Cross-regional water transfer project intelligent scheduling method and system

Abstract

The invention relates to the technical field of intelligent water conservancy, and discloses a cross-regional water transfer project intelligent scheduling method and system, and the method comprises the steps: building a digital twin system based on a geographic information system, hydrological monitoring data and a spatial topological structure, and integrating a meteorological evolution prediction model, a basin hydrological response model and a water demand prediction model; predicting a water demand and an adjustable water amount by using a space-time convolutional neural network and a gating circulation unit; establishing a multi-objective optimization model taking water supply benefit, ecological influence and energy consumption cost as optimization objectives; an optimal water transfer scheme is generated through a Markov decision process and multi-agent cooperation; and carrying out robustness evaluation on the scheme and generating an emergency scheduling plan. According to the method, the scheduling efficiency and adaptability of the water transfer project are remarkably improved, and efficient configuration of water resources and quick response under extreme situations are achieved.

Description

Technical Field

[0001] The present invention relates to the field of smart water conservancy technology, and in particular to a method and system for intelligent scheduling of cross-regional water diversion projects. Background Art

[0002] Interregional water transfer projects are crucial infrastructure for alleviating the uneven spatial and temporal distribution of water resources, but their scheduling faces numerous technical challenges. Traditional scheduling methods rely primarily on manual experience and static rules, making them difficult to address, among other constraints, dynamic changes in meteorological conditions, fluctuating water demand, and ecological protection requirements. Existing technologies have the following limitations:

[0003] Insufficient data integration and the fragmented and independent nature of meteorological, hydrological, and demand data limit the accuracy of forecasting models. A single optimization objective, focused primarily on water supply efficiency while ignoring ecological impacts and energy costs, makes it difficult to maximize overall benefits. Emergency response is delayed, lacking the ability to proactively predict and rapidly make decisions for extreme scenarios (such as droughts and floods). The dispatching system lacks coordination, with decision-making at each node relying on central control and lacking flexibility. With the intensification of climate change and the growth of water demand, an intelligent dispatching method is urgently needed that can enhance the scientific nature, adaptability, and robustness of water diversion projects through multi-source data fusion and multi-objective dynamic optimization. Summary of the Invention

[0004] The present invention provides an intelligent scheduling method and system for cross-regional water diversion projects, the main purpose of which is to solve the problems of difficulty in multi-source data fusion, insufficient dynamic scheduling optimization, and delayed emergency response in extreme scenarios in cross-regional water diversion projects.

[0005] To achieve the above objectives, the present invention provides an intelligent scheduling method for a cross-regional water diversion project, comprising:

[0006] Based on geographic information systems, hydrological monitoring data, and spatial topology, a digital twin system integrating meteorological evolution prediction models, watershed hydrological response models, and water demand forecasting benchmark models is established. The digital twin system is connected to meteorological satellite data, hydrological sensor data, and water demand data in real time.

[0007] Predicting the water demand of each water receiving area and the adjustable water volume of the water source in the future period based on the input data of the digital twin system;

[0008] Based on the prediction results, a multi-objective optimization model is established with water supply benefits, ecological impact and energy consumption costs as optimization targets;

[0009] generating an optimal water diversion scheme for a cross-regional water diversion project based on the multi-objective optimization model and the Markov decision process;

[0010] The robustness evaluation of the optimal water diversion scheme is carried out, and emergency dispatch plans under extreme scenarios are generated based on the evaluation results of the robustness evaluation.

[0011] Optionally, the digital twin system integrating the meteorological evolution prediction model, the watershed hydrological response model and the water demand prediction benchmark model is established based on the geographic information system, hydrological monitoring data and spatial topology structure, including:

[0012] Compare and analyze real-time meteorological satellite data with historical meteorological patterns to establish a meteorological evolution prediction model;

[0013] Construct a watershed hydrological response model based on the spatiotemporal distribution characteristics of hydrological sensor data;

[0014] Based on the periodic characteristics of water demand data and regional water use patterns, a water demand forecasting benchmark model is established;

[0015] The meteorological evolution prediction model, the watershed hydrological response model and the water demand prediction benchmark model are integrated to obtain a digital twin system.

[0016] Optionally, the predicting of the water demand of each water receiving area and the adjustable water volume of the water source in a future time period based on the input data of the digital twin system includes:

[0017] A spatiotemporal convolutional neural network is used to extract the spatiotemporal distribution feature matrix of precipitation from meteorological satellite data.

[0018] The gated recurrent unit is used to capture the temporal variation characteristics of runoff in the hydrological sensor data;

[0019] The spatial and temporal distribution characteristic matrix of precipitation and the temporal variation characteristic sequence of runoff are integrated to generate a joint prediction result including the water demand prediction value and the adjustable water volume prediction value.

[0020] Optionally, the multi-objective optimization model based on the prediction results is established with water supply benefits, ecological impacts and energy consumption costs as optimization targets, including:

[0021] Construct a water supply benefit evaluation function, the output value of which reflects the quantitative index of water supply priority and satisfaction of the water receiving area;

[0022] Design an ecological impact assessment function that includes the river ecological base flow constraint, and the output value of the ecological impact assessment function represents the quantitative index of the degree to which the river ecological base flow meets the standard;

[0023] Establish an energy cost function that reflects the operating efficiency of the pumping station. The output value of the energy cost function reflects the quantitative index of the operating efficiency of the pumping station.

[0024] A multi-objective optimization model is constructed based on water supply benefit evaluation function, ecological impact evaluation function and energy consumption cost function.

[0025] Optionally, generating an optimal water diversion plan for a cross-regional water diversion project based on the multi-objective optimization model and the Markov decision process includes:

[0026] Convert the multi-objective optimization model into a collaborative decision-making problem of Markov decision process;

[0027] Formulate multi-agent collaboration rules based on the spatial topological structure of cross-regional water diversion projects;

[0028] The policy gradient algorithm is applied to solve the problem and output the optimal scheduling instruction set for each node in the cross-regional water diversion project.

[0029] Optionally, converting the multi-objective optimization model into a collaborative decision-making problem of a Markov decision process includes:

[0030] Define a multidimensional state vector containing the water level, flow rate and energy consumption of each node;

[0031] Construct a set of feasible scheduling actions, where the set of feasible scheduling actions satisfies the hydraulic constraints of the pipe network;

[0032] A state transition probability matrix considering the time delay effect is established to generate a Markov decision process framework.

[0033] Optionally, the robustness evaluation of the optimal water diversion scheme includes:

[0034] Constructing a collection of extreme scenarios encompassing drought, flooding, and pollution events;

[0035] Simulate and implement the optimal water diversion plan under a set of extreme scenarios and record performance indicator data;

[0036] Based on the simulation results, a visual assessment report is generated that includes vulnerable node identification and risk transmission paths.

[0037] Optionally, generating an emergency dispatch plan under extreme scenarios based on the robustness evaluation results includes:

[0038] Determine the key control points for emergency dispatch based on the key vulnerable nodes and risk transmission paths identified by robustness assessment;

[0039] Construct a hierarchical response strategy system based on the event type characteristics in the extreme scenario set, which includes dispatch plans corresponding to events of different severity levels;

[0040] Generate water source dispatch sequence in emergency state based on water source supply capacity and water receiving area priority;

[0041] Formulate a pump station parameter adjustment plan that matches the water source scheduling sequence based on the hydraulic characteristics of the pipeline network and the operating constraints of the pump station;

[0042] Integrate key control points, hierarchical response strategy system, water source scheduling sequence and pump station parameter adjustment plan to generate emergency scheduling plans for extreme scenarios.

[0043] Optionally, after generating an emergency dispatch plan for an extreme scenario based on the evaluation result of the robustness evaluation, the method further includes:

[0044] Compare and analyze the generated emergency dispatch plan with historical dispatch cases to extract the emergency dispatch plan feature vector;

[0045] Filter out reference cases based on the plan feature vector and historical scheduling case library;

[0046] Extract successful scheduling experience from reference cases and optimize emergency scheduling plans based on the current project operation status.

