Multi-time scale water dispatching method based on irrigation area

Abstract

The present invention discloses a multi-time scale water dispatching method based on an irrigation area, which relates to the technical field of intelligent water resource management, and includes the following steps: constructing a multi-time scale prediction model, realizing minute-level and hour-level water demand prediction, and completing the water demand trend prediction for the next few days to weeks; establishing a multi-objective optimization and constraint model, dynamically adjusting the target weight and reconstructing the model, and using a non-equilibrium dynamic constraint correction algorithm and fuzzy game to coordinately optimize the multi-objective optimization model; responding to emergencies through disturbance monitoring and response mechanisms, and adjusting water supply priorities; dynamically reconstructing constraints and optimization models based on real-time data, hierarchically regulating water supply priorities according to the actual conditions of the irrigation area, and designing robust dispatching execution strategies to respond to anomalies. The problems existing in traditional dispatching methods can be solved, and multi-level dispatching decisions can be realized, and optimization and adjustment can be made for short-term and long-term needs, so as to improve the intelligence, flexibility and robustness of the irrigation area water supply system.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent water resource management, and more specifically, to a multi-time scale water dispatching method based on irrigation areas. Background Art

[0002] With climate change, increasing water demand and limited water resources, water scheduling in irrigation areas faces increasingly complex challenges. Traditional water scheduling methods have problems such as delayed information transmission, low scheduling accuracy and inability to respond quickly to sudden changes. In order to achieve scientific management of water resources, it is necessary to rely on advanced technical means, such as sensor networks, big data analysis and optimization algorithms, to achieve accurate and efficient water scheduling in irrigation areas. In irrigation area water resource management, traditional water scheduling methods often lack accurate grasp and flexible response to water demand at different time scales. In short-term scheduling, it is difficult to adjust the water supply plan in time according to the real-time changing meteorological and hydrological conditions and sudden water demand, resulting in low water supply efficiency and waste of water resources. In terms of medium- and long-term scheduling, the long-term water use trend and seasonal changes are often not fully considered, which makes the water supply plan out of touch with actual demand, affecting the sustainable development of the irrigation area.

[0003] Deficiencies of existing technologies:

[0004] Existing methods have deficiencies in multi-source data fusion, multi-objective optimization, and closed-loop control of the scheduling process, and are unable to achieve efficient use of water resources and maximize comprehensive benefits. Traditional scheduling methods may rely on experience or simplified models, lack accurate prediction and real-time adjustment capabilities, and are prone to over-extraction of water sources or insufficient water supply. Traditional methods lack sufficient flexibility and are difficult to cope with complex and changing climate conditions and sudden water shortages or surpluses. In order to better cope with the complex and changing water resource demands of irrigation areas, a multi-time scale water scheduling method is proposed. This method combines short-term and long-term water scheduling, and achieves more flexible and accurate water resource management by comprehensively considering the demands of different time scales. Based on this, a new water scheduling method based on multi-objective dynamic optimization and adaptive regulation is proposed, which can realize multi-level scheduling decisions, optimize and adjust short-term and long-term demands, and improve the intelligence, flexibility and robustness of the irrigation area water supply system.

[0005] In view of the above problems, the present invention proposes a solution. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a multi-time-scale water scheduling method based on irrigation areas, and solves the problems raised in the above-mentioned background technology by providing a multi-time-scale water scheduling method based on irrigation areas.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The multi-time scale water scheduling method based on irrigation areas includes the following steps:

[0009] Build a multi-time scale prediction model, use a hybrid deep learning model guided by the attention mechanism to achieve minute-level and hour-level water demand forecasts, and use the XGBoost model based on trend analysis to complete the water demand trend forecast for the next few days to weeks;

[0010] Dynamically adjust constraints according to real-time data, establish multi-objective optimization and constraint models, introduce reinforcement learning self-evolution strategy, dynamically adjust target weights and reconstruct models, and use non-equilibrium dynamic constraint correction algorithm and fuzzy game to collaboratively optimize multi-objective optimization models;

[0011] Implement multi-time scale flexible regulation, divide the dispatching strategy into different time scales and dynamically adjust the dispatching strategy, respond to emergencies through disturbance monitoring and response mechanisms, and adjust water supply priorities;

[0012] Based on real-time data, constraints and optimization models are dynamically reconstructed, water supply priorities are adjusted in layers according to the actual situation of the irrigation area, and robust scheduling execution strategies are designed to deal with anomalies.

