Wetland group dynamic water distribution method and device based on multi-target dynamic water distribution optimization

Abstract

The invention provides a wetland group dynamic water distribution method and device based on multi-target dynamic water distribution optimization, and relates to the technical field of dynamic water distribution, and the method comprises the following steps: obtaining predetermined target data; the target data is data obtained after the prediction demand data and the real-time environment data are fused into a multi-dimensional feature vector and feature weights are dynamically adjusted; generating a Pareto optimal solution set based on the target data in combination with an NSGA-III algorithm; dynamically adjusting the target weight through deep reinforcement learning; determining a water distribution scheme based on the Pareto optimal solution set and the target weight; and executing the water distribution scheme through the digital twin platform, monitoring execution process data, and adjusting the Pareto optimal solution set based on the execution process data. In the mode, the multi-dimensional demand conflict is solved, and the adaptability and balance of the water distribution strategy are improved.

Description

Technical Field

[0001] This invention relates to the field of dynamic water distribution technology, and in particular to a dynamic water distribution method and apparatus for wetland communities based on multi-objective dynamic water distribution optimization. Background Technology

[0002] Constructed wetlands are ecological engineering systems built to simulate natural wetlands. They utilize plants, microorganisms, and substrates to purify water, combining water quality improvement, ecological restoration, and water resource regulation. Constructed wetland management has a significant impact on improving water purification efficiency, optimizing resource utilization, and ensuring flood and drought resistance. The core of constructed wetland management lies in dynamic water allocation to ensure ecological balance, synergistic effects between water purification and flood and drought control, improve resource efficiency, and respond to sudden environmental changes.

[0003] In water distribution using related technologies, the use of fixed models for water distribution greatly increases multi-dimensional demand conflicts and reduces the adaptability and balance of water distribution strategies. Summary of the Invention

[0004] In view of this, the purpose of this invention is to provide a dynamic water allocation method and device for wetland groups based on multi-objective dynamic water allocation optimization, which solves the multi-dimensional demand conflict and improves the adaptability and balance of the water allocation strategy.

[0005] In a first aspect, embodiments of the present invention provide a dynamic water allocation method for wetland communities based on multi-objective dynamic water allocation optimization. The method includes: acquiring predetermined target data; the target data being data obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights; generating a Pareto optimal solution set based on the target data and the NSGA-III algorithm; dynamically adjusting the target weights through deep reinforcement learning; determining a water allocation scheme based on the Pareto optimal solution set and the target weights; executing the water allocation scheme through a digital twin platform and monitoring the execution process data, and adjusting the Pareto optimal solution set based on the execution process data.

[0006] In a preferred embodiment of the present invention, the above-mentioned generation of Pareto optimal solution set based on target data and the NSGA-III algorithm includes: initializing the population; randomly generating multiple candidate solutions based on the constraints in the target data; performing non-dominated sorting on the population; dividing the candidate solutions into multiple frontier levels according to the dominance relationship; wherein the first level among the multiple frontier levels is the Pareto frontier solution; generating a set of uniformly distributed reference points using the Das-Dennis method for the three optimization objectives of ecology, economy, and society; performing genetic operations on the current population based on the frontier levels and reference points by simulating binary crossover and polynomial mutation to generate offspring populations; merging the parent populations and offspring populations, and repeating the non-dominated sorting, reference point association, and genetic operations until the pre-set convergence conditions are met, and outputting the Pareto optimal solution set.

[0007] In a preferred embodiment of the present invention, the above-described non-dominated sorting of the population and the division of candidate solutions into multiple frontier levels based on dominance relationships includes: traversing each candidate solution in the population, comparing the first candidate solution with multiple second candidate solutions pairwise; the second candidate solutions are candidate solutions in the population other than the first candidate solution; if the first candidate solution is not inferior to the second candidate solution on all optimization objectives, and is superior to the second candidate solution on at least one optimization objective, then the first candidate solution is determined to dominate the second candidate solution; the solution set dominated by each candidate solution and the number of times it is dominated by other candidate solutions are counted; all candidate solutions with a dominance count of zero are assigned to the first frontier level; the candidate solutions of the first frontier level are removed from the population, and the dominance count of the remaining candidate solutions is updated, until all candidate solutions are assigned to the corresponding frontier level.

[0008] In a preferred embodiment of the present invention, the target weights include: ecological target weights, economic target weights, and social target weights; dynamically adjusting the target weights through deep reinforcement learning includes: pre-assigning initial weights to the ecological target, economic target, and social target respectively; defining a state space, action space, and reward function for reinforcement learning; learning and updating the policy based on real-time environmental data in the state space and the reward function; and adjusting the initial weights using the updated policy to obtain the target weights.

[0009] In a preferred embodiment of the present invention, the state space includes: real-time water level and meteorological forecast data obtained from the target data; the action space is the adjustment action of the ecological target weight, economic target weight and social target weight; and the reward function is determined based on ecological benefits, economic consumption and social security indicators.

[0010] In a preferred embodiment of the present invention, the above-mentioned determination of water allocation scheme based on Pareto optimal solution set and target weight includes: determining the comprehensive score of each solution in Pareto optimal solution set based on dynamically adjusted target weight; selecting the solution with the highest comprehensive score from Pareto optimal solution set as water allocation scheme; and determining execution parameters based on water allocation scheme; the execution parameters include water allocation volume for each wetland, gate opening degree and water allocation time window.

[0011] In a preferred embodiment of the present invention, adjusting the Pareto optimal solution set based on execution process data includes: determining real-time data after the water distribution scheme is executed based on the execution process data; comparing the real-time data with the predicted data of the digital twin platform to obtain the data deviation; if the data deviation exceeds a preset deviation threshold, regenerating the Pareto optimal solution set using the NSGA-III algorithm.