[0047] In order to solve the above problems, the present invention also provides an intelligent scheduling system for cross-regional water diversion projects, the system comprising:

[0048] A digital twin system building module is used to build a digital twin system that integrates a meteorological evolution prediction model, a watershed hydrological response model, and a water demand prediction benchmark model based on a geographic information system, hydrological monitoring data, and spatial topology. The digital twin system is connected to meteorological satellite data, hydrological sensor data, and water demand data in real time.

[0049] A water volume prediction module, configured to predict the water demand of each water receiving area and the adjustable water volume of the water source in the future period based on the input data of the digital twin system;

[0050] A multi-objective optimization model building module is used to establish a multi-objective optimization model based on the prediction results, with water supply benefits, ecological impacts and energy consumption costs as optimization objectives;

[0051] A water diversion scheme generation module, configured to generate an optimal water diversion scheme for a cross-regional water diversion project based on the multi-objective optimization model and the Markov decision process;

[0052] The emergency dispatch plan generation module is used to perform robustness evaluation on the optimal water diversion plan and generate emergency dispatch plans under extreme scenarios based on the evaluation results of the robustness evaluation.

[0053] The present invention builds a digital twin system, integrates meteorological satellites, hydrological sensors and water demand data, combines spatiotemporal convolutional neural networks (to extract spatiotemporal characteristics of precipitation) with gated recurrent units (to capture temporal changes in runoff), and achieves high-precision joint prediction of water demand and adjustable water volume, solving the problems of data fragmentation and large prediction deviation in traditional methods; establishes an optimization model with water supply efficiency, ecological base flow compliance rate and energy consumption cost as goals, and transforms them through the Markov decision process, designs multi-agent collaboration rules based on spatial topological structure, and uses the policy gradient algorithm to generate the optimal scheduling instructions for each node, achieving the coordinated optimization of multi-objective balance and local-global decision-making, significantly improving resource allocation efficiency; by constructing a set of extreme scenarios (such as droughts and pollution events), simulating and evaluating the vulnerable nodes and risk transmission paths of water diversion plans, generating hierarchical response strategies and water source scheduling sequences, and optimizing plans based on historical case libraries, forming a "prediction-assessment-response" closed loop, enabling the system to quickly adjust pump station parameters and water source allocation in emergency situations, reducing the impact of extreme events. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic diagram of a flow chart of an intelligent scheduling method for a cross-regional water diversion project provided by one embodiment of the present invention;

[0055] Figure 2 This is a functional module diagram of an intelligent scheduling system for a cross-regional water diversion project provided by one embodiment of the present invention;

[0056] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0057] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] The embodiment of the present application provides a method for intelligent scheduling of cross-regional water diversion projects. The execution subject of the method for intelligent scheduling of cross-regional water diversion projects includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for intelligent scheduling of cross-regional water diversion projects can be executed by software or hardware installed on a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0059] Reference Figure 1 FIG. 1 is a flow chart of a method for intelligently scheduling a cross-regional water diversion project according to an embodiment of the present invention. In this embodiment, the method for intelligently scheduling a cross-regional water diversion project includes:

[0060] S1. Based on geographic information systems, hydrological monitoring data, and spatial topology, a digital twin system integrating meteorological evolution prediction models, watershed hydrological response models, and water demand prediction benchmark models is established. The digital twin system has real-time access to meteorological satellite data, hydrological sensor data, and water demand data.

[0061] In an embodiment of the present invention, a geographic information system (GIS) can be used in cross-regional water diversion projects to present the geographical locations and spatial relationships of water sources, pumping stations, water pipelines, reservoirs, and receiving areas. Hydrological monitoring data refers to data on the flow, water level, water quality, and other aspects of water bodies obtained through various hydrological monitoring equipment, which can reflect the real-time status of water resources. The spatial topological structure depicts the connection relationships and logical layout between the various components of the water diversion project (such as water sources, pumping stations, water pipelines, etc.).

[0062] In detail, the digital twin system is a system that digitally maps physical entities (i.e., cross-regional water diversion projects) and simulates their operating status in virtual space with the help of real-time access data; meteorological satellite data is data on meteorological conditions collected by meteorological satellites, covering information such as temperature, precipitation, and wind speed; hydrological sensor data is real-time data about water bodies obtained through hydrological sensors, such as water level, flow, water quality, etc.; water demand data reflects relevant data on water demand in each receiving area, such as water consumption in different time periods and different regions.

[0063] Specifically, real-time access means that data delay does not exceed 30 minutes. For example, meteorological satellite data is updated every 15 minutes via L-band broadcast, and hydrological sensor data is reported every 10 minutes via the NB-IoT network.

[0064] For example, in the South-to-North Water Diversion Project, the use of a geographic information system (GIS) allows for a clear visualization of the locations of various water sources, pumping stations, water transmission lines, and receiving areas. Hydrological monitoring data provides real-time insights into water levels and flows at each node. This information is integrated to construct the spatial topology of the project, enabling the development of a digital twin system. This system is then integrated with meteorological satellite data to understand the impact of weather changes on water diversion, hydrological sensor data to provide real-time water conditions, and water demand data to understand water usage in various regions.

[0065] In an embodiment of the present invention, the digital twin system integrating a meteorological evolution prediction model, a watershed hydrological response model, and a water demand prediction benchmark model is established based on a geographic information system, hydrological monitoring data, and a spatial topological structure, including:

[0066] Compare and analyze real-time meteorological satellite data with historical meteorological patterns to establish a meteorological evolution prediction model;

[0067] Construct a watershed hydrological response model based on the spatiotemporal distribution characteristics of hydrological sensor data;

[0068] Based on the periodic characteristics of water demand data and regional water use patterns, a water demand forecasting benchmark model is established;

[0069] The meteorological evolution prediction model, the watershed hydrological response model and the water demand prediction benchmark model are integrated to obtain a digital twin system.

[0070] In an embodiment of the present invention, the meteorological satellite data acquired in real time is compared and analyzed with the historical meteorological patterns. Establishing a meteorological evolution prediction model means collecting a large amount of historical meteorological data, applying data analysis and machine learning algorithms, and summarizing the historical meteorological patterns, wherein the historical meteorological patterns are meteorological change laws and patterns summarized based on meteorological data from past years; then, the meteorological satellite data is acquired in real time, and compared with the historical meteorological patterns to find similarities and differences. These comparison results are combined with meteorological principles to establish a meteorological evolution prediction model, wherein the meteorological evolution prediction model is a mathematical model for predicting future meteorological change trends.

[0071] For example, by analyzing precipitation data from the past decade, we can summarize the seasonal and cyclical changes in precipitation and form a historical meteorological pattern. This can then be compared with real-time meteorological satellite data. If we find that the current meteorological conditions are similar to a certain period in history and that there are specific meteorological change trends in the future, we can establish a prediction model based on this.

[0072] Specifically, the meteorological evolution prediction model can predict in advance the impact of meteorological changes on water resources, such as how increases or decreases in precipitation affect water volume, and how temperature changes affect evaporation. The meteorological evolution prediction model can address extreme events brought about by climate change, predicting meteorological changes in advance and enabling water diversion projects to prepare for these events, thereby improving their adaptability and stability.

[0073] In detail, constructing a watershed hydrological response model based on the spatiotemporal distribution characteristics of hydrological sensor data refers to collecting hydrological sensor data from different locations within a certain period of time and analyzing the distribution characteristics of these data in time and space. The spatiotemporal distribution characteristics of hydrological sensor data refer to the changing patterns and distribution characteristics of hydrological sensor data in time and space, such as changes in water level and flow in different time periods and geographical locations. Using the hydrological model construction method, combined with the topography, landform, soil and other geographical information of the watershed, a watershed hydrological response model is constructed. The watershed hydrological response model is a mathematical model that describes the response of hydrological processes in the watershed to factors such as precipitation and evaporation.

[0074] For example, by analyzing the water level and flow data of multiple hydrological sensors in a certain river basin in different seasons, it was found that there was a certain response pattern after precipitation. Combined with the terrain characteristics of the river basin, a model that can reflect this response relationship was constructed.

[0075] In general, the watershed hydrological response model can accurately simulate the hydrological processes within the watershed and predict the changes in water level and flow at different locations and times. The watershed hydrological response model is constructed through real-time hydrological sensor data and can promptly reflect the hydrological dynamics within the watershed.

