CN118014326A

CN118014326A - Method and system for planning river basin water resource scheduling in water network system

Info

Publication number: CN118014326A
Application number: CN202410420993.1A
Authority: CN
Inventors: 郭旭宁; 刘为锋; 李云玲; 杜涛; 朱非林; 陈娟; 杜二虎; 徐斌; 李金明; 杨青素; 钟平安
Original assignee: China Renewable Energy Engineering Institute
Current assignee: China Renewable Energy Engineering Institute
Priority date: 2024-04-09
Filing date: 2024-04-09
Publication date: 2024-05-10
Anticipated expiration: 2044-04-09
Also published as: CN118014326B

Abstract

The invention discloses a method and a system for planning river basin water resource scheduling in a water network system, which are used for collecting data and constructing a quantitative index system of historical meteorological factors and water situation space-time distribution characteristics; calculating and predicting the first industrial water demand, the second industrial water demand, the third industrial water demand, the ecological water demand and the living water demand of the research area; calculating the drainage flow demand of all initial sections by RVMF method, constructing a generalized simulation model of the water resource system, and screening out important sections; and constructing a water resource optimization scheduling model, solving the water resource optimization scheduling model by adopting a plurality of multi-objective algorithms to obtain a non-inferior solution set, and optimizing the non-inferior solution set by adopting an MLAA-VIKOR model to obtain a river basin water resource scheduling plan. According to the invention, weather factor forecast is used for predicting water conditions, the three-generation water demand and the drainage flow are comprehensively considered, the management and scheduling efficiency of water resources is improved, the water resource supply is ensured, and the flood risk is reduced.

Description

Method and system for planning river basin water resource scheduling in water network system

Technical Field

The invention relates to a method for planning and planning water resource scheduling of a river basin in a water network system.

Background

The water network system is a comprehensive system integrating functions of optimizing configuration of water resources, flood control and disaster reduction of a river basin, protection of a water ecological system and the like on the basis of natural rivers and lakes, taking drainage engineering as a channel, regulation and storage engineering as nodes and intelligent regulation as means, is an effective measure for solving the problem of uneven spatial distribution of water resources, improving the water resource guarantee rate of a water receiving area, relieving the contradiction between supply and demand of water resources in a water-deficient area and realizing reasonable configuration of the water resources, and is an important way for promoting economic development and comprehensive development and utilization of the water resources in the water-deficient area.

Along with the development of social economy and the increasing importance of water resource management, the demands of an intelligent water resource scheduling system are increasing, at present, a few researches on a water resource scheduling planning method are related to hydrologic models, optimization algorithms and data acquisition technologies, however, the traditional water resource scheduling method is often limited by the problems of strong experience, high computational complexity, low efficiency and the like, the increasing water resource management demands are difficult to meet, and the defects of high computational complexity, poor instantaneity, poor adaptability and the like in the prior art still exist, so that an intelligent and automatic water resource scheduling system needs to be developed, and the water resource allocation can be automatically adjusted according to real-time data and demands, so that the water resource utilization efficiency is improved and the risk is reduced.

The invention provides a river basin water resource scheduling plan programming method and system in a water network system, which solve the problems existing at present, improve the effective management and scheduling of water resources and ensure the sustainable utilization of water resources and the protection of ecological environment.

Disclosure of Invention

The invention aims to provide a river basin water resource scheduling planning method in a water network system so as to solve the problems in the prior art. On the other hand, a river basin water resource scheduling plan making system in the water network system is provided.