[0012] Secondly, embodiments of the present invention also provide a dynamic water distribution device for wetland communities based on multi-objective dynamic water distribution optimization, comprising: a target data acquisition module for acquiring predetermined target data; the target data is data obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights; a Pareto optimal solution set generation module for generating a Pareto optimal solution set based on the target data and the NSGA-III algorithm; a target weight adjustment module for dynamically adjusting the target weights through deep reinforcement learning; a water distribution scheme determination module for determining a water distribution scheme based on the Pareto optimal solution set and the target weights; and a Pareto optimal solution set adjustment module for executing the water distribution scheme through a digital twin platform and monitoring the execution process data, and adjusting the Pareto optimal solution set based on the execution process data.

[0013] Thirdly, embodiments of the present invention also provide an electronic device, including a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the wetland group dynamic water allocation method based on multi-objective dynamic water allocation optimization described in the first aspect.

[0014] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are invoked and executed by a processor, the computer-executable instructions cause the processor to implement the wetland group dynamic water allocation method based on multi-objective dynamic water allocation optimization described in the first aspect.

[0015] The embodiments of the present invention bring the following beneficial effects: This invention provides a method and apparatus for dynamic water allocation in wetland communities based on multi-objective dynamic water allocation optimization. The method involves acquiring pre-determined target data, which is obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights. Based on the target data, a Pareto optimal solution set is generated using the NSGA-III algorithm. The target weights are then dynamically adjusted through deep reinforcement learning. A water allocation scheme is determined based on the Pareto optimal solution set and the target weights. The water allocation scheme is executed through a digital twin platform, and the execution process data is monitored. The Pareto optimal solution set is then adjusted based on the execution process data. This approach resolves multi-dimensional demand conflicts and improves the adaptability and balance of the water allocation strategy.

[0016] Other features and advantages of this disclosure will be set forth in the following description, or some features and advantages may be inferred from the description or determined without doubt, or may be learned by practicing the techniques described above.

[0017] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a dynamic water allocation method for wetland communities based on multi-objective dynamic water allocation optimization, provided as an embodiment of the present invention; Figure 2 A flowchart of another dynamic water allocation method for wetland groups based on multi-objective dynamic water allocation optimization provided in an embodiment of the present invention; Figure 3 A flowchart of another dynamic water allocation method for wetland groups based on multi-objective dynamic water allocation optimization provided in an embodiment of the present invention; Figure 4 A schematic diagram of a dynamic water distribution device for wetland groups based on multi-objective dynamic water distribution optimization provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Constructed wetlands are ecological engineering systems built to simulate natural wetlands. They utilize plants, microorganisms, and substrates to purify water, combining water quality improvement, ecological restoration, and water resource regulation. Constructed wetland management has a significant impact on improving water purification efficiency, optimizing resource utilization, and ensuring flood and drought resistance. The core of constructed wetland management lies in dynamic water allocation to ensure ecological balance, synergistic effects between water purification and flood and drought control, improve resource efficiency, and respond to sudden environmental changes.

[0022] In water distribution using related technologies, the use of fixed models for water distribution greatly increases multi-dimensional demand conflicts and reduces the adaptability and balance of water distribution strategies.

[0023] Based on this, the present invention provides a dynamic water allocation method and apparatus for wetland groups based on multi-objective dynamic water allocation optimization. This method involves acquiring pre-determined target data, which is obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights. A Pareto optimal solution set is generated based on the target data using the NSGA-III algorithm. The target weights are dynamically adjusted through deep reinforcement learning. A water allocation scheme is determined based on the Pareto optimal solution set and the target weights. The water allocation scheme is executed through a digital twin platform, and the execution process data is monitored. The Pareto optimal solution set is then adjusted based on the execution process data. This approach resolves multi-dimensional demand conflicts and improves the adaptability and balance of the water allocation strategy.

[0024] To facilitate understanding of this embodiment, a detailed description of a dynamic water allocation method for wetland communities based on multi-objective dynamic water allocation optimization, as disclosed in this embodiment of the invention, will be provided first.

[0025] Example 1 This invention provides a dynamic water allocation method for wetland communities based on multi-objective dynamic water allocation optimization. Figure 1 This is a flowchart illustrating a dynamic water allocation method for wetland communities based on multi-objective dynamic water allocation optimization, provided as an embodiment of the present invention. Figure 1 As shown, the dynamic water allocation method for wetland communities based on multi-objective dynamic water allocation optimization may include the following steps: Step S101: Obtain the predetermined target data.

[0026] The target data is the data obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights.

[0027] The target data may include real-time data and predicted data. Real-time data may include current water level, soil moisture and water quality indicators, such as pH and dissolved oxygen. Predicted data may include water demand and weather forecasts for a specified future time output by the federated learning model, such as rainfall and evaporation. The specified future time may be, for example, the next 24 hours.

[0028] Step S102: Generate a Pareto optimal solution set based on the target data and the NSGA-III algorithm.

[0029] Specifically, generating a Pareto optimal solution set based on target data and the NSGA-III algorithm can include: initializing the population; randomly generating multiple candidate solutions based on the constraints in the target data; performing non-dominated sorting on the population and dividing the candidate solutions into multiple frontier levels according to dominance relationships; wherein the first level among the multiple frontier levels is the Pareto front solution; generating a set of uniformly distributed reference points using the Das-Dennis method for the three optimization objectives of ecology, economy, and society; performing genetic operations on the current population based on the frontier levels and reference points through simulated binary crossover and polynomial mutation to generate offspring populations; merging the parent populations with the offspring populations, and repeating the non-dominated sorting, reference point association, and genetic operations until the pre-set convergence conditions are met, and outputting the Pareto optimal solution set.