[0076] In detail, establishing a water demand forecasting benchmark model based on the cyclical characteristics of water demand data and regional water use patterns refers to collecting water demand data from each water receiving area and analyzing its cyclical characteristics and regional water use patterns. The cyclical characteristics of water demand data refer to the cyclical changes in water demand data over time, such as seasonal water use changes, differences in water use between weekdays and weekends, etc. Regional water use patterns refer to the water use characteristics and patterns in different regions due to factors such as economic development level, population density, and industrial structure. Using data analysis and statistical methods, combined with factors affecting water demand, such as population growth, economic development, and climate change, a water demand forecasting benchmark model is established. The water demand forecasting benchmark model is a mathematical model used to predict future water demand in each water receiving area.

[0077] For example, by analyzing a city's water use data over the past few years, it was found that water demand in summer was significantly higher than in winter, and that water use patterns in commercial and residential areas were different. Combined with the city's population growth forecast and economic development plan, a water demand forecast benchmark model was established.

[0078] In detail, the meteorological evolution prediction model adopts the ARIMA (3,1,2) time series model, inputs the precipitation intensity of meteorological satellite data, and outputs the predicted value; the basin hydrological response model is constructed based on the Saint-Venant equations.

[0079] Overall, the water demand forecasting benchmark model can predict water demand in each receiving area in advance, allowing dispatchers to plan water transfers appropriately and avoid water shortages or oversupply. By analyzing the cyclical characteristics of water demand and regional water usage patterns, and incorporating multiple influencing factors, the water demand forecasting benchmark model can more accurately predict water demand.

[0080] Furthermore, the input and output interfaces of the meteorological evolution prediction model, the watershed hydrological response model, and the water demand forecasting benchmark model were unified to enable intercommunication and data sharing. These models were then integrated into the digital twin system framework, enabling comprehensive simulation and prediction of the water diversion project's operational status through data interaction and collaborative computing.

[0081] Furthermore, the integration method of the digital twin system includes: using the RESTful API interface specification to establish a data interaction channel between the meteorological evolution prediction model, the watershed hydrological response model and the water demand prediction benchmark model, ensuring that the format of the input and output data of each model is unified to the JSON standard; through the timestamp synchronization mechanism, the real-time data of each model is time-aligned, and the time resolution is set to 1 hour; using the microservice architecture to deploy three sub-models, and realizing dynamic resource allocation through Kubernetes container orchestration, for example, the meteorological evolution prediction model occupies a computing resource weight of 40%, the watershed hydrological response model 30%, and the water demand prediction benchmark model 30%; setting a data verification layer to perform outlier detection (using the 3σ principle) and missing value filling (using linear interpolation) on the input data.

[0082] For example, the meteorological evolution prediction model outputs {timestamp: ISO8601, precipitation: mm / h, region code: Geohash7} in JSON format. The watershed hydrological response model uses ETL tools to convert the data format and uses a sliding time window (window size 6 hours, step size 1 hour) for time alignment.

[0083] S2. Predicting the water demand of each water receiving area and the adjustable water volume of the water source in the future period based on the input data of the digital twin system.

[0084] In an embodiment of the present invention, the input data of the digital twin system refers to the meteorological satellite data, hydrological sensor data and water demand data, etc., which are accessed by the digital twin system in real time in step S1. These data reflect the internal and external environmental conditions of the water diversion project; the future period is a time interval relative to the current time, and water demand and adjustable water volume need to be predicted, such as the next week, month, etc.; the water demand of the receiving area refers to the amount of water required by each receiving area in the future period to meet the water use requirements of life, production, ecology, etc.; the adjustable water volume of the water source refers to the amount of water that can be allocated by each water source to supply the water receiving area in the future period, which is affected by factors such as the water volume of the water source itself and the regulation capacity of the water conservancy facilities.

[0085] In an embodiment of the present invention, the predicting of the water demand of each water receiving area and the adjustable water volume of the water source in a future period based on the input data of the digital twin system includes:

[0086] A spatiotemporal convolutional neural network is used to extract the spatiotemporal distribution feature matrix of precipitation from meteorological satellite data.

[0087] The gated recurrent unit is used to capture the temporal variation characteristics of runoff in the hydrological sensor data;

[0088] The spatial and temporal distribution characteristic matrix of precipitation and the temporal variation characteristic sequence of runoff are integrated to generate a joint prediction result including the water demand prediction value and the adjustable water volume prediction value.

[0089] In an embodiment of the present invention, using a spatiotemporal convolutional neural network to extract a spatiotemporal distribution feature matrix of precipitation from meteorological satellite data refers to inputting meteorological satellite data into the spatiotemporal convolutional neural network. The spatiotemporal convolutional neural network performs a convolution operation on the data through a convolution layer to extract spatial features of different scales, and reduces the dimension of the features through a pooling layer to reduce the amount of computation. At the same time, a loop structure is used to process data changes in the time dimension, thereby extracting the spatiotemporal distribution features of precipitation. The extracted features are organized into a matrix form to obtain a spatiotemporal distribution feature matrix of precipitation. The spatiotemporal convolutional neural network is a deep learning neural network that combines spatial and temporal processing capabilities. It can extract and analyze features of data with spatiotemporal characteristics and performs well in processing data with spatiotemporal variation patterns, such as meteorological satellite data. The spatiotemporal distribution feature matrix of precipitation is a two-dimensional or multi-dimensional matrix used to represent the distribution characteristics of precipitation at different time and spatial locations, such as the amount of precipitation in different regions at different times and changes in precipitation intensity.

[0090] Furthermore, the network structure of the spatiotemporal convolutional neural network includes 4 layers of 3D convolutional layers (kernel size 3×3×3, step size 1×1×1), 2 layers of spatiotemporal attention modules, and the activation function uses LeakyReLU.

[0091] Specifically, the three-dimensional convolution kernel (width × height × time) simultaneously extracts spatial and temporal features, captures the propagation dynamics of precipitation within the region, and identifies the spatiotemporal diffusion pattern of precipitation from upstream to downstream in the water source area; the convolution window size is 3 pixels × 3 pixels × 3 time steps, which balances local feature extraction and computational efficiency, and can detect precipitation changes in an area of approximately 50 km * 50 km within 3 hours; the step size is 1 × 1 × 1 to retain high-resolution spatiotemporal information and ensure that the output matrix is aligned with the spatiotemporal dimensions of the input data; the two-layer spatiotemporal attention module enhances the feature weights of key areas / periods, highlighting the impact of the heavy rain center area and the key periods of the flood season; the LeakyReLU alleviates the gradient disappearance, retains negative value information, and handles zero values (no rain) and extreme values (heavy rain) in precipitation data.

[0092] Furthermore, the loss function is:

[0093]

[0094] Where L is the mean square error (MSE) loss value; N is the number of training samples; i is the sample index, and the i-th sample can represent the satellite data at 8:00 on July 1, 2020; is the predicted precipitation matrix of the i-th sample, predicting the precipitation in the reservoir catchment area in the next 24 hours (mm / h); is the true precipitation matrix of the i-th sample; ‖*‖ is the L2 norm.

[0095] For example, when processing meteorological satellite data for a certain area, the spatiotemporal convolutional neural network can identify precipitation patterns in different seasons and geographical locations, such as the distribution patterns of afternoon thunderstorms in certain mountainous areas in summer, and present these patterns in matrix form.

[0096] In detail, capturing the temporal change feature sequence of runoff in hydrological sensor data through a gated recurrent unit means inputting the hydrological sensor data into the gated recurrent unit in chronological order. The gated recurrent unit controls the transmission and forgetting of information by updating the gate and resetting the gate, so that it can learn the temporal dependency and change law of the runoff data. At each time step, the gated recurrent unit outputs a new hidden state based on the current input and the hidden state of the previous moment. These hidden states are arranged in sequence to obtain the temporal change feature sequence of runoff. Among them, the gated recurrent unit is a variant of the recurrent neural network. It solves the gradient vanishing or gradient exploding problems that are prone to occur in traditional recurrent neural networks when processing long sequence data by introducing a gating mechanism, and can better capture the temporal information in the data. The temporal change feature sequence of runoff is a sequence arranged in chronological order, which records the changing characteristics of runoff at different times, such as the increase and decrease trend of flow, water level fluctuation, etc.