[0013] In a preferred embodiment, the multi-time scale prediction model includes a short-term prediction model and a medium- to long-term prediction model;

[0014] The short-term prediction model includes regularly collecting historical water demand data of the irrigation area, obtaining water level, water flow, and meteorological forecast data in real time, and cleaning and normalizing the data;

[0015] The processed data is divided into training set, validation set and test set. The training set is used to train the hybrid deep learning model (Attention - GRU + XGBoost) guided by the attention mechanism, and the model parameters are adjusted. The model performance is evaluated using the validation set until the model achieves good prediction accuracy on the validation set.

[0016] Input real-time data into the trained model to obtain minute-by-minute and hour-by-hour water demand forecasts;

[0017] The medium- and long-term forecasting model includes collecting, collating and analyzing historical water use, rainfall and soil moisture data from multiple years;

[0018] Input the sorted data into the XGBoost model based on trend analysis, set model parameters, perform training and optimization, and evaluate model performance through cross-validation;

[0019] Combined with future weather forecast data, the trained model is used to generate water demand trend forecasts for the next few days to weeks.

[0020] In a preferred embodiment, the specific process of establishing the multi-objective optimization and constraint model is as follows:

[0021] According to the irrigation area water resources management objectives, determine the optimization objective function of minimizing water supply and demand deviation, optimizing ecological water use ratio, and maximizing water quality compliance rate;

[0022] Set corresponding constraints based on the actual capacity of the reservoir, the total historical water demand of the basin, ecological and environmental requirements, and the range of changes in real-time monitoring data.

[0023] In a preferred embodiment, the specific process of introducing the reinforcement learning self-evolution strategy, dynamically adjusting the target weight and reconstructing the model is as follows:

[0024] Define the state space, action space and reward function, and initialize the reinforcement learning agent model;

[0025] Use historical data and real-time monitoring data to train the reinforcement learning agent model and continuously adjust the target weights to achieve the optimal scheduling effect in different scenarios;

[0026] Regularly collect real-time monitoring and historical data, and dynamically adjust the upper limit of reservoir capacity, ecological water demand ratio and total water demand constraints of the basin based on the non-equilibrium dynamic constraint correction algorithm to achieve rapid adaptive reconstruction of the optimization model.

[0027] In a preferred embodiment, the collaborative optimization process of the multi-objective optimization model using the non-equilibrium dynamic constraint correction algorithm and fuzzy game is as follows:

[0028] Regularly analyze the degree of conflict between objective functions and identify the objective pairs with more prominent conflicts;

[0029] According to the degree of conflict, a multi-objective collaborative optimization strategy based on fuzzy game is used to dynamically assign weights, find the Pareto optimal solution, and balance the conflicting objectives;

[0030] Combined with real-time data, rolling adjustments are made to optimization plans for different time periods to ensure the effectiveness of the scheduling plan.

[0031] In a preferred embodiment, the multi-time scale elastic regulation process is as follows:

[0032] Receive short-term and mid- to long-term water demand forecasts in real time;

[0033] Develop corresponding dispatch strategies according to different time scales, and give priority to meeting emergency water needs in the short term (0-6 hours) based on minute-level and hour-level forecast results;

[0034] In the medium term (6-24 hours), water supply plans are adjusted using a combination of short-term forecasts and real-time data;

[0035] In the long term (more than 24 hours), water resource storage and allocation planning is carried out based on medium- and long-term forecast results.

[0036] In a preferred embodiment, the process of adjusting the water supply priority by using the disturbance monitoring and response mechanism to cope with emergencies is as follows:

[0037] Use the disturbance monitoring module to continuously monitor real-time data and calculate the deviation between real-time data and predicted results;

[0038] When the deviation exceeds the threshold, the disturbance response mechanism is triggered and the algorithm based on disturbance sensitivity analysis is started;

[0039] The algorithm calculates the impact range of disturbance events on water scheduling, and dynamically generates adjustment plans based on the impact range of different types of disturbance events to adjust water supply priority and water supply volume.