[0030] The process of non-dominated sorting of the population and dividing candidate solutions into multiple frontier levels based on dominance relationships can include: traversing every candidate solution in the population and comparing the first candidate solution with multiple second candidate solutions pairwise; the second candidate solutions are all candidate solutions in the population other than the first candidate solution; if the first candidate solution is not inferior to the second candidate solution on all optimization objectives and is superior to the second candidate solution on at least one optimization objective, then the first candidate solution is determined to dominate the second candidate solution; the solution set dominated by each candidate solution and the number of times it is dominated by other candidate solutions are counted; all candidate solutions with a dominance count of zero are assigned to the first frontier level; the candidate solutions in the first frontier level are removed from the population, and the dominance count of the remaining candidate solutions is updated, until all candidate solutions are assigned to the corresponding frontier level.

[0031] Specifically, the population is first initialized: N candidate solutions are randomly generated, for example, N=100, and each solution includes: water allocation amount. Gate opening Water distribution sequence .

[0032] Then perform non-dominated sorting: stratify the solution set based on the objective function value using the following formula: , , The first layer of non-dominated solutions is selected by iterating through all candidate solutions and comparing each solution with every other solution. If solution A is not inferior to solution B on all objective functions, and is strictly superior on at least one objective function, then solution A is said to dominate solution B. Solutions not dominated by any other solution are non-dominated solutions. All non-dominated solutions are assigned to the first layer of Pareto fronts to form the initial optimal solution set.

[0033] Then, reference points are generated: these reference points are uniformly distributed in the three-dimensional target space (ecological, economic, and social) to ensure solution set diversity. The number of reference points, M, is represented as: Where H is the number of segments in each objective direction (i.e., the number of evenly divided intervals), for example: H=4 means that each objective dimension is divided into 4 segments; D is the number of objective functions (dimensions) of the optimization problem, for example: D=3 corresponds to three objectives: ecological, economic and social; M is the total number of reference points generated by the calculation.

[0034] Therefore, when H=4 and D=3, A total of 20 uniformly distributed reference points were generated.

[0035] Finally, crossover and mutation are performed: Crossover and mutation generate offspring by simulating binary crossover (SBX) and combining the superior characteristics of the parent generation, ensuring that the solution set inherits advantages; multinomial mutation introduces random perturbations, breaking through local optima and exploring new solution spaces. Both work together to maintain population diversity, avoid premature convergence, and ensure the widespread availability and balanced distribution of Pareto front solutions. Simulated binary crossover (SBX) generates offspring solutions.

[0036] Crossover probability refers to the probability that two parent individuals will exchange genes to generate offspring in a genetic algorithm. It is usually taken as 0.5 to 1. A high probability is conducive to promoting the transmission of good traits and accelerating convergence.

[0037] The mutation probability refers to the probability that an individual's genes will undergo random changes. The mutation probability (n is the dimension of the variable) is usually between 0.001 and 0.01. A low probability is beneficial for maintaining population diversity and avoiding local optima.

[0038] Step S103: Dynamically adjust the target weights through deep reinforcement learning.

[0039] The target weights include: ecological target weights, economic target weights, and social target weights.

[0040] Specifically, dynamically adjusting target weights through deep reinforcement learning can include: pre-assigning initial weights to ecological, economic, and social goals respectively; defining a state space, action space, and reward function for reinforcement learning; learning and updating a policy based on real-time environmental data and the reward function in the state space; and adjusting the initial weights using the updated policy to obtain the target weights.

[0041] The state space includes real-time water level and meteorological forecast data obtained from the target data; the action space consists of the adjustment actions for the weights of ecological, economic, and social objectives; and the reward function is determined based on ecological benefits, economic consumption, and social security indicators.

[0042] Among them, NSGA-III is used to generate Pareto solutions, and deep reinforcement learning (PPO) is used to adjust the weights of ecological, economic and social objectives in real time to resolve multi-dimensional demand conflicts and improve the adaptability and balance of water allocation strategies.

[0043] Among these features, the target weights can be dynamically optimized based on real-time environmental changes. Deep reinforcement learning (PPO) can learn the optimal strategy through interaction by defining a state space (such as real-time water level and weather forecast), an action space (dynamically adjusting the weights of ecological / economic / social objectives), and a reward function (indicators such as vegetation cover and energy consumption).

[0044] The target weights can be ecological target weights, economic target weights, and social target weights. Initially, initial weights are assigned to ecological, economic, and social targets, such as ecological target weights. =0.6, Economic Target Weight =0.3, Social Goal Weight =0.1.

[0045] The reward function involves ecological rewards, economic penalties, and social penalties. For ecological rewards: a 1% increase in vegetation cover results in a +10 reward; for economic penalties: a 1 kWh increase in energy consumption results in a -5 penalty; a 1 yuan overrun in irrigation costs results in a -3 penalty; for social penalties: exceeding the water level... Punishment -20.

[0046] Specifically, interaction data (dynamic-action-reward) can be collected based on the current strategy, and action advantage value can be calculated using generalized advantage estimation (GAE). .

[0047] Therefore, maximize the objective function This can be expressed by the following formula: .

[0048] in, For the current policy (neural network) in state Select action The probability of; The probability of choosing the same action for the old strategy (the strategy before the update) is used to constrain the magnitude of the strategy update; The dominance function measures the action. The relative merits of strategies compared to the average (e.g., the degree of comprehensive improvement in ecological, economic, and social goals). The cut-off threshold limits the policy update step size to prevent training instability.

[0049] Finally, the target weights are adjusted based on the optimized strategy. For example, the initial weights of ecological, economic, and social goals are adjusted as follows: Ecological goal weight. =0.65, economic target weight =0.25, Social Goal Weight =0.1.