[0097] For example, for the hydrological sensor data of a river, the gated recurrent unit can capture the river's runoff change trends in the rainy and dry seasons, as well as the flow fluctuation characteristics during the flood and dry seasons.

[0098] Specifically, runoff directly reflects the amount of water available at a water source, and accurately capturing its temporal variations is crucial for predicting this amount. The temporal variation characteristics of runoff, derived through gated recurrent units, reflect the dynamics of runoff and provide a reliable basis for subsequent predictions of available water.

[0099] In detail, fusing the precipitation spatiotemporal distribution feature matrix and the runoff time series variation feature sequence to generate a joint prediction result including water demand prediction values and adjustable water volume prediction values means fusing the precipitation spatiotemporal distribution feature matrix and the runoff time series variation feature sequence by weighted summation. During the fusion process, corresponding weights can be assigned according to the importance of different features. The fused features are input into a multi-layer perceptron. By learning the relationship between the features in the historical data and the water demand and adjustable water volume, the fused features are analyzed and predicted, and finally a joint prediction result including the water demand prediction values of each receiving area and the adjustable water volume prediction values of each water source is output.

[0100] For example, the spatiotemporal distribution feature matrix of precipitation and the time series change feature sequence of runoff are spliced according to certain weights and input into the multi-layer perceptron. The multi-layer perceptron calculates the input according to the parameters obtained through training and outputs the predicted values of water demand of each receiving area and adjustable water volume of each water source in the future period.

[0101] In general, precipitation and runoff are two important factors affecting water demand and adjustable water volume. Fusing their characteristics can comprehensively consider the impact of both and generate more accurate and comprehensive joint prediction results.

[0102] S3. Based on the prediction results, a multi-objective optimization model is established with water supply benefits, ecological impact and energy consumption costs as optimization targets.

[0103] In an embodiment of the present invention, the prediction result refers to the water demand of each water receiving area and the adjustable water volume of the water source in the future time period obtained based on the input data of the digital twin system in step S2; water supply benefit refers to the benefit generated by the water diversion project in meeting the water demand of the water receiving area, which can be measured by the water supply priority and satisfaction of the water receiving area; ecological impact refers to the impact of the water diversion project on the river ecosystem, which is mainly reflected by the degree of compliance of the river ecological base flow; energy consumption cost refers to the energy cost consumed by the operation of the pump station in the water diversion project, which is related to the operating efficiency of the pump station; the multi-objective optimization model comprehensively considers multiple interrelated and possibly conflicting goals, and seeks a mathematical model that can make these goals achieve the optimal or near-optimal state at the same time under certain constraints.

[0104] In general, the multi-objective optimization model comprehensively considers the three important goals of water supply efficiency, ecological impact and energy consumption cost. It can ensure the water demand of the receiving area while taking into account ecological protection and reducing energy consumption costs, thereby maximizing the overall benefits of the water diversion project, avoiding the problem of traditional methods easily falling into local optimality, and improving the dynamic adaptability of the water diversion project.

[0105] In an embodiment of the present invention, the multi-objective optimization model based on the prediction results with water supply benefits, ecological impact and energy consumption costs as optimization targets includes:

[0106] Construct a water supply benefit evaluation function, the output value of which reflects the quantitative index of water supply priority and satisfaction of the water receiving area;

[0107] Design an ecological impact assessment function that includes the river ecological base flow constraint, and the output value of the ecological impact assessment function represents the quantitative index of the degree to which the river ecological base flow meets the standard;

[0108] Establish an energy cost function that reflects the operating efficiency of the pumping station. The output value of the energy cost function reflects the quantitative index of the operating efficiency of the pumping station.

[0109] A multi-objective optimization model is constructed based on water supply benefit evaluation function, ecological impact evaluation function and energy consumption cost function.

[0110] Specifically, constructing a water supply benefit evaluation function involves the following steps: First, factors influencing the water supply priority of a receiving area, such as its population, economic development level, and whether it is a key industrial zone, are identified, and each factor is assigned a corresponding weight. Then, water supply satisfaction is calculated based on the water demand and actual water supply situation of the receiving area. This water supply priority and water supply satisfaction are comprehensively considered to construct a mathematical function to represent water supply benefits. For example, a weighted summation approach can be used, multiplying the water supply priority and water supply satisfaction by their respective weights and then adding them together to obtain the output value of the water supply benefit evaluation function.

[0111] In detail, the water supply priority of the receiving area refers to the order of water supply determined based on factors such as the importance of the receiving area and the urgency of water use; the water supply satisfaction of the receiving area refers to the ratio of the actual water volume obtained by the receiving area to the required water volume, reflecting the degree to which water demand is met.

[0112] For example, for a water diversion project involving multiple receiving areas, receiving area A is a densely populated urban area, while receiving area B is an agricultural irrigation area. When determining water supply priorities, receiving area A can be given a higher weight. By monitoring the water demand and actual water supply of each receiving area in real time, the water supply satisfaction is calculated, and the value of the water supply benefit evaluation function is derived.

[0113] Specifically, designing an ecological impact assessment function that incorporates a river's ecological baseflow constraint involves the following steps: First, determine the river's ecological baseflow value based on the river's ecological characteristics and relevant standards. Then, monitor the river's actual flow in real time and compare it with the ecological baseflow value. This function constructs a quantitative indicator of the river's ecological baseflow compliance with the standard. For example, the ratio of actual flow to ecological baseflow can be used as a quantitative indicator of compliance. A ratio greater than 1 indicates compliance; a ratio less than 1 indicates non-compliance.

[0114] In detail, the ecological base flow of a river is the minimum flow that needs to be maintained in the river to maintain the basic structure and function of the river ecosystem; the degree of compliance of the ecological base flow of a river is the ratio or difference between the actual river flow and the ecological base flow, reflecting the satisfaction of the ecological base flow of the river.

[0115] For example, if the ecological base flow of a river is 10 cubic meters per second, and the actual flow of the river is obtained in real time through hydrological monitoring equipment, if the actual flow is 12 cubic meters per second, the quantitative index of compliance is 1.2; if the actual flow is 8 cubic meters per second, the quantitative index of compliance is 0.8.

[0116] Specifically, establishing an energy cost function that reflects the operating efficiency of a pumping station involves the following steps: First, factors influencing the station's operating efficiency and energy cost are analyzed, such as the pumping station's flow rate, head, motor power, and operating time. Then, based on the pumping station's operating principles and relevant physical formulas, a mathematical relationship is established between energy cost and these factors. For example, by calculating the pumping station's power consumption at different flow rates and heads and incorporating cost factors such as electricity prices, an energy cost function can be derived. This function takes the pumping station's operating parameters as input and outputs a quantitative indicator of energy cost.

[0117] In detail, the operating efficiency of a pumping station refers to the efficiency of energy conversion and utilization in the process of water transportation by the pumping station, which is related to factors such as the equipment performance and operating parameters of the pumping station.

[0118] For example, if the motor power of a pumping station is P, the operating time is t, and the electricity price is c, the energy consumption cost can be expressed as C = P × t × c. By analyzing and modeling the energy consumption costs under different operating conditions, the energy consumption cost function is obtained.

[0119] Furthermore, constructing a multi-objective optimization model based on the water supply benefit evaluation function, the ecological impact evaluation function, and the energy consumption cost function involves using a weighted summation method to multiply the water supply benefit evaluation function by a weight, the ecological impact evaluation function by a weight, and the energy consumption cost function by a weight to obtain a comprehensive objective function. A genetic algorithm is used to optimize F while satisfying constraints such as the adjustable water volume of the water source and the operating capacity of the pumping station.