[0040] In a preferred embodiment, the process of dynamically reconstructing constraints and optimizing models based on real-time data is as follows:

[0041] Comprehensively calculate the deviation between real-time data and prediction, and compare the deviation with the established threshold to determine whether dynamic constraint reconstruction is needed;

[0042] When the dynamic constraint reconstruction conditions are met, the reservoir capacity constraints are adjusted according to the future inflow forecast results, and the ecological water use target priority is increased according to the actual situation;

[0043] Through a dynamic optimization algorithm based on a feedback loop, the expected results set by the current scheduling plan in the scheduling execution record are compared with the real-time monitoring data to calculate the deviation;

[0044] According to the calculated deviation, the objective function weight of the multi-objective optimization model is adjusted, the multi-objective model is re-optimized, and the revised optimization strategy is applied in the next scheduling.

[0045] In a preferred embodiment, the water supply priority is regulated in layers according to the actual situation of the irrigation area, and a robust scheduling execution strategy is designed to deal with abnormal processes as follows:

[0046] According to the actual situation of the irrigation area, the irrigation area is divided into key areas, sub-key areas and general areas;

[0047] Based on real-time monitoring and prediction results, water supply priorities are adjusted dynamically, and water supply to key areas is prioritized when water resources are tight;

[0048] When an abnormality occurs in equipment or communication, the backup scheduling plan is activated, and the medium- and long-term prediction results of the prediction module are used to provide a redundant water supply plan to ensure uninterrupted scheduling.

[0049] The technical effects and advantages of the multi-time scale water dispatching method of the present invention based on irrigation areas are as follows:

[0050] 1. The present invention can carry out refined management according to the water resource demand in different time ranges through multi-time scale scheduling, so that the allocation of water resources is more scientific and reasonable, the utilization efficiency of water resources is improved, and the waste of water resources is reduced. The scheduling of different time scales takes into account factors such as seasonal changes and weather changes, so it can adjust the water resource allocation plan in time, flexibly respond to short-term fluctuations and long-term changes in water volume in the irrigation area, and ensure the normal irrigation of crops in the irrigation area. The multi-time scale scheduling method combines short-term and long-term demand forecasts, can more accurately predict future water demand, and carry out scheduling planning in advance, which improves the scientificity and predictability of water resource management. The multi-time scale scheduling method can effectively distinguish the water demand in different time periods, such as the difference between short-term daily water demand and long-term seasonal irrigation demand, so as to reasonably arrange water resource supply and avoid water surplus or shortage. Accurate water scheduling at multiple time scales is achieved, the timeliness and accuracy of water supply are improved, and the water demand at different time scales in the irrigation area is met. Through the multi-objective optimization model and its self-evolution mechanism, the water supply and demand, ecological water use and water quality compliance are effectively balanced, and the comprehensive benefits of water resource utilization are improved. 2. The present invention can reasonably allocate water resources, avoid waste, and reduce water resource consumption in the irrigation area by accurately dispatching water. Reasonable water dispatching ensures timely irrigation of crops, can improve the growth efficiency and yield of crops, and thus increase the benefits of agricultural production in the irrigation area. The long-term dispatching takes into account ecological protection and sustainable use of water resources, avoids excessive extraction of water resources, protects the ecological environment of the water source, and promotes the long-term sustainable management of water resources. Through fine short-term and long-term water dispatching, water resources can be effectively allocated, the drought resistance of the irrigation area in drought periods can be enhanced, and the impact of natural disasters on agricultural production can be reduced. With the development of technology, the multi-time scale water dispatching method can be combined with modern information technology to automate dispatching, improve the intelligence of the dispatching process, reduce errors caused by human factors, and improve the precision and efficiency of dispatching. The method can adjust the water resource dispatching plan in real time according to the changes in demand at different time scales, and improve the emergency response capability of the irrigation area in the face of emergencies. The flexible control mechanism enhances the dispatching system's ability to adapt to emergencies and real-time changes, ensuring the stability and reliability of water supply. The optimized execution and closed-loop linkage mechanism improves the dispatching system's response speed and robustness, ensuring the efficient operation of the entire dispatching process. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a structural schematic diagram of the multi-time scale water scheduling method based on irrigation areas of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Embodiment 1, Figure 1 The present invention provides a multi-time scale water dispatching method based on irrigation areas.