[0050] Step S104: Determine the water allocation scheme based on the Pareto optimal solution set and the target weight.

[0051] Specifically, the water allocation scheme is determined based on the Pareto optimal solution set and the target weights, including: determining the comprehensive score of each solution in the Pareto optimal solution set based on the dynamically adjusted target weights; selecting the solution with the highest comprehensive score from the Pareto optimal solution set as the water allocation scheme; and determining the execution parameters based on the water allocation scheme. The execution parameters include the water allocation volume of each wetland, the gate opening degree, and the water allocation time window.

[0052] To ensure the generated water distribution scheme meets the actual constraints, constraint processing and scheme verification are required. For example, if the water volume exceeds the limit, a penalty is imposed on solutions that violate the constraints. The penalty term is as follows: ;in, Indicates the penalty coefficient. .

[0053] This requires verifying the feasibility of the plan, and calculating the transmission time based on the channel length L and the flow velocity v. ,make sure ;examine The water demand must be greater than the ecological water demand, that is For example, the water requirement of mangrove ecosystems .

[0054] Specifically, the solution with the highest overall score is selected from the Pareto front, for example, with an ecological score accounting for 60%; then the water allocation for each wetland is determined. For example, the concentration in mangrove areas is 1200 m³ / d, and in agricultural areas it is 800 m³ / d; and the execution parameters, such as gate opening, are determined. =75%, with water distribution time from 6:00 to 8:00 daily; the effect of the scheme is simulated in the digital twin platform to verify the water level changes and ecological impact.

[0055] Step S105: Execute the water distribution plan through the digital twin platform and monitor the execution process data, and adjust the Pareto optimal solution set based on the execution process data.

[0056] Among these features, a high-precision three-dimensional hydrological model of the wetland cluster can be constructed in real time using digital twin technology. This model integrates sensor networks (water level, meteorology, water quality) with external data (weather forecasts, remote sensing images) to dynamically perceive the environmental status. After generating a water allocation plan based on a multi-objective optimization model, the plan's effects are simulated using a digital twin platform to verify its feasibility and risks. Water allocation commands are then precisely executed through intelligent gates (PID control) and solar-powered water pumps. In conjunction with the isolated forest algorithm, anomalies (such as equipment failure or sudden rainstorms) are detected in real time, triggering an emergency mode (activating backup water sources and adjusting weights to prioritize the core ecological area). Simultaneously, a closed-loop feedback mechanism continuously calibrates model parameters to ensure that the system dynamically optimizes with environmental changes, achieving precise, intelligent, and highly robust water resource allocation.

[0057] Specifically, adjusting the Pareto optimal solution set based on execution process data can include: determining real-time data after the water distribution scheme is executed based on the execution process data; comparing the real-time data with the predicted data from the digital twin platform to obtain the data deviation; and if the data deviation exceeds a pre-set deviation threshold, regenerating the Pareto optimal solution set using the NSGA-III algorithm.

[0058] The purpose of dynamic feedback and iterative optimization is to continuously optimize the model based on actual execution results. The model is updated by real-time monitoring of indicators such as water level, vegetation cover, and energy consumption after execution. If the actual results deviate from the prediction (e.g., water level error ≥ 5%), NSGA-III is triggered to regenerate the solution set and update the federated learning model parameters, thereby improving the accuracy of water demand prediction.

[0059] Among them, regarding the digital twin modeling of the digital twin platform: a three-dimensional hydrological model of wetlands (resolution 1m×1m) can be constructed based on remote sensing and GIS, integrating parameters such as soil permeability coefficient and vegetation transpiration rate, accessing sensor, meteorological and satellite data in real time, and calibrating the model accuracy through historical data (water level error ≤5%), establishing a virtual-real synchronized twin platform to support dynamic monitoring and closed-loop control.

[0060] Specifically, three-dimensional hydrological modeling is performed: a three-dimensional topographic model of the wetland complex (1m×1m resolution) is established based on remote sensing imagery and a geographic information system (GIS). Hydrological parameters are integrated, including soil permeability coefficient, vegetation transpiration rate, and channel slope. Data access is achieved through real-time access to sensor data (water level, soil moisture, water quality), weather forecasts (rainfall, evaporation), and satellite remote sensing data (NDVI vegetation index). Model calibration is performed using historical data to verify model accuracy (e.g., water level simulation error ≤5%) and to adjust parameters (e.g., Manning coefficient).

[0061] Specifically, real-time data monitoring and synchronization are performed: data acquisition, with the sensor network uploading data (water level, soil moisture, equipment status) every 5 minutes; data cleaning, using wavelet decomposition and cubic polynomial for noise reduction and filling; and digital twin synchronization, updating parameters such as real-time water level and vegetation coverage in the 3D model to generate a visualization interface.

[0062] Specifically, optimal control commands can be generated based on the results of digital twin simulations, including: scenario simulation: inputting the water distribution scheme generated by the multi-objective optimization model to simulate water level changes, vegetation response and energy consumption in the next 6 hours; risk assessment: detecting potential risks (such as water level exceeding limits or channel blockage); command generation: if the simulation results meet the constraints, output execution commands (gate opening, pump flow); otherwise, trigger the optimization model to recalculate.

[0063] Specifically, decision-making instructions can be translated into physical equipment actions, including: intelligent gate control and setting target water levels. The error is calculated using the following formula. The output is controlled by the following formula: ,in, , , Adjust the gate opening (precision) (1cm), to bring the actual water level close to the target value; solar-powered water pump control, based on water distribution... Adjust the flow rate: Flow rate = Qitirrigation (e.g.) =1200 m3 / d, t=8 h, flow rate=150 m3 / h), dynamically adjust the pump speed (accuracy) 5%), matching the water distribution sequence.