[0120] In detail, the entropy weight method is used to calculate the weights of each objective function, where the water supply benefit weight w1, ecological impact weight w2, and energy consumption cost weight w3 satisfy w1+w2+w3=1. The specific calculation formula is:

[0121]

[0122] Among them, w i is the weight of the i-th objective, which is used to balance the priorities of water supply, ecology, and energy consumption in the multi-objective optimization model; E i is the information entropy of the i-th target, reflecting the degree of data dispersion (the smaller the entropy, the greater the weight); E j is the information entropy of the jth target. The smaller the entropy, the greater the fluctuation of the target data and the more significant the impact on scheduling. n is the number of evaluation samples, which is usually the number of historical scheduling cases or the size of the extreme scenario set. If the data of the past five years are analyzed, n = 60 (monthly statistics); k is the sample index, which identifies each independent historical or simulated data point. k = 1 can represent the scheduling data of January 2023. ik It is the standardized value of the kth sample under the i-th goal, reflecting the relative importance of the goal in a specific sample; i is the objective function category index (i = 1, 2, 3 correspond to water supply benefit, ecological impact, and energy consumption cost, respectively).

[0123] Specifically, when implementing the entropy weight method, it is necessary to generate a sample set (sample number n≥30) based on historical scheduling data or simulation scenarios, and calculate the information entropy after standardizing each target data. For example, the standardized value p of the water supply benefit target 1k Take the proportion of water supply in that month to the total water supply of the sample, and the ecological impact target p 2k Take the ratio of the ecological base flow compliance rate of the month to the total compliance rate of the sample, and the energy consumption cost target p 3k Take the energy consumption ratio of the current month. The final weight w i It is obtained by reverse normalization of entropy values, ensuring that important targets (such as water supply benefits) receive higher weights when data fluctuate greatly.

[0124] Alternatively, the assignment rule of “water supply priority weight” in the water supply benefit function can be linearly distributed according to population density (people / km 2 ×0.2), for example, the water supply benefit weight of the urban area is 0.85; the specific value of the ecological base flow constraint (such as the minimum flow value) can be taken as 30% of the multi-year average flow, for example: the Yellow River A section is 120m 3 / s; the electricity price parameter of the energy consumption cost function can refer to the local time-of-use electricity price table, for example: 0.8 yuan / kWh during peak hours.

[0125] Furthermore, the solution steps of the multi-objective optimization model include: using the NSGA-II algorithm for 500 generations, with the population size set to 100, the crossover probability to 0.9, and the mutation probability to 0.1, and selecting the optimal solution from the Pareto solution set that satisfies the water supply benefit ≥ 80% and the ecological base flow compliance rate ≥ 90%.

[0126] S4. Generate an optimal water diversion plan for the cross-regional water diversion project based on the multi-objective optimization model and Markov decision process.

[0127] In an embodiment of the present invention, the multi-objective optimization model refers to the model constructed in step S3, which takes water supply efficiency, ecological impact and energy consumption cost as optimization objectives, and is used to comprehensively consider multiple objectives of the water diversion project; the optimal water diversion plan refers to a water diversion plan that can achieve the optimal or near-optimal goals such as water supply efficiency, ecological impact and energy consumption cost in the multi-objective optimization model under various constraints, and is specifically reflected in the optimal scheduling instruction set of each node.

[0128] In general, based on the multi-objective optimization model constructed in step S3, the multi-objective optimization model is transformed into a collaborative decision-making problem of the Markov decision process, and then the multi-agent collaboration rules are formulated according to the spatial topological structure of the cross-regional water diversion project. Finally, the policy gradient algorithm is used to solve and obtain the optimal scheduling instruction set for each node.

[0129] In an embodiment of the present invention, generating an optimal water diversion plan for a cross-regional water diversion project based on the multi-objective optimization model and the Markov decision process includes:

[0130] Convert the multi-objective optimization model into a collaborative decision-making problem of Markov decision process;

[0131] Formulate multi-agent collaboration rules based on the spatial topological structure of cross-regional water diversion projects;

[0132] The policy gradient algorithm is applied to solve the problem and output the optimal scheduling instruction set for each node in the cross-regional water diversion project.

[0133] In detail, the spatial topological structure of the cross-regional water diversion project describes the spatial position relationship and connection structure between each node (water source, pumping station, water pipeline network, reservoir, etc.) in the water diversion project; the multi-agent collaboration rules refer to the rules for collaboration and interaction between each node (agent) in the cross-regional water diversion project, which are used to coordinate the decision-making and actions of each node to achieve overall optimal scheduling.

[0134] In detail, the formulation of multi-agent collaboration rules based on the spatial topological structure of the cross-regional water diversion project includes the following steps: first, the spatial topological structure of the cross-regional water diversion project is analyzed to determine the upstream and downstream relationships, connection paths, etc. between each node; then, according to the objectives and constraints of the multi-objective optimization model, the collaboration rules between each node are formulated.

[0135] For example, regulations might stipulate that when water is sufficient, upstream nodes prioritize supplying water to important downstream nodes; or that when a node fails, other related nodes should respond to emergencies. These rules must balance the local interests of each node with overall goals, ensuring multi-objective optimization during collaboration.

[0136] Specifically, multi-agent collaboration rules can regulate the collaborative behavior between nodes and improve their efficiency. Through collaboration, nodes can share information and cooperate with each other to jointly respond to various situations in the water diversion project, thereby achieving overall optimal scheduling and improving the operational efficiency and reliability of the water diversion project.

[0137] Furthermore, the multi-agent collaboration rules work in conjunction with the Markov decision process framework. In a Markov decision process, each node makes decisions based on its own state and feasible scheduling actions. The multi-agent collaboration rules specify how nodes communicate and collaborate with each other, ensuring that their decisions are coordinated and achieve multi-objective optimization.

[0138] In detail, the application of the policy gradient algorithm to solve and output the optimal scheduling instruction set for each node in the cross-regional water diversion project includes the following steps: the Markov decision process framework and multi-agent collaboration rules generated previously are used as input, and the policy gradient algorithm is applied to solve. Among them, the policy gradient algorithm continuously tries different decision strategies and calculates the benefits under each strategy according to the state transition probability matrix and the objective function of the multi-objective optimization model; then, the strategy is adjusted according to the benefits, and the strategy that makes the objective function optimal is gradually found; finally, according to the optimal strategy, the scheduling actions that each node should take under different states are determined to form the optimal scheduling instruction set.

[0139] For example, the Deep Deterministic Policy Gradient (DDPG) algorithm is used, which combines deep neural networks and policy gradient methods. It can handle problems in continuous action spaces and is suitable for decision-making of continuous actions such as flow regulation of pump stations in water diversion projects.

[0140] Specifically, the policy gradient algorithm is an optimization algorithm used to solve Markov decision processes. It optimizes the objective function (in this case, the comprehensive objective of the multi-objective optimization model) by continuously adjusting the decision-making strategy. The optimal scheduling instruction set is a set of optimal scheduling actions that each node should take under the conditions that meet the multi-objective optimization model and multi-agent collaboration rules. These instructions can achieve optimal or near-optimal goals for water diversion projects, such as water supply benefits, ecological impact, and energy consumption costs.

[0141] In an embodiment of the present invention, converting the multi-objective optimization model into a collaborative decision-making problem of a Markov decision process includes:

[0142] Define a multidimensional state vector containing the water level, flow rate and energy consumption of each node;

[0143] Construct a set of feasible scheduling actions, where the set of feasible scheduling actions satisfies the hydraulic constraints of the pipe network;

[0144] A state transition probability matrix considering the time delay effect is established to generate a Markov decision process framework.

[0145] In an embodiment of the present invention, defining a multidimensional state vector containing the water level, flow rate, and energy consumption of each node includes the following steps: collecting real-time water level, flow rate, and energy consumption data of each node in the cross-regional water diversion project (such as water sources, pumping stations, reservoirs, etc.), arranging these data in a certain order, and forming a multidimensional vector. For example, assuming there are n nodes, each node has three state variables: water level h, flow rate q, and energy consumption e. Then the multidimensional state vector can be expressed as s = [h1, q1, e1, h2, q2, e2, ..., h n ,q n ,e n ].

[0146] Specifically, a multidimensional state vector is used to describe the state information of each node in a cross-regional water diversion project at a certain moment, including variables in multiple dimensions such as water level, flow rate, and energy consumption. The multidimensional state vector is an important component of the Markov decision process framework. As a representation of the state, it is used to subsequently construct a set of feasible scheduling actions and a state transition probability matrix.