[0054] S10, build a multi-time scale prediction model, realize minute-level and hour-level water demand prediction through a hybrid deep learning model guided by the attention mechanism, and use the XGBoost model based on trend analysis to complete the water demand trend prediction for the next few days to weeks;

[0055] The multi-time scale prediction model includes a short-term prediction model and a medium- and long-term prediction model;

[0056] Short-term forecasting model:

[0057] Regularly collect historical water demand data of irrigation areas, obtain water level, water flow, and weather forecast data in real time, and clean and normalize the data;

[0058] Construct a hybrid deep learning model (Attention - GRU + XGBoost) guided by the attention mechanism, integrate historical water demand data, real-time water level, water flow, and meteorological forecast data, capture the local characteristics and long-term dependencies of time series data, realize minute-level and hour-level water demand forecasting, and provide a basis for real-time scheduling;

[0059] The processed data is divided into training set, validation set and test set according to a certain ratio. The training set is used to train the hybrid deep learning model (Attention - GRU + XGBoost) guided by the attention mechanism, and the model parameters are adjusted. The model performance is evaluated using the validation set until the model achieves good prediction accuracy on the validation set.

[0060] Input real-time data into the trained model to obtain minute-level and hour-level water demand forecast results for real-time scheduling decisions;

[0061] Medium and long-term forecast model:

[0062] Collect historical water use, rainfall, and soil moisture data from multiple years, organize and analyze them;

[0063] Input the sorted data into the XGBoost model based on trend analysis, set model parameters, perform training and optimization, and evaluate model performance through cross-validation;

[0064] Combined with regional future meteorological forecast data, the trained model is used to extract long-term change trends and seasonal fluctuations, generate water demand trend forecasts for the next few days to weeks, and assist in medium- and long-term water supply planning.

[0065] The construction of multi-time scale prediction model plays an important role in the multi-time scale water dispatching method based on irrigation area, which is mainly reflected in the following aspects:

[0066] Optimize water resource management: The multi-time scale prediction model can accurately predict the available amount of water resources according to different time scales (such as hours, days, months, seasons, etc.), providing accurate data support for water scheduling in irrigation areas. By predicting short-term, medium-term and long-term water resources, managers can make reasonable water allocations at each time scale to ensure that the water needs of various areas in the irrigation area are met, while avoiding water waste or over-extraction;

[0067] Improve irrigation efficiency: The multi-time scale prediction model can reasonably arrange irrigation plans based on the prediction of climate factors such as precipitation, temperature, evaporation, etc. in different time periods in the future. This can help optimize the timing and amount of irrigation, improve the efficiency of the irrigation system, avoid over-irrigation or under-irrigation, and improve the efficiency of water resource utilization in agricultural production;

[0068] Improve scheduling flexibility: With the complexity of weather changes, seasonal changes, and water source changes, water resource scheduling faces many uncertainties. Multi-time scale prediction models can provide multi-level foresight for scheduling decisions, making irrigation area water resource scheduling more flexible and able to be dynamically adjusted at different time scales. For example, during periods of low seasonal precipitation, measures such as reservoir water storage and water conservation can be taken in advance, while during periods of high short-term precipitation, irrigation plans can be adjusted in a timely manner to reduce unnecessary water consumption;

[0069] Reduce risks and respond to emergencies: By predicting future short-term and long-term climate change, potential risks in water resource scheduling can be effectively identified. For example, if extreme weather events such as droughts and floods are predicted to occur in the future, managers can take countermeasures in advance, such as adjusting water source scheduling and increasing water resource reserves, thereby reducing the impact of extreme weather on water scheduling in irrigation areas;

[0070] Enhance system sustainability: Multi-time scale prediction models help achieve sustainability in irrigation water scheduling. By combining long-term water resource trends and short-term changes, water allocation can be optimized to not only meet current water needs, but also ensure the rational use of water resources and avoid resource depletion caused by excessive scheduling in the short term. This multi-level scheduling method helps maintain the long-term stability of ecosystems and agricultural production;

[0071] Support for intelligent decision-making and optimization: Based on multi-time scale prediction models, we can combine artificial intelligence, big data analysis and other technical means to provide more intelligent decision-making support. For example, we can use machine learning algorithms to analyze historical data and meteorological data of water resources, train multi-time scale prediction models, and provide real-time water scheduling suggestions and optimize water allocation.