[0064] Specifically, it can perform anomaly detection and emergency response, including: For anomaly detection, the Isolation Forest algorithm was used: Constructing an isolation tree (iTree): Randomly selecting features and split values, recursively dividing the data space into sub-regions until each sub-region contains only one sample or reaches the tree height limit; Anomaly characteristics: Due to their large differences from most data, anomalies are usually isolated in shallower layers (shorter paths); Calculating path length: Recording the average path length required for each sample to be isolated in all isolation trees. The shorter the path length, the higher the probability of sample anomalies; Calculate the anomaly score: Analyze the sensor data stream and calculate the anomaly score: ,in, This represents the mean path length of each data point in the isolated tree. The path length normalization factor is used; a score close to 1 indicates an anomaly, while a score close to 0 indicates normality. Specifically, if the anomaly score is greater than 0.7, it is considered an abnormal event (such as being affected by heavy rain, equipment failure, etc.).

[0065] Regarding the emergency mode trigger: Backup water sources are activated, and underground wells or reservoirs are automatically opened to ensure water supply to the core ecological area; a conservative strategy is switched, and in mangrove reserves: minimum ecological water demand is maintained. =1000m3 / d, agricultural areas: water allocation reduced to 50% (e.g.) =400m3 / d); manual intervention notification, sending alerts to administrators, such as via APP push and SMS.

[0066] Specifically, system parameters can be optimized based on actual performance results: Performance evaluation: Compare the deviations between the digital twin predictions and the actual water level and vegetation cover (allowable error ≤ 5%); Parameter calibration: If the deviation exceeds the limit, adjust the hydrological model parameters (such as permeability coefficient) or the weights of the federated learning model; Model retraining: Collect new data and update the LSTM water demand prediction model (full training once every 7 days).

[0067] Specifically, it can provide a human-computer interaction interface to support decision monitoring: Cloud platform: WebGL 3D interface displays real-time water level, water distribution scheme simulation, and equipment status; historical data backtracking (such as water level change curves over the past 30 days). Mobile APP: Real-time receiving of alerts (such as "Water level in the mangrove area is below the threshold"); remote manual control (such as forcibly closing a gate).

[0068] This approach employs NSGA-III to generate Pareto solutions and combines deep reinforcement learning (PPO) to adjust the weights of ecological, economic, and social objectives in real time. This resolves multi-dimensional demand conflicts, improves the adaptability and balance of water allocation strategies, and constructs a high-precision three-dimensional hydrological model. It integrates PID algorithms and solar-powered equipment to precisely execute water allocation commands. Real-time data feedback dynamically calibrates the model, achieving closed-loop management of the entire process from "perception to decision-making to execution to optimization." Based on the isolated forest algorithm, it detects data anomalies in real time (such as heavy rain or equipment failure), automatically activates backup water sources, and adjusts water allocation strategies to ensure water supply to core ecological areas, enhancing the system's resilience and emergency response efficiency.

[0069] The wetland group dynamic water allocation method based on multi-objective dynamic water allocation optimization provided in this invention can achieve the following: First, it acquires pre-determined target data, which is obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights. Then, it generates a Pareto optimal solution set based on the target data using the NSGA-III algorithm. Next, it dynamically adjusts the target weights through deep reinforcement learning. Finally, it determines the water allocation scheme based on the Pareto optimal solution set and the target weights. The water allocation scheme is executed through a digital twin platform, and the execution process data is monitored. The Pareto optimal solution set is then adjusted based on the execution process data. This approach resolves multi-dimensional demand conflicts and improves the adaptability and balance of the water allocation strategy.

[0070] Example 2 This invention also provides another dynamic water allocation method for wetland groups based on multi-objective dynamic water allocation optimization; this method is implemented on the basis of the method in the above embodiments; the method focuses on describing the specific implementation of obtaining predetermined target data.

[0071] Figure 2 A flowchart of another wetland group dynamic water allocation method based on multi-objective dynamic water allocation optimization provided in an embodiment of the present invention is shown below. Figure 2 As shown, acquiring predetermined target data may include the following steps: Step S201: Real-time environmental data of the wetland cluster is collected through a multi-source sensor network with pre-optimized layout.

[0072] Multi-source sensor networks can include soil moisture monitoring equipment, water quality monitoring equipment, meteorological monitoring equipment, etc.

[0073] Environmental data may include soil moisture data, water quality data, meteorological data, etc.

[0074] Step S202: Preprocess the environmental data to obtain standardized environmental data.

[0075] Specifically, preprocessing environmental data to obtain standardized environmental data may include: using wavelet decomposition to remove high-frequency noise from the environmental data and reconstructing the denoised data; after reconstructing the denoised data, filling in missing data points through cubic polynomial interpolation to obtain standardized data.

[0076] In order to improve data quality, environmental data can be preprocessed.

[0077] The process of removing high-frequency noise from environmental data using wavelet decomposition and reconstructing the denoised data can include: removing high-frequency noise from environmental data using wavelet decomposition using the following formula: ;in, is the original signal, representing the input data to be decomposed; k is the translation parameter, representing the position of the wavelet function on the time axis; The scaling coefficients are derived from the scaling function. Generation, reflecting the low-frequency components of the signal, ; For detail coefficients, derived from wavelet functions generate, .

[0078] Among them, soft thresholds can be applied. High-frequency noise is removed from detail coefficients; among which, Let N be the noise standard deviation and N be the number of data points. The denoised data is reconstructed using the processed coefficients. .