[0147] Specifically, the set of feasible scheduling actions refers to the set of scheduling actions that can be taken by each node in the water diversion project under the current state and that meet the hydraulic constraints of the pipeline network. For example, a pump station can adjust flow rate, open or close valves, etc.

[0148] In detail, the hydraulic constraints of the pipeline network refer to the restrictions imposed by the physical characteristics and operation requirements of the pipeline network in the water diversion project, such as flow conservation and pressure limitation.

[0149] Specifically, constructing a set of feasible scheduling actions involves determining the possible scheduling actions for each node based on the network structure and hydraulic characteristics of the water diversion project, such as adjusting the flow rate of a pump station or opening or closing a valve. These actions are then screened to remove those that do not meet the hydraulic constraints of the network.

[0150] For example, hydraulic calculation software (such as EPANET) can be used to simulate the water flow in the pipe network under different scheduling actions to determine whether constraints such as flow conservation and pressure balance are met. Ultimately, a set of feasible scheduling actions that meet the constraints is obtained.

[0151] In general, the set of feasible scheduling actions is a key element in the Markov decision process. Together with the multidimensional state vector, it determines the probability of state transitions. When constructing the subsequent state transition probability matrix, the probability of transitioning from one state to another after taking a feasible action needs to be considered.

[0152] Specifically, the state transition probability matrix describes the probability of transitioning from one state to another in a Markov decision process. In water diversion projects, it represents the probability of transitioning to the next state after taking a feasible scheduling action in the current state. The state transition probability matrix is the core component of the Markov decision process framework. Together with the multidimensional state vector and the set of feasible scheduling actions, it constitutes a complete Markov decision process.

[0153] Specifically, the time delay effect refers to the time delay between when a scheduling action is taken and when it actually affects the project status. For example, adjusting the flow rate at a pumping station takes time to affect the water level and flow rate at downstream nodes.

[0154] In detail, establishing a state transition probability matrix considering the time delay effect and thus generating a Markov decision process framework includes the following steps: first, based on historical data and the physical characteristics of the project, the possibility of state transition after a certain time delay after taking different feasible scheduling actions under different states is analyzed. For example, under a certain water level and flow state, the changes in the water level and flow of the downstream node after a period of time after the pump station increases a certain flow; then, these transfer possibilities are organized into a matrix form, namely the state transition probability matrix. When establishing the matrix, the time delay effect must be fully considered to ensure that the matrix can accurately reflect the actual state transition process.

[0155] For example, for state s i Take action a j Then transfer to state s k The probability P(s k ∣s i ,a j), all state transition probabilities are organized into a matrix form to obtain the state transition probability matrix, which is combined with the multidimensional state vector and the set of feasible scheduling actions to generate a Markov decision process framework.

[0156] Furthermore, a state transition probability matrix that considers time delay effects can more accurately describe the dynamic changes in the water diversion project, making the Markov decision process more realistic. This matrix can be used to predict the changing trends in the project state under different scheduling actions, providing a more reliable basis for decision-making.

[0157] S5. Conduct a robustness assessment on the optimal water diversion scheme and generate an emergency dispatch plan under extreme scenarios based on the robustness assessment results.

[0158] In an embodiment of the present invention, robustness assessment is an evaluation of the ability of the optimal water diversion scheme to maintain normal operation and achieve expected goals when facing extreme scenarios or uncertain factors such as drought, floods, pollution, etc.; extreme scenarios refer to unconventional events that may have a serious impact on cross-regional water diversion projects, such as long-term droughts leading to water source depletion, sudden floods damaging facilities, and water pollution affecting water quality; emergency dispatch plans are formulated for extreme scenarios and are used to guide the reasonable dispatch of water diversion projects in emergency situations and ensure water supply safety and ecological safety.

[0159] In general, first, a set of extreme scenarios is constructed to simulate the implementation of the optimal water diversion plan under these scenarios, performance indicator data is collected and a visual evaluation report is generated to complete the robustness evaluation; then, based on the evaluation results, the key control points of emergency dispatch are determined, a hierarchical response strategy system is constructed, a water source dispatch sequence and pump station parameter adjustment plan are generated, and finally, an emergency dispatch plan is integrated to form it; after the plan is generated, it is further optimized by comparing and analyzing it with historical cases.

[0160] In an embodiment of the present invention, the robustness evaluation of the optimal water diversion scheme includes:

[0161] Constructing a collection of extreme scenarios encompassing drought, flooding, and pollution events;

[0162] Simulate and implement the optimal water diversion plan under a set of extreme scenarios and record performance indicator data;

[0163] Based on the simulation results, a visual assessment report is generated that includes vulnerable node identification and risk transmission paths.

[0164] In detail, performance indicator data are used to measure the performance of various aspects of the water diversion plan during operation, such as water supply satisfaction rate, ecological base flow compliance rate, energy consumption cost, etc.

[0165] In detail, constructing a set of extreme scenarios including droughts, floods and pollution events includes the following steps: collecting relevant data on historical droughts, floods and pollution events, including information such as the time, location, duration, scope of impact, and severity of the events; using meteorological models (such as the WRF model), hydrological models (such as the HEC-HMS model) and water quality models (such as the EFDC model), combined with the actual situation of water diversion projects, to simulate extreme events of different intensities and types.

[0166] Specifically, a drought scenario is defined as a continuous period of no effective precipitation for more than 30 days and the reservoir water storage is below the dead water level; a flood scenario is defined as a 24-hour surface rainfall exceeding the 100-year return standard; and a pollution event is defined as a water quality index exceeding the surface water Class III standard by more than 10 times.

[0167] In detail, meteorological industry standards (such as "GB / T20481-2017 Meteorological Drought Level") can be cited to define drought thresholds, or a basis for statistical analysis of historical data can be provided.

[0168] For example, the WRF model is used to simulate drought scenarios with long periods of no precipitation, setting different drought durations and precipitation reduction ranges; the HEC-HMS model is used to simulate flood scenarios caused by heavy rains, considering different rainfall amounts and runoff speeds; and the EFDC model is used to simulate pollution events such as industrial wastewater leakage, setting different pollutant types and leakage amounts, and organizing these simulated extreme events into a set of extreme scenarios.

[0169] Specifically, simulating the optimal water diversion plan under a set of extreme scenarios and recording performance data includes the following steps: Using the constructed set of extreme scenarios as input, the digital twin system simulates the optimal water diversion plan. During the simulation, the system monitors and records performance data for various aspects of the water diversion project in real time. For example, sensors and data acquisition systems are used to obtain the actual water supply and water supply satisfaction rate for each receiving area; hydrological monitoring equipment is used to record river flow and ecological base flow compliance rates; and energy consumption costs are extracted from pump station operation data. This data is then organized and stored according to different extreme scenarios and time series.

[0170] In detail, the performance indicator data include: water supply guarantee rate, ecological base flow compliance rate and energy consumption cost deviation rate, among which, water supply guarantee rate = actual water supply / water demand × 100%, qualified threshold ≥ 85%; ecological base flow compliance rate = actual flow / ecological base flow threshold × 100, qualified threshold ≥ 80%; energy consumption cost deviation rate = |actual energy consumption - predicted energy consumption| / predicted energy consumption × 100%, allowable deviation ≤ 15%.

[0171] Furthermore, when any indicator exceeds a threshold, it is determined that the robustness of the solution does not meet the standard.

[0172] In detail, generating a visual assessment report containing vulnerable node identification and risk transmission paths based on simulation results includes the following steps: analyzing the recorded performance indicator data, using cluster analysis to identify nodes whose performance indicators change significantly under different extreme scenarios, and determining them as vulnerable nodes. By analyzing the connection relationship and data transmission between nodes, combined with the spatial topological structure of the water diversion project, the risk transmission path between each node is sorted out. Using visualization tools (such as ArcGIS, ECharts), the vulnerable nodes and risk transmission paths are displayed in the form of charts, maps, etc. to generate a visual assessment report. For example, the location of the vulnerable node is marked on the ArcGIS map, and the risk transmission path is represented by different colors and lines.