[0072] S20, dynamically adjusts constraints according to real-time data, establishes a multi-objective optimization model, introduces a reinforcement learning self-evolution strategy, dynamically adjusts the objective weights and reconstructs the model, and uses a non-equilibrium dynamic constraint correction algorithm and fuzzy game to collaboratively optimize the multi-objective model;

[0073] The specific process of dynamically adjusting the constraints according to real-time data and establishing a multi-objective optimization model is as follows:

[0074] According to the irrigation area water resources management objectives, determine the objective function of minimizing water supply and demand deviation, optimizing ecological water use ratio, and maximizing water quality compliance rate;

[0075] Set corresponding constraints based on the actual capacity of the reservoir, the total historical water demand of the basin, ecological and environmental requirements, and the range of changes in real-time monitoring data;

[0076] The specific calculation formula for establishing a multi-objective optimization model is as follows:

[0077] Among them, the optimization objective function calculation formula is as follows: ;

[0078] The constraint formula is as follows: ;

[0079] In the formula, The deviation between water supply and demand is is the water supply and demand deviation weight factor, is the actual water demand, is the short-term forecast of water demand. is the long-term forecast of water demand; is the ecological water use ratio, is the ecological water use ratio weight factor, is the ecological water consumption, is the total water supply; is the water quality compliance rate, is the weight factor of water quality compliance rate, Is the water quality and quantity that meet the standards? is the total water volume; M is the optimization objective function; is the water demand of each region, is the water volume in the reservoir at time t, and are the minimum and maximum capacities, It is the minimum ecological water requirement.

[0080] The specific process of introducing the reinforcement learning self-evolution strategy, dynamically adjusting the target weight and reconstructing the model is as follows:

[0081] Define the state space, action space and reward function, and initialize the reinforcement learning agent model;

[0082] Use historical data and real-time monitoring data to train the reinforcement learning agent model and continuously adjust the target weights to achieve the optimal scheduling effect in different scenarios;

[0083] Regularly collect real-time monitoring and historical data, and dynamically adjust the upper limit of reservoir capacity, ecological water demand ratio and total water demand constraints of the basin according to the non-equilibrium dynamic constraint correction algorithm to achieve rapid adaptive reconstruction of the optimization model;

[0084] The collaborative optimization process of the multi-objective model using the non-equilibrium dynamic constraint correction algorithm and fuzzy game is as follows:

[0085] Regularly analyze the degree of conflict between objective functions and identify the objective pairs with more prominent conflicts;

[0086] According to the degree of conflict, a multi-objective collaborative optimization strategy based on fuzzy game is used to dynamically assign weights, find the Pareto optimal solution, and balance the conflicting objectives;

[0087] Combined with real-time monitoring data, rolling adjustments are made to the optimization plans for different time periods to ensure the effectiveness of the scheduling plan.

[0088] S30, implement multi-time scale flexible regulation, divide different time scales to dynamically adjust the dispatching strategy, respond to emergencies through disturbance monitoring and response mechanisms, and adjust water supply priorities;

[0089] The process of implementing multi-time scale elastic regulation is as follows:

[0090] Receive short-term and mid- to long-term water demand forecasts in real time;

[0091] Develop corresponding dispatch strategies according to different time scales, and give priority to meeting emergency water needs in the short term (0-6 hours) based on minute-level and hour-level forecast results;

[0092] In the medium term (6-24 hours), combine short-term forecasts with real-time monitoring data to analyze water use trends in the next 24 hours and dynamically adjust water supply plans based on water demand and water source conditions in different regions;

[0093] In the long term (more than 24 hours), water resource storage and allocation planning is carried out based on medium- and long-term forecast results.

[0094] The dispatching strategies between different time scales should maintain smooth information transmission, ensure the coordination between short-term dispatching and medium- and long-term dispatching, and make dynamic adjustments according to actual conditions to ensure that the dispatching plan can flexibly respond to changing demand and water source conditions at different time scales. Build an efficient dispatching system and use advanced data analysis and prediction technologies (such as AI, big data analysis, etc.) to provide a scientific basis for decision-making.