According to one aspect of the application, there is provided a method for scheduling water resources in a river basin in a water network system, comprising the steps of:

step S1, collecting historical meteorological factor data and water regime data of a research area, and historical data of water demand in a Feng Ping model year; analyzing and constructing a quantitative index system aiming at the time-space distribution characteristics of the historical meteorological factors and the water conditions; a pre-built decision tree machine learning module is adopted to establish a mapping relation between historical meteorological factors and water conditions;

S2, predicting the first industrial water demand of a research area according to historical data of the base Yu Fengping model annual water demand, and verifying by adopting two-dimensional joint distribution; constructing a prediction method set at least comprising two prediction models, and cross predicting and verifying second industrial water demand, third industrial water demand, ecological water demand and living water demand of a research area; calculating the sum of water demand, ecological water demand and living water demand of the first industry, the second industry and the third industry to obtain the total water demand of the drainage basin of the research area;

s3, acquiring and taking all hydrologic stations, main and branch flow junctions, drainage basin outlets and administrative district junctions with water contradiction in a research area as initial sections, and calculating the drainage flow requirements of all the initial sections by adopting a RVMF method; calling the constructed water resource system generalized simulation model, and calculating and screening out important sections;

s4, inputting the water conditions of the research area, the total water demand of the drainage basin of the research area and the drainage flow demand of the important section into a pre-constructed water resource optimal scheduling model, solving the water resource optimal scheduling model by adopting at least two multi-objective algorithms to obtain a non-inferior solution set, screening an optimal solution from the non-inferior solution set by adopting an MLAA-VIKOR model, and taking the optimal solution as a river drainage basin water resource scheduling plan.

According to one aspect of the present application, the step S1 is further:

S11, determining a range of a research area, and collecting historical meteorological factor data, water regime data and Feng Ping years old water demand historical data of the research area, wherein the historical meteorological factors of the research area comprise rainfall, air temperature, wind speed, relative humidity and sunshine hours;

S12, extracting historical meteorological factors and water regime data of a research area, analyzing the space-time distribution characteristics of the historical meteorological factors and the water regime, and respectively constructing a quantitative index system of the space-time distribution characteristics of the historical meteorological factors and the water regime by adopting a factor analysis method;

S13, discretizing a quantitative index system sample of the time-space distribution characteristics of the historical meteorological factors and the water conditions to obtain a preset quantity of CNN-transducer training samples, and training a CNN-transducer model through the CNN-transducer training samples;

S14, respectively reducing weights of quantitative index systems of time-space distribution characteristics of historical meteorological factors and water conditions by using a trained CNN-transducer model to obtain characteristic indexes of the historical meteorological factors and the water conditions, wherein rainfall, air temperature, wind speed, relative humidity and sunshine hours in the historical meteorological factors respectively comprise a, b, c, d, e characteristic indexes, the water conditions comprise f characteristic indexes, and a, b, c, d, e, f is a natural number larger than 0;

And S15, aiming at the characteristic indexes of the historical meteorological factors and the water conditions after weight reduction, a decision tree machine learning module is adopted to establish a mapping relation from the historical meteorological factors to the water conditions.

According to an aspect of the present application, the step S13 is further:

Step S13a, historical meteorological factors and water regime data are called, a sample set is generated, and the sample set is recorded as: { x _ij |i=1, 2, …, m; j= 1,2, …, n }, where i and j are the sample number and index number, respectively, and m and n are the sample number and index number, respectively;

Step S13b, based on the sample set, determining positive ideal points { x _j ⁺ |j= 1,2, …, n } and negative ideal points { x _j ^- |j= 1,2, …, n };

Step S13c, randomly generating k random numbers obeying uniform distribution by adopting a random simulation method, performing discretization calculation on index values between positive ideal points and negative ideal points, and generating k training samples serving as CNN-transducer input items; k is a natural number less than n;

step S13d, sequentially calculating Euclidean distances between each training sample and the positive ideal point and the negative ideal point;

Step S13e, sequentially calculating the Euclidean distance between each training sample and the negative ideal point divided by the sum of the Euclidean distances between the training sample and the positive ideal point and the Euclidean distance between the training sample and the negative ideal point to obtain the fitness coefficient of the training sample, and taking the fitness coefficient as a CNN-transducer output item;

And step S13f, training the CNN-transducer and storing.