[0079] Specifically, filling in missing data points using cubic polynomial interpolation to obtain standardized data can include: filling in missing data points... Based on adjacent time points and Data , Construct a cubic polynomial: ; satisfy boundary conditions and Furthermore, the first and second derivatives are continuous.

[0080] Therefore, for missing points The interpolation is calculated using a polynomial for the corresponding interval: Standardized data was obtained after filling in the missing points.

[0081] Step S203: Process the standardized environmental data based on the pre-trained federated learning model and output the predicted water demand values ​​for each wetland.

[0082] Among them, the federated learning model (LSTM+FedAvg) can be used to predict the water demand of each wetland in the future target time based on the input historical water level, rainfall and evaporation.

[0083] Specifically, the overall structure of the local model can be: ;in, , , , These are the weight matrices of the LSTM unit, corresponding to the forget gate, input gate, output gate, and candidate memory unit, respectively; , , , For the bias term of the LSTM cell; , , These are the activation values ​​for the forget gate, input gate, and output gate, respectively. , These are candidate memory units and the memory unit at the current moment; This represents the current hidden state.

[0084] The loss function is represented by the mean squared error: Used to measure predicted water demand Compared with actual value The deviation.

[0085] Step S204: The predicted water demand value is fused with the real-time data into a multi-dimensional feature vector and the feature weights are dynamically adjusted to obtain the target data.

[0086] Specifically, fusing water demand forecasts and real-time data into a multi-dimensional feature vector and dynamically adjusting the feature weights to obtain target data can include: normalizing and merging water demand forecasts and real-time data into a multi-dimensional feature vector; and dynamically adjusting feature weights based on wetland type to obtain target data.

[0087] Each wetland node uploads local model parameters. The following formula is used to calculate a weighted average based on the proportion of data volume to ensure that the global model reflects the data distribution of all wetlands: ;in, Let k be the number of data points in the k-th node. .

[0088] By parameters Add Gaussian noise differential privacy protection. Noise levels are adjusted according to privacy budgets, such as =1.

[0089] Specifically, the preprocessed environmental data is input into the global LSTM model, which outputs a water demand sequence: The Monte Carlo Dropout method is used to calculate the confidence intervals of the predicted values: ;in, To predict the average water demand, Here, M is the standard deviation of the predicted value, M is the number of Monte Carlo samplings, and z is the Z-score. For example, z=1.96 corresponds to a 95% confidence level.

[0090] Specifically, data such as water demand forecasts, real-time water levels, and weather forecasts are normalized and merged into a multi-dimensional feature vector: Then, based on the wetland type, such as ecological zone / agricultural zone / flood control zone, the feature weights are dynamically adjusted.

[0091] The wetland dynamic water allocation method based on multimodal sensing provided in this invention can collect environmental data of wetland groups in real time through a pre-optimized multi-source sensor network. The environmental data is preprocessed to obtain standardized environmental data. A pre-trained federated learning model processes the standardized environmental data, outputting predicted water demand values ​​for each wetland. These predicted values ​​are then fused with real-time data to form a multi-dimensional feature vector, and the feature weights are dynamically adjusted to obtain target data. A water allocation scheme is generated based on the target data and executed after simulation verification using a digital twin platform. This approach improves ecological connectivity, ensures ecological balance, and saves resources.

[0092] Example 3 This invention also provides another dynamic water distribution method for wetland groups based on multi-objective dynamic water distribution optimization; this method is implemented on the basis of the method in the above embodiments; this method focuses on describing the specific implementation of the layout optimization method of the multi-source sensor network.

[0093] This invention also provides another dynamic water allocation method for wetland groups based on multimodal sensing; this method is implemented on the basis of the method in the above embodiments; this method focuses on describing the specific implementation of the layout optimization method of the multi-source sensor network.

[0094] Figure 3 A flowchart of another dynamic water allocation method for wetland groups based on multimodal sensing provided in an embodiment of the present invention is shown below. Figure 3 As shown, the layout optimization method for this multi-source sensor network can include the following steps: Step S301: Obtain the geographical coordinates of the wetland area to form a wetland plane.

[0095] The geographical coordinates of the wetland area can be latitude and longitude or two-dimensional plane coordinates.

[0096] Step S302: Randomly select a preset number of initial cluster centers within the wetland plane area; the preset number is the number of sensor nodes in the multi-source sensor network.

[0097] K initial cluster centers are randomly selected within the wetland plane, where K is the number of sensor nodes, which can be determined based on the wetland area and resolution requirements.

[0098] Step S303: Assign the geographic coordinates of each wetland area to the nearest cluster center to form multiple clusters, and determine the mean geographic coordinates of each point in each cluster as the target cluster center.

[0099] The geographic coordinates of each wetland area can be calculated using the following formula. Determine the geographical coordinates of the wetland area With all cluster centers Euclidean distance: The following formula is used to assign the geographic coordinates of wetland areas to the clusters corresponding to the cluster centers with the smallest distances, based on Euclidean distance: For each cluster, the mean geographic coordinates of each point in the cluster are determined as the target cluster center.

[0100] Step S304: Determine the target cluster center and the change in center between cluster centers.

[0101] Specifically, determining the target cluster centers and the centroid variation among cluster centers can include: redetermining the centroid coordinates of each cluster using the following formula: ;in, This represents the set of data points in the j-th cluster. and The target cluster centers are represented; after each recalculation, the change in center between the target cluster centers and the existing cluster centers is determined using the following formula: ;in, and Indicates the cluster center.

[0102] Step S205: When the center change is less than the preset change threshold, or after reaching the maximum number of iterations, output the optimal arrangement position of a preset number of sensor nodes.

[0103] The threshold for the change can be set to 1~1.5m.