[0173] Furthermore, the visual assessment report can clearly and intuitively present the weak links and risk transfer of the optimal water diversion plan under extreme scenarios.

[0174] In an embodiment of the present invention, generating an emergency dispatch plan under extreme scenarios based on the robustness evaluation results includes:

[0175] Determine the key control points for emergency dispatch based on the key vulnerable nodes and risk transmission paths identified by robustness assessment;

[0176] Construct a hierarchical response strategy system based on the event type characteristics in the extreme scenario set, which includes dispatch plans corresponding to events of different severity levels;

[0177] Generate water source dispatch sequence in emergency state based on water source supply capacity and water receiving area priority;

[0178] Formulate a pump station parameter adjustment plan that matches the water source scheduling sequence based on the hydraulic characteristics of the pipeline network and the operating constraints of the pump station;

[0179] Integrate key control points, hierarchical response strategy system, water source scheduling sequence and pump station parameter adjustment plan to generate emergency scheduling plans for extreme scenarios.

[0180] Specifically, vulnerable nodes are key nodes that are easily affected in extreme scenarios and may cause problems in the operation of water diversion projects, such as water sources, pumping stations, and key water pipeline sections; risk transmission paths are the ways in which risks are transmitted and spread between various nodes and links of water diversion projects in extreme scenarios.

[0181] In detail, based on the key vulnerable nodes and risk transmission paths identified by the robustness assessment, the key control points for emergency dispatch are determined, which includes the following steps: According to the vulnerable nodes and risk transmission paths identified in the visual assessment report, combined with the function and importance of the water diversion project, the nodes and paths that have a greater impact on the overall operation of the project are screened out as key control points.

[0182] For example, if a pumping station is a key hub for water transportation to multiple water receiving areas and is shown to be vulnerable to flood scenarios in the robustness assessment, then the pumping station can be identified as a critical control point; for a water pipeline section connecting multiple important nodes, if there is a risk transmission path passing through it, it can also be identified as a critical control point.

[0183] Specifically, integrating critical control points, a hierarchical response strategy system, a water source scheduling sequence, and pump station parameter adjustment plans to generate emergency dispatch plans for extreme scenarios involves the following steps: Analyzing the characteristics of drought, flooding, and pollution events within the extreme scenario set, such as the duration and severity of droughts, the flow rate and inundation range of floods, and the concentration and diffusion rate of pollutants. Based on these characteristics, each event type is classified into different severity levels, such as mild, moderate, and severe. A corresponding dispatch plan is developed for each level.

[0184] For example, in the case of drought, the water allocation ratio can be adjusted in mild droughts to give priority to key water-receiving areas; in moderate droughts, backup water sources can be activated; and in severe droughts, water use by some non-critical users can be restricted.

[0185] In detail, the hierarchical response strategy system can take different response measures according to the severity of extreme events, making emergency dispatch more scientific and reasonable and improving resource utilization efficiency.

[0186] In detail, generating a water source dispatching sequence in an emergency state based on the water source supply capacity and the priority of the water receiving area includes the following steps: collecting the water supply capacity data of each water source, including the water reserve, adjustable water volume, water quality, etc. of the water source; and determining the water use priority of each water receiving area, which can be divided according to factors such as the population size, economic development level, and important industrial layout of the water receiving area.

[0187] Specifically, in an emergency, a water source dispatch sequence is developed using linear programming or dynamic programming algorithms based on the water source's supply capacity and the priority of the receiving areas. For example, high-quality, high-volume water sources are allocated to receiving areas with high water priority, ensuring water demand in key areas.

[0188] Specifically, developing a pump station parameter adjustment plan that matches the water source scheduling sequence, combining the hydraulic characteristics of the pipeline network and the operational constraints of the pump stations, involves the following steps: Using a pipeline network hydraulic model (such as EPANET) to analyze the hydraulic characteristics of the pipeline network under different operating conditions, including flow velocity, pressure distribution, and flow distribution, Furthermore, the operational constraints of the pump stations, such as their maximum flow rate, head range, and equipment start / stop restrictions, are considered. Based on the water source scheduling sequence, a nonlinear programming algorithm is used to calculate the optimal operating parameters of the pump stations, such as flow adjustment amplitude, head adjustment value, and start / stop times, to form a pump station parameter adjustment plan. For example, when the water source scheduling sequence requires an increase in water supply to a specific area, the required flow rate and head increase for the corresponding pump station is calculated based on the hydraulic characteristics of the pipeline network and the operational constraints of the pump stations.

[0189] Specifically, the identified critical control points, the established hierarchical response strategy system, the generated water source dispatch sequence, and the pump station parameter adjustment plan should be summarized and organized, and compiled into an emergency dispatch plan document in a logical order and format. The document should clearly define the specific content, applicable scenarios, and implementation process of each section to ensure the plan's operability. For example, it should detail how to implement the hierarchical response strategy for critical control points under different extreme scenarios and severity levels, as well as the corresponding water source dispatch and pump station parameter adjustment procedures.

[0190] In an embodiment of the present invention, after generating an emergency dispatch plan for an extreme scenario based on the evaluation result of the robustness evaluation, the method further includes:

[0191] Compare and analyze the generated emergency dispatch plan with historical dispatch cases to extract the emergency dispatch plan feature vector;

[0192] Filter out reference cases based on the plan feature vector and historical scheduling case library;

[0193] Extract successful scheduling experience from reference cases and optimize emergency scheduling plans based on the current project operation status.

[0194] Specifically, the generated emergency dispatch plan is compared and analyzed with historical dispatch cases. Extracting the emergency dispatch plan's feature vector involves the following steps: Establishing a historical dispatch case library collects dispatch cases from previous water diversion projects under different circumstances, including normal operation dispatch cases and emergency dispatch cases. The generated emergency dispatch plan is analyzed to extract key information and features, such as event type, response measures, and involved nodes and equipment. These features are quantified and encoded to form a plan feature vector. For example, the event type is represented by a numerical code, and the response measures are represented by different parameters. These codes and parameters are then combined into a vector. The historical dispatch cases are processed using the TF-IDF algorithm to extract the case feature vectors.

[0195] Specifically, selecting reference cases based on the emergency plan feature vector and the historical dispatch case library includes the following steps: The generated emergency dispatch plan feature vector is calculated to measure similarity with the feature vectors of each case in the historical dispatch case library, using a cosine similarity method. Based on the similarity, historical cases with a high degree of similarity to the emergency dispatch plan are selected as reference cases. For example, a similarity threshold can be set to select historical cases with similarities above the threshold.

[0196] In detail, the selected reference cases can provide practical experience and reference for the optimization of emergency dispatch plans, and help discover possible problems and improvement directions in the plans.

[0197] Specifically, extracting successful scheduling experiences from reference cases and optimizing the emergency scheduling plan based on the current project's operating status involves the following steps: Detailed analysis of the selected reference cases is conducted to extract successful scheduling experiences and effective measures. Simultaneously, operational data for the current water diversion project is collected, including water levels, flow rates, and equipment operating conditions at each node. Successful scheduling experiences are combined with the current project's operating status to adjust and optimize the emergency scheduling plan. For example, if a specific water allocation method was used in a reference case under similar extreme circumstances and achieved good results, and the current project's water source conditions are similar, this allocation method can be incorporated into the emergency scheduling plan.

[0198] In detail, by optimizing the emergency dispatch plan by combining successful dispatch experience and the current project operation status, the plan can be made more in line with actual needs and the feasibility and effectiveness of the plan in practical applications can be improved.

[0199] like Figure 2 , which is a functional module diagram of an intelligent scheduling system for a cross-regional water diversion project provided by one embodiment of the present invention.

[0200] The intelligent scheduling system 100 for inter-regional water diversion projects described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the intelligent scheduling system 100 for inter-regional water diversion projects can include a digital twin system construction module 101, a water volume prediction module 102, a multi-objective optimization model construction module 103, a water diversion plan generation module 104, and an emergency scheduling plan generation module 105. The modules described in the present invention, also referred to as units, refer to a series of computer program segments that can be executed by an electronic device processor and perform fixed functions, and are stored in the memory of the electronic device.