According to one aspect of the present application, the step S2 is further:

S21, calling collected historical data of Feng Ping model year water demand, predicting the first industry water demand of a research area through a Feng Ping model year method, and adopting two-dimensional joint distribution verification;

S22, constructing a prediction method set at least comprising two prediction models, calling the prediction models according to a preset rule to alternately predict and verify the second industrial water demand, the third industrial water demand, the ecological water demand and the living water demand of the research area; the prediction model comprises a trend extension module, a multiple regression module and a multiple machine learning module;

Step S23, calculating the sum of the first industrial water demand, the second industrial water demand, the third industrial water demand, the ecological water demand and the living water demand as the total water demand of the study area.

According to an aspect of the application, the step S21 is further:

step S21a, calling historical data of the water demand of the Fengping model year, and obtaining first industrial water demand of each model year to form a first industrial water demand sequence V1; calculating the first industrial water demand of each model year in a preset period through a preset water demand calculation model to form a first industrial water demand sequence V2;

Step S21b, constructing and verifying two-dimensional joint distribution of the first industrial water demand through the first industrial water demand sequence V1 and the first industrial water demand sequence V2;

Step S21c, calculating the first industrial water demand in the current year based on the two-dimensional joint distribution of the first industrial water demand.

According to an aspect of the application, the step S22 is further:

Step S22a, a training data set is constructed, and a prediction method set at least comprises two prediction models, wherein the prediction models comprise a trend epitaxy module, a multiple regression module and a multiple machine learning module;

Step S22b, dividing training data into K folds, wherein K is a natural number larger than 5;

Step S22b, in each round of training, sequentially taking one of the data as a test set and the K-1 data as a training set; according to a preset rule, predicting the second industrial water demand and the third industrial water demand of a research area by adopting a trend epitaxy module, a multiple regression module and a multiple machine learning module respectively, predicting the ecological water demand of the research area by adopting the trend epitaxy module and the multiple regression module respectively, and predicting the living water demand by adopting the multiple regression module and the multiple machine learning module respectively; calculating the performance index of each prediction model and the average performance index of all prediction rounds in each prediction round;

Step S22c, selecting an optimal method corresponding to the second industrial water demand, the third industrial water demand, the ecological water demand and the living water demand according to the average performance index;

step S22d, calculating the second industrial water demand, the third industrial water demand, the ecological water demand and the living water demand of the research area by selecting an optimal method one by one.

According to one aspect of the present application, the step S3 is further:

s31, taking all hydrological sites, main and branch flow junctions, drainage basin outlets and administrative district junctions with water contradiction in a research area as initial sections;

S32, calculating ecological flow in a river channel by adopting a RVMF method, and taking the maximum value of the ecological flow and the flow requirements of downstream life, industry, irrigation and shipping as the lower drainage flow requirement of each section;

and S33, constructing a generalized simulation model of the water resource system, sequentially taking the drainage flow requirements of all the initial sections as model input, and calculating and screening important sections.

According to an aspect of the present application, the step S33 is further:

S33a, constructing a generalized simulation model of a water resource system;

step S33b, setting the downstream flow demand of the section as an upstream and downstream section control condition, setting the water consumption in the flow area as an upstream and downstream section interval water taking condition, and setting the section as an unimportant section if the influence of the selected section on the water consumption of the flow area is smaller than a threshold value in a threshold value time;

And step S33c, repeating the operation on all the sections, screening out unimportant sections, and taking the rest sections as important sections.

According to one aspect of the present application, the step S4 is further:

S41, constructing a water resource optimization scheduling model, wherein an objective function is the water demand of a sub-basin and the drainage flow under an important section, and the constraint condition is engineering capacity;

S42, inputting the water conditions of the research area, the total water demand of the drainage basin of the research area and the drainage flow demand under the important section into a water resource optimization scheduling model, and solving the water resource optimization scheduling model by adopting a plurality of multi-objective algorithms to obtain a non-inferior solution set;

And S43, adopting an MLAA-VIKOR model to conduct optimization on the non-inferior solution set, and obtaining an optimal solution which is a river basin water resource scheduling plan.

According to an aspect of