[0095] Through reasonable time scale scheduling strategies, efficient use and sustainable development of water resources can be achieved, and the stability of water supply and the ability to respond to emergencies can be ensured.

[0096] The process of adjusting water supply priority by using disturbance monitoring and response mechanism to deal with emergencies is as follows:

[0097] Use the disturbance monitoring module to continuously monitor real-time data and calculate the deviation between the monitored data and the predicted results;

[0098] When the deviation exceeds the threshold, the disturbance response mechanism is triggered and the algorithm based on disturbance sensitivity analysis is started;

[0099] The algorithm calculates the impact range of disturbance events on water scheduling, and dynamically generates adjustment plans based on the impact range of different types of disturbance events to adjust water supply priority and water supply volume.

[0100] The threshold value in the process of adjusting water supply priority is obtained as follows:

[0101] Collect real-time data of the irrigation area over a long period of time and the corresponding water dispatching conditions, including various data in different seasons and weather conditions, such as water level, water flow, and water demand, analyze the range and characteristics of these data under normal conditions and when disturbance events occur, and find out the general rules of the deviation between real-time data and predicted results under different conditions, which will serve as the basis for determining the threshold;

[0102] Combining the experience and expertise of experts in the field of water conservancy, experts give suggestions and guidance on the reasonable range of thresholds based on their long-term practical experience in irrigation district management and water dispatch, taking into account factors such as the characteristics of the irrigation district, water resources conditions, and water demand characteristics, and determine the thresholds based on comprehensive expert opinions;

[0103] By using the numerical simulation model of the irrigation district, different disturbance scenarios and deviation levels are set to simulate various possible situations, and the operating effect and stability of the water scheduling system under different deviations are observed. Through the evaluation of the simulation results, a threshold range is determined that can reasonably trigger the disturbance response mechanism, and will not be triggered frequently due to too small deviations, nor cause serious problems in the scheduling system due to too large deviations.

[0104] The calculation method of the threshold value in the process of adjusting water supply priority is as follows:

[0105] Calculate the standard deviation σ of the historical deviation between real-time data and forecast results. Usually, the threshold can be set to kσ, where k is a constant, usually between 1.5 and 3. For example, if k=2, it means that when the deviation exceeds 2 times the standard deviation of the historical deviation, the disturbance response mechanism is triggered.

[0106] S40 dynamically reconstructs constraints and optimization models based on real-time data, adjusts water supply priorities in layers according to the actual conditions of the irrigation area, and designs robust scheduling execution strategies to deal with anomalies.

[0107] The process of dynamically reconstructing constraints and optimizing models based on real-time data is as follows:

[0108] Monitor reservoir capacity and water demand data in real time, calculate the deviation between real-time monitoring data and prediction, and compare the deviation with the established threshold to determine whether dynamic constraint reconstruction is needed;

[0109] When the dynamic constraint reconstruction conditions are met, the reservoir capacity constraints are adjusted according to the future inflow forecast results, and the ecological water use target priority is increased according to the actual situation;

[0110] Through a dynamic optimization algorithm based on a feedback loop, the expected results set by the scheduling plan extracted from the scheduling execution record are compared with the real-time monitoring data to calculate the deviation;

[0111] According to the calculated deviation, the objective function weight of the multi-objective optimization model is adjusted, the multi-objective model is re-optimized, and the revised optimization strategy is applied in the next scheduling.

[0112] The thresholds are obtained in the process of real-time data dynamic reconstruction constraints and optimization models as follows:

[0113] Conduct sensitivity analysis on multi-objective optimization models and constraint models to observe the impact of changes in different parameters and variables on model results, determine which parameters or variables changes will have a key impact on the stability and effectiveness of the model, and determine the corresponding thresholds for these key factors;

[0114] Refer to the data deviations in the irrigation district's previous water scheduling process, when the existing scheduling schemes and models no longer work effectively or fail to achieve the expected scheduling effect, and use these data deviations that cause problems as an important reference for determining the threshold;

[0115] Consult the relevant standards, specifications and guidelines of the water conservancy industry to see if there are any regulations or suggestions on the allowable range of data deviation in water scheduling in similar irrigation areas. Based on the actual situation of the irrigation area, make appropriate adjustments and refinements to the relevant values ​​in these standards and specifications to determine the thresholds that are suitable for this irrigation area.