[0104] The wetland dynamic water allocation method based on multi-objective dynamic water allocation optimization provided in this invention employs the K-means clustering algorithm to optimize the sensor network layout. By analyzing the geographical coordinates of the wetland area, sensor nodes are distributed in optimal locations to maximize monitoring coverage and minimize data redundancy. Iterative updates to the cluster centers ensure balanced node coverage of key areas, improving data acquisition efficiency, reducing communication energy consumption and equipment redundancy, and meeting the needs of dynamic wetland monitoring.

[0105] Example 4 Corresponding to the above method embodiments, this invention provides a dynamic water distribution device for wetland communities based on multi-objective dynamic water distribution optimization. Figure 4 This is a schematic diagram of a dynamic water distribution device for wetland communities based on multi-objective dynamic water distribution optimization, provided in an embodiment of the present invention. Figure 4 As shown, the wetland group dynamic water distribution device based on multi-objective dynamic water distribution optimization may include: The target data acquisition module 401 is used to acquire pre-determined target data; the target data is the data obtained by fusing the predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights.

[0106] The Pareto optimal solution set generation module 402 is used to generate a Pareto optimal solution set based on the target data and the NSGA-III algorithm.

[0107] The target weight adjustment module 403 is used to dynamically adjust the target weights through deep reinforcement learning.

[0108] The water distribution scheme determination module 404 is used to determine the water distribution scheme based on the Pareto optimal solution set and the target weight.

[0109] The Pareto optimal solution set adjustment module 405 is used to execute the water distribution plan through the digital twin platform and monitor the execution process data, and adjust the Pareto optimal solution set based on the execution process data.

[0110] The wetland group dynamic water distribution device based on multi-objective dynamic water distribution optimization provided in this invention can acquire pre-determined target data. This target data is obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights. Based on the target data, a Pareto optimal solution set is generated using the NSGA-III algorithm. The target weights are dynamically adjusted through deep reinforcement learning. A water distribution scheme is determined based on the Pareto optimal solution set and the target weights. The water distribution scheme is executed through a digital twin platform, and the execution process data is monitored. The Pareto optimal solution set is then adjusted based on the execution process data. This approach resolves multi-dimensional demand conflicts and improves the adaptability and balance of the water distribution strategy.

[0111] In some embodiments, the Pareto optimal solution set generation module is further configured to initialize the population, randomly generate multiple candidate solutions based on the constraints in the target data, perform non-dominated sorting on the population, and divide the candidate solutions into multiple frontier levels according to the dominance relationship; wherein, the first level among the multiple frontier levels is the Pareto frontier solution; generate a set of uniformly distributed reference points using the Das-Dennis method for the three optimization objectives of ecology, economy and society; perform genetic operations on the current population based on the frontier levels and reference points by simulating binary crossover and polynomial mutation to generate offspring populations; merge the parent populations with the offspring populations, and repeat the non-dominated sorting, reference point association and genetic operations until the pre-set convergence conditions are met, and output the Pareto optimal solution set.

[0112] In some embodiments, the Pareto optimal solution set generation module is further configured to traverse each candidate solution in the population, compare the first candidate solution with each of the multiple second candidate solutions pairwise; the second candidate solutions are candidate solutions in the population other than the first candidate solution; if the first candidate solution is not inferior to the second candidate solution on all optimization objectives, and is superior to the second candidate solution on at least one optimization objective, then the first candidate solution is determined to dominate the second candidate solution; the solution set dominated by each candidate solution and the number of times it is dominated by other candidate solutions are counted; all candidate solutions with a domination count of zero are assigned to the first frontier level; the candidate solutions of the first frontier level are removed from the population, and the domination count of the remaining candidate solutions is updated, until all candidate solutions are assigned to the corresponding frontier level.

[0113] In some embodiments, the target weights include: ecological target weights, economic target weights, and social target weights; the target weight adjustment module is further used to pre-allocate initial weights to the ecological target, economic target, and social target respectively; define the state space, action space, and reward function for reinforcement learning; perform policy learning and updating based on real-time environmental data in the state space and the reward function; and adjust the initial weights using the updated policy to obtain the target weights.

[0114] In some embodiments, the state space includes: real-time water level and meteorological forecast data obtained from the target data; the action space is the adjustment action of the ecological target weight, economic target weight and social target weight; and the reward function is determined based on ecological benefits, economic consumption and social security indicators.

[0115] In some embodiments, the water allocation scheme determination module is further configured to determine the comprehensive score of each solution in the Pareto optimal solution set based on the dynamically adjusted target weight; select the solution with the highest comprehensive score from the Pareto optimal solution set as the water allocation scheme; and determine the execution parameters based on the water allocation scheme; the execution parameters include the water allocation volume of each wetland, the gate opening degree, and the water allocation time window.

[0116] In some embodiments, the Pareto optimal solution set adjustment module is further configured to determine the real-time data after the water distribution scheme is executed based on the execution process data; compare the real-time data with the prediction data of the digital twin platform to obtain the data deviation; if the data deviation exceeds the preset deviation threshold, the Pareto optimal solution set is regenerated using the NSGA-III algorithm.

[0117] The device provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0118] Example 5 This invention also provides an electronic device for running the above-described wetland group dynamic water allocation method based on multi-objective dynamic water allocation optimization; see also Figure 5 The diagram shows the structure of an electronic device, which includes a memory 500 and a processor 501. The memory 500 stores one or more computer instructions, which are executed by the processor 501 to implement the above-mentioned dynamic water distribution method for wetland groups based on multi-objective dynamic water distribution optimization.

[0119] Furthermore, Figure 5 The electronic device shown also includes a bus 502 and a communication interface 503. The processor 501, the communication interface 503 and the memory 500 are connected via the bus 502.