[0201] In this embodiment, the functions of each module / unit are as follows:

[0202] The digital twin system construction module 101 is used to establish a digital twin system that integrates a meteorological evolution prediction model, a watershed hydrological response model, and a water demand prediction benchmark model based on a geographic information system, hydrological monitoring data, and a spatial topological structure. The digital twin system is connected to meteorological satellite data, hydrological sensor data, and water demand data in real time;

[0203] The water volume prediction module 102 is used to predict the water demand of each water receiving area and the adjustable water volume of the water source in the future period based on the input data of the digital twin system;

[0204] The multi-objective optimization model building module 103 is used to establish a multi-objective optimization model with water supply benefits, ecological impacts and energy consumption costs as optimization objectives based on the prediction results;

[0205] The water diversion scheme generating module 104 is used to generate an optimal water diversion scheme for the inter-regional water diversion project based on the multi-objective optimization model and the Markov decision process;

[0206] The emergency dispatch plan generation module 105 is used to perform a robustness evaluation on the optimal water diversion plan and generate an emergency dispatch plan under extreme scenarios based on the evaluation results of the robustness evaluation.

[0207] In the several embodiments provided by the present invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the module division is merely a logical function division, and other division methods may be used in actual implementation.

[0208] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of the modules may be selected to achieve the purpose of the solution of this embodiment according to actual needs.

[0209] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional modules.

[0210] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0211] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0212] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent scheduling of cross-regional water diversion projects, characterized in that: The method comprises: Based on geographic information systems, hydrological monitoring data, and spatial topology, a digital twin system integrating meteorological evolution prediction models, watershed hydrological response models, and water demand forecasting benchmark models is established. The digital twin system is connected to meteorological satellite data, hydrological sensor data, and water demand data in real time. Predicting the water demand of each water receiving area and the adjustable water volume of the water source in the future period based on the input data of the digital twin system; Based on the prediction results, a multi-objective optimization model is established with water supply benefits, ecological impact and energy consumption costs as optimization targets; generating an optimal water diversion scheme for a cross-regional water diversion project based on the multi-objective optimization model and the Markov decision process; The robustness evaluation of the optimal water diversion scheme is carried out, and based on the evaluation results of the robustness evaluation, emergency dispatch plans under extreme scenarios are generated.

2. The intelligent scheduling method for a cross-regional water diversion project according to claim 1, characterized in that: The digital twin system based on geographic information system, hydrological monitoring data and spatial topology structure is established to integrate meteorological evolution prediction model, watershed hydrological response model and water demand prediction benchmark model, including: Compare and analyze real-time meteorological satellite data with historical meteorological patterns to establish a meteorological evolution prediction model; Construct a watershed hydrological response model based on the spatiotemporal distribution characteristics of hydrological sensor data; Based on the periodic characteristics of water demand data and regional water use patterns, a water demand forecasting benchmark model is established; The meteorological evolution prediction model, the watershed hydrological response model and the water demand prediction benchmark model are integrated to obtain a digital twin system.

3. The intelligent scheduling method for a cross-regional water diversion project according to claim 1, characterized in that: The prediction of the water demand of each water receiving area and the adjustable water volume of the water source in the future period based on the input data of the digital twin system includes: A spatiotemporal convolutional neural network is used to extract the spatiotemporal distribution feature matrix of precipitation from meteorological satellite data. The gated recurrent unit is used to capture the temporal variation characteristics of runoff in the hydrological sensor data; The spatial and temporal distribution characteristic matrix of precipitation and the temporal variation characteristic sequence of runoff are integrated to generate a joint prediction result including the water demand prediction value and the adjustable water volume prediction value.

4. The intelligent scheduling method for a cross-regional water diversion project according to claim 1, characterized in that: The multi-objective optimization model established based on the prediction results with water supply benefits, ecological impact and energy consumption costs as optimization objectives includes: Construct a water supply benefit evaluation function, the output value of which reflects the quantitative index of water supply priority and satisfaction of the water receiving area; Design an ecological impact assessment function that includes the river ecological base flow constraint, and the output value of the ecological impact assessment function represents the quantitative index of the degree to which the river ecological base flow meets the standard; Establish an energy cost function that reflects the operating efficiency of the pumping station. The output value of the energy cost function reflects the quantitative index of the operating efficiency of the pumping station. A multi-objective optimization model is constructed based on water supply benefit evaluation function, ecological impact evaluation function and energy consumption cost function.

5. The intelligent scheduling method for a cross-regional water diversion project according to claim 1, characterized in that: The generating of the optimal water diversion scheme for the inter-regional water diversion project based on the multi-objective optimization model and the Markov decision process includes: Convert the multi-objective optimization model into a collaborative decision-making problem of Markov decision process; Formulate multi-agent collaboration rules based on the spatial topological structure of cross-regional water diversion projects; The policy gradient algorithm is applied to solve the problem and output the optimal scheduling instruction set for each node in the cross-regional water diversion project.

6. The intelligent scheduling method for a cross-regional water diversion project according to claim 5, characterized in that: The collaborative decision-making problem of converting the multi-objective optimization model into a Markov decision process includes: Define a multidimensional state vector containing the water level, flow rate and energy consumption of each node; Construct a set of feasible scheduling actions, where the set of feasible scheduling actions satisfies the hydraulic constraints of the pipe network; A state transition probability matrix considering the time delay effect is established to generate a Markov decision process framework.

7. The intelligent scheduling method for a cross-regional water diversion project according to claim 1, characterized in that: The robustness evaluation of the optimal water diversion scheme includes: Constructing a collection of extreme scenarios encompassing drought, flooding, and pollution events; Simulate and implement the optimal water diversion plan under a set of extreme scenarios and record performance indicator data; Based on the simulation results, a visual assessment report is generated that includes vulnerable node identification and risk transmission paths.

8. The intelligent scheduling method for a cross-regional water diversion project according to claim 7, characterized in that: The evaluation results based on the robustness evaluation generate emergency dispatch plans for extreme scenarios, including: Determine the key control points for emergency dispatch based on the key vulnerable nodes and risk transmission paths identified by robustness assessment; Construct a hierarchical response strategy system based on the event type characteristics in the extreme scenario set, which includes dispatch plans corresponding to events of different severity levels; Generate water source dispatch sequence in emergency state based on water source supply capacity and water receiving area priority; Formulate a pump station parameter adjustment plan that matches the water source scheduling sequence based on the hydraulic characteristics of the pipeline network and the operating constraints of the pump station; Integrate key control points, hierarchical response strategy system, water source scheduling sequence and pump station parameter adjustment plan to generate emergency scheduling plans for extreme scenarios.

9. The intelligent scheduling method for a cross-regional water diversion project according to claim 7, characterized in that: After generating an emergency dispatch plan under an extreme scenario based on the evaluation result of the robustness evaluation, the method further includes: Compare and analyze the generated emergency dispatch plan with historical dispatch cases to extract the emergency dispatch plan feature vector; Filter out reference cases based on the plan feature vector and historical scheduling case library; Extract successful scheduling experience from reference cases and optimize emergency scheduling plans based on the current project operation status.

10. An intelligent scheduling system for cross-regional water diversion projects, characterized in that: The system comprises: A digital twin system building module is used to build a digital twin system that integrates a meteorological evolution prediction model, a watershed hydrological response model, and a water demand prediction benchmark model based on a geographic information system, hydrological monitoring data, and spatial topology. The digital twin system is connected to meteorological satellite data, hydrological sensor data, and water demand data in real time. A water volume prediction module, configured to predict the water demand of each water receiving area and the adjustable water volume of the water source in the future period based on the input data of the digital twin system; A multi-objective optimization model building module is used to establish a multi-objective optimization model based on the prediction results, with water supply benefits, ecological impacts and energy consumption costs as optimization objectives; A water diversion scheme generation module, configured to generate an optimal water diversion scheme for a cross-regional water diversion project based on the multi-objective optimization model and the Markov decision process; The emergency dispatch plan generation module is used to perform robustness evaluation on the optimal water diversion plan and generate emergency dispatch plans under extreme scenarios based on the evaluation results of the robustness evaluation.

Images