[0116] The calculation method of the threshold in the process of real-time data dynamic reconstruction constraint and optimization model is as follows:

[0117] Calculate the rate of change of the objective function in the multi-objective optimization model under different data deviations. For example, when the rate of change of objective functions such as water supply and demand deviation, ecological water use ratio or water quality compliance rate exceeds a certain value, it is considered necessary to perform dynamic constraint reconstruction. Assuming that the rate of change of the objective function M, ΔM / M, exceeds r% (r is a predetermined constant), dynamic constraint reconstruction is triggered, where r% is the corresponding threshold.

[0118] The water supply priority is adjusted in layers according to the actual situation of the irrigation area, and a robust scheduling execution strategy is designed to deal with abnormalities:

[0119] According to the actual situation of the irrigation area, the irrigation area is divided into key areas, sub-key areas and general areas;

[0120] Based on real-time monitoring and prediction results, water supply priorities are adjusted dynamically, and water supply to key areas is prioritized when water resources are tight;

[0121] When an abnormality occurs in equipment or communication, the backup scheduling plan is activated, and the medium- and long-term prediction results of the prediction module are used to provide a redundant water supply plan to ensure uninterrupted scheduling.

[0122] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0123] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0124] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0125] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0126] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0127] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-time scale water scheduling method based on irrigation areas, characterized in that: The following steps are involved: Build a multi-time scale prediction model, use a hybrid deep learning model guided by the attention mechanism to achieve minute-level and hour-level water demand forecasts, and use the XGBoost model based on trend analysis to complete the water demand trend forecast for the next few days to weeks; Dynamically adjust constraints according to real-time data, establish multi-objective optimization and constraint models, introduce reinforcement learning self-evolution strategy, dynamically adjust target weights and reconstruct models, and use non-equilibrium dynamic constraint correction algorithm and fuzzy game to collaboratively optimize multi-objective optimization models; Implement multi-time scale flexible regulation, divide the dispatching strategy into different time scales and dynamically adjust the dispatching strategy, respond to emergencies through disturbance monitoring and response mechanisms, and adjust water supply priorities; Dynamically reconstruct constraints and optimization models based on real-time data, adjust water supply priorities in layers according to the actual conditions of the irrigation area, and design robust scheduling execution strategies to deal with abnormalities; The multi-time scale prediction model includes a short-term prediction model and a medium- and long-term prediction model; The short-term prediction model includes regularly collecting historical water demand data of the irrigation area, obtaining water level, water flow, and meteorological forecast data in real time, and cleaning and normalizing the data; The processed data is divided into training set, validation set and test set. The training set is used to train the hybrid deep learning model Attention - GRU + XGBoost guided by the attention mechanism, and the model parameters are adjusted. The model performance is evaluated using the validation set until the model achieves good prediction accuracy on the validation set. Input real-time data into the trained model to obtain minute-by-minute and hour-by-hour water demand forecasts; The medium- and long-term forecasting model includes collecting, collating and analyzing historical water use, rainfall and soil moisture data from multiple years; Input the sorted data into the XGBoost model based on trend analysis, set model parameters, perform training and optimization, and evaluate model performance through cross-validation; Combined with future weather forecast data, the trained model is used to generate water demand trend forecasts for the next few days to weeks; The specific process of establishing the multi-objective optimization and constraint model is as follows: According to the irrigation area water resources management objectives, determine the optimization objective function of minimizing water supply and demand deviation, optimizing ecological water use ratio, and maximizing water quality compliance rate; Set corresponding constraints based on the actual capacity of the reservoir, the total historical water demand of the basin, ecological and environmental requirements, and the range of changes in real-time monitoring data.

2. The multi-time scale water dispatching method based on irrigation area according to claim 1 is characterized in that: The specific calculation formula for establishing the multi-objective optimization and constraint model is as follows: Among them, the optimization objective function calculation formula is as follows: ; The constraint formula is as follows: ; In the formula, The deviation between water supply and demand is is the water supply and demand deviation weight factor, is the actual water demand, is the short-term forecast of water demand. is the long-term forecast of water demand; is the ecological water use ratio, is the ecological water use ratio weight factor, is the ecological water consumption, is the total water supply; is the water quality compliance rate, is the weight factor of water quality compliance rate, Is the water quality and quantity that meet the standards? is the total water volume; M is the optimization objective function; is the water demand of each region, is the water volume in the reservoir at time t, and are the minimum and maximum capacities, It is the minimum ecological water requirement.