[0120] The memory 500 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 503 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 502 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0121] Processor 501 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 501 or by instructions in software form. Processor 501 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a readily available storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 500, and processor 501 reads information from memory 500 and, in conjunction with its hardware, completes the steps of the method described in the foregoing embodiments.

[0122] This invention also provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are called and executed by a processor, they cause the processor to implement the aforementioned dynamic water allocation method for wetland groups based on multi-objective dynamic water allocation optimization. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0123] The computer program product for performing a dynamic water allocation method for wetland groups based on multi-objective dynamic water allocation optimization provided in this embodiment of the invention includes a computer-readable storage medium storing non-volatile program code executable by a processor. The instructions included in the program code can be used to execute the methods in the preceding method embodiments. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0124] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0125] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0126] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0127] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0128] If the functionality is implemented as a software functional unit and sold or used as an independent product, it can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0129] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, 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, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A wetland group dynamic water distribution method based on multi-objective dynamic water distribution optimization, characterized in that, The method includes: Acquire predetermined target data; the target data is the data obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights. Based on the target data, a Pareto optimal solution set is generated using the NSGA-III algorithm; Dynamically adjust target weights through deep reinforcement learning; The water allocation scheme is determined based on the Pareto optimal solution set and the target weights; The water distribution scheme is executed through a digital twin platform, and the execution process data is monitored. The Pareto optimal solution set is then adjusted based on the execution process data.

2. The method of claim 1, wherein, The generation of the Pareto optimal solution set based on the target data and the NSGA-III algorithm includes: Initialize the population and randomly generate multiple candidate solutions based on the constraints in the target data; The population is sorted non-dominated, and the candidate solutions are divided into multiple front levels according to the dominance relationship; wherein, the first level among the multiple front levels is the Pareto front solution; To address the three optimization objectives of ecology, economy, and society, the Das-Dennis method is used to generate a set of uniformly distributed reference points. Based on the aforementioned frontier level and the reference point, genetic operations are performed on the current population through simulated binary crossover and polynomial mutation to generate offspring populations. The parent population is merged with the offspring population, and non-dominated sorting, reference point association, and genetic operations are repeated until the pre-set convergence condition is met, and the Pareto optimal solution set is output.

3. The method of claim 2, wherein, The non-dominated sorting of the population, and the division of the candidate solutions into multiple frontier levels based on dominance relationships, includes: Iterate through each candidate solution in the population, and compare the first candidate solution with each of the multiple second candidate solutions in pairs; the second candidate solutions are the candidate solutions in the population other than the first candidate solution. If the first candidate solution is not inferior to the second candidate solution on all optimization objectives, and is superior to the second candidate solution on at least one optimization objective, then the first candidate solution is determined to dominate the second candidate solution. Count the solution set dominated by each candidate solution and the number of times it is dominated by other candidate solutions; All candidate solutions with a dominance count of zero are assigned to the first frontier level; Remove candidate solutions from the first frontier level from the population and update the dominance count of the remaining candidate solutions until all candidate solutions are assigned to the corresponding frontier level.

4. The method of claim 1, wherein, The target weights include: ecological target weights, economic target weights, and social target weights; the dynamic adjustment of target weights through deep reinforcement learning includes: Initial weights are pre-assigned to the ecological objectives, the economic objectives, and the social objectives, respectively. Define the state space, action space, and reward function for reinforcement learning; Based on the real-time environmental data in the state space and the reward function, policy learning and updating are performed; The target weights are obtained by adjusting the initial weights using the updated strategy.

5. The method according to claim 4, characterized in that, The state space includes: real-time water level and meteorological forecast data obtained from the target data; the action space is the adjustment action of the ecological target weight, the economic target weight and the social target weight; the reward function is determined based on ecological benefits, economic consumption and social security indicators.

6. The method according to claim 1, characterized in that, The process of determining the water allocation scheme based on the Pareto optimal solution set and the target weights includes: Based on the dynamically adjusted target weights, the comprehensive score of each solution in the Pareto optimal solution set is determined; The solution with the highest comprehensive score is selected from the Pareto optimal solution set as the water allocation scheme. Based on the water distribution plan, the execution parameters are determined; the execution parameters include the water distribution volume of each wetland, the gate opening degree, and the water distribution time window.

7. The method according to claim 1, characterized in that, The adjustment of the Pareto optimal solution set based on the execution process data includes: Based on the execution process data, determine the real-time data after the water distribution plan is executed; The real-time data is compared with the predicted data from the digital twin platform to obtain the data deviation; If the data deviation exceeds a preset deviation threshold, the Pareto optimal solution set is regenerated using the NSGA-III algorithm.

8. A dynamic water distribution device for wetland communities based on multi-objective dynamic water distribution optimization, characterized in that, The device includes: The target data acquisition module is used to acquire predetermined target data; the target data is the data obtained by fusing predicted demand data and real-time environmental data into a multi-dimensional feature vector and dynamically adjusting the feature weights. The Pareto optimal solution set generation module is used to generate a Pareto optimal solution set based on the target data and the NSGA-III algorithm. The target weight adjustment module is used to dynamically adjust the target weights through deep reinforcement learning; The water distribution scheme determination module is used to determine the water distribution scheme based on the Pareto optimal solution set and the target weight; The Pareto optimal solution set adjustment module is used to execute the water distribution scheme through a digital twin platform and monitor the execution process data, and adjust the Pareto optimal solution set based on the execution process data.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing computer-executable instructions that can be executed by the processor, the processor executing the computer-executable instructions to implement the wetland group dynamic water allocation method based on multi-objective dynamic water allocation optimization as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement the wetland group dynamic water allocation method based on multi-objective dynamic water allocation optimization as described in any one of claims 1 to 7.

Images