is ≤80% of the sub-catchment area; The total area of ​​variable porosity permeable pavement is ≥ 30% of the hardened area of ​​the sub-catchment area; Total volume of the regulating reservoir ≤ catchment area of ​​the sub-catchment area × 50mm; Construction cost ≤ preset budget threshold.

5. The campus LID facility rainfall-runoff regulation system based on SWMM as described in claim 2, characterized in that, The model predictive control algorithm uses the SWMM model as the internal predictive model. The corresponding control logic is that the prediction time domain is set to 1-6 hours, and the rolling optimization cycle is 5-15 minutes. The input data includes real-time monitoring data from the data acquisition module and simulation data of the SWMM model at the current moment. The maximum water level of the pipeline node is controlled at 70%-80% of the design water level, and the total runoff reduction rate is ≥60%. The output control commands include variable pore opening, height of the sunken green space water storage layer, opening of the overflow valve of the regulating tank, and speed of the drainage pump.

6. The campus LID facility rainfall-runoff regulation system based on SWMM as described in claim 1, characterized in that, The intelligent LID facility includes variable porosity permeable pavement, adjustable water storage layer sunken green space, and controllable overflow storage tank. The variable porosity permeable pavement adopts a double-layer structure, with the upper layer being porous concrete with a porosity of 15%-25% and the lower layer being a variable porosity adjustment layer. The dynamic adjustment of the operating parameters is achieved by controlling the pore opening from 0-100% through an electric push rod, so as to realize the dynamic adjustment of the infiltration rate from 0.5-5m / h. The adjustable water storage layer sunken green space is equipped with a liftable water-resistant plate. The dynamic adjustment of the operating parameters is achieved by controlling the thickness of the water storage layer from 50 to 300 mm through a stepper motor, and by using a soil moisture sensor for feedback adjustment, with the error being less than the preset error. The controllable overflow storage tank is equipped with an electric overflow valve and a variable frequency drainage pump. The dynamic adjustment of the operating parameters is achieved by the PLC controller receiving instructions from the algorithm integration module, responding and adjusting the overflow threshold (0.5-2.0m) and pump speed (0-100m) within a preset time. 3 / h.

7. The campus LID facility rainfall-runoff regulation system based on SWMM as described in any one of claims 1 to 6, characterized in that, The feedback calibration module is used to collect actual rainfall, water level, and flow data after LID facility regulation, and to calculate the actual total runoff Qactual and peak flow Qp. The SWMM model was used to simulate the total runoff Q and peak flow Qp under the same operating conditions. The least squares method is used to calculate the error values ​​E1=|Qactual-Qsimulated| / Qactual and E2=|Qpactual-Qpsimulated| / Qpactual. When E1>10% or E2>15%, calibration is initiated. The calibrated model parameters are synchronously updated to the constraints of the multi-objective genetic algorithm to optimize the next generation of Pareto solutions; The calibrated error data is used as a prediction bias compensation term for the model predictive control algorithm to correct the control command; The calibrated extreme condition data are added to the training sample library of the deep reinforcement learning algorithm to improve the model's generalization ability.

8. A method for regulating rainfall-runoff in campus LID facilities based on SWMM, characterized in that, The SWMM-based campus LID facility rainfall-runoff regulation method applies the SWMM-based campus LID facility rainfall-runoff regulation system as described in any one of claims 1 to 7, and the method includes: The data acquisition module collects basic data and real-time monitoring data of the campus area, and outputs the basic data and real-time monitoring data into a standardized dataset. The basic data includes topographic elevation, land use, pipeline topology, rainfall meteorology, and soil permeability coefficient, and the real-time monitoring data includes rainfall, water level, and flow rate. The model building module constructs an SWMM model based on the standardized dataset, divides the study area into several independent sub-catchments and several pipe network nodes according to the confluence path, and outputs the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipe network, and the initial simulation data of the model. The sub-catchments are used to match the deployment range of LID facilities, and the pipe network nodes are used to monitor the hydraulic status. The algorithm integration module uses the SWMM model as the simulation core and integrates multi-objective genetic algorithm, model prediction control algorithm and deep reinforcement learning algorithm according to the preset three-level logic to output corresponding control commands; The LID facility control module responds to the control commands output by the algorithm integration module and dynamically adjusts the operating parameters of the intelligent LID facility. The spatial layout of the intelligent LID facility corresponds one-to-one with the sub-catchment areas output by the model construction module, and its initial capacity parameters are determined based on the impermeability and area in the catchment area's runoff generation parameters. The feedback calibration module collects the actual runoff data after it has been adjusted by the LID facility control module. It compares the actual runoff data with the runoff data under the corresponding operating conditions output by the SWMM model simulation. When the error exceeds the preset threshold, it dynamically calibrates the runoff generation and collection parameters of the catchment area and the hydraulic characteristic parameters of the pipeline network. The calibrated parameters are then sent back to the algorithm integration module for model and algorithm optimization to form a closed loop.

9. The campus LID facility rainfall-runoff regulation method based on SWMM as described in claim 8, characterized in that, The multi-objective genetic algorithm receives the runoff generation and confluence parameters of the catchment area, the hydraulic characteristic parameters of the pipeline network, and the initial simulation data of the model output by the model building module. Using the area and impermeability of each sub-catchment area in the runoff generation and confluence parameters, it determines the upper limit of the LID facility deployment area and the minimum proportion constraint of permeable paving. Using the design water depth of each node in the pipeline network hydraulic characteristic parameters, it determines the pipeline network carrying capacity threshold constraint. Using the total runoff and peak flow rate in the initial simulation data of the model as optimization benchmarks, it solves for the target static configuration scheme of the LID facilities and outputs the Pareto target facility configuration scheme as initial control parameters. The Pareto target facility configuration scheme includes the LID facility type, deployment location, and initial parameters. The model predictive control algorithm receives real-time monitoring data from the data acquisition module, rolling simulation data from the SWMM model, and initial control parameters output by the multi-objective genetic algorithm. It outputs dynamic control commands to the LID facility regulation module and simultaneously feeds back regulation data to the training environment of the deep reinforcement learning algorithm. The deep reinforcement learning algorithm receives extreme rainfall simulation data output by the model building module, regulation data from the model prediction control algorithm, and error data output by the subsequent feedback calibration module, and trains to generate adaptive enhanced control commands under extreme conditions.

10. The campus LID facility rainfall-runoff regulation method based on SWMM as described in claim 9, characterized in that, The multi-objective genetic algorithm, the model prediction and control algorithm, and the deep reinforcement learning algorithm are called sequentially in runtime order, and the data interaction relationships at different stages include: When in the offline pre-optimization stage, the multi-objective genetic algorithm is invoked, with a single optimization cycle of 24-72 hours. The output of the multi-objective genetic algorithm is used as the initial control parameters of the model prediction control algorithm. When in the online operation phase, the model prediction control algorithm is called cyclically, with a control cycle of 5-15 minutes per cycle. The output of the model prediction control algorithm is a dynamic control command, and the running data is synchronized to the deep reinforcement learning algorithm training set. When in an extreme response phase, the deep reinforcement learning algorithm is triggered, wherein the adaptive enhanced control command output by the deep reinforcement learning algorithm is used to replace or supplement the dynamic control command output by the model predictive control algorithm.

Images