3. The multi-time scale water dispatching method based on irrigation area according to claim 2 is characterized in that: The specific process of introducing the reinforcement learning self-evolution strategy, dynamically adjusting the target weight and reconstructing the model is as follows: Define the state space, action space and reward function, and initialize the reinforcement learning agent model; Use historical data and real-time monitoring data to train the reinforcement learning agent model and continuously adjust the target weights to achieve the optimal scheduling effect in different scenarios; Regularly collect real-time monitoring and historical data, and dynamically adjust the upper limit of reservoir capacity, ecological water demand ratio and total water demand constraints of the basin based on the non-equilibrium dynamic constraint correction algorithm to achieve rapid adaptive reconstruction of the optimization model.

4. The multi-time scale water dispatching method based on irrigation area according to claim 3 is characterized in that: The collaborative optimization process of the multi-objective optimization model using the non-equilibrium dynamic constraint correction algorithm and fuzzy game is as follows: Regularly analyze the degree of conflict between objective functions and identify the objective pairs with more prominent conflicts; According to the degree of conflict, a multi-objective collaborative optimization strategy based on fuzzy game is used to dynamically assign weights, find the Pareto optimal solution, and balance the conflicting objectives; Combined with real-time data, rolling adjustments are made to optimization plans for different time periods to ensure the effectiveness of the scheduling plan.

5. The multi-time scale water dispatching method based on irrigation area according to claim 4 is characterized in that: The process of implementing multi-time scale elastic regulation is as follows: Receive short-term and mid- to long-term water demand forecasts in real time; Develop corresponding dispatch strategies based on different time scales, and prioritize emergency water needs based on minute-level and hour-level forecast results within the short term of 0 to 6 hours; Adjust water supply plans in the medium term (6-24 hours) by combining short-term forecasts with real-time data; When the water level exceeds 24 hours, water resource reserve and allocation planning is carried out based on the medium- and long-term forecast results.

6. The multi-time scale water dispatching method based on irrigation area according to claim 5 is characterized in that: The process of adjusting water supply priority by using disturbance monitoring and response mechanism to deal with emergencies is as follows: Use the disturbance monitoring module to continuously monitor real-time data and calculate the deviation between real-time data and predicted results; When the deviation exceeds the threshold, the disturbance response mechanism is triggered and the algorithm based on disturbance sensitivity analysis is started; The algorithm calculates the impact range of disturbance events on water scheduling, and dynamically generates adjustment plans based on the impact range of different types of disturbance events to adjust water supply priority and water supply volume.

7. The multi-time scale water dispatching method based on irrigation area according to claim 6 is characterized in that: The process of dynamically reconstructing constraints and optimizing models based on real-time data is as follows: Comprehensively calculate the deviation between real-time data and prediction, and compare the deviation with the established threshold to determine whether dynamic constraint reconstruction is needed; When the dynamic constraint reconstruction conditions are met, the reservoir capacity constraints are adjusted according to the future inflow forecast results, and the ecological water use target priority is increased according to the actual situation; Through a dynamic optimization algorithm based on a feedback loop, the expected results set by the current scheduling plan in the scheduling execution record are compared with the real-time monitoring data to calculate the deviation; According to the calculated deviation, the objective function weight of the multi-objective optimization model is adjusted, the multi-objective model is re-optimized, and the revised optimization strategy is applied in the next scheduling.

8. The multi-time scale water dispatching method based on irrigation area according to claim 7 is characterized in that: The process of hierarchically regulating water supply priorities according to the actual situation of the irrigation area and designing a robust scheduling execution strategy to deal with abnormalities is as follows: According to the actual situation of the irrigation area, the irrigation area is divided into key areas, sub-key areas and general areas; Based on real-time monitoring and prediction results, water supply priorities are adjusted dynamically, and water supply to key areas is prioritized when water resources are tight; When an abnormality occurs in equipment or communication, the backup scheduling plan is activated, and the medium- and long-term prediction results of the prediction module are used to provide a redundant water supply plan to ensure uninterrupted scheduling.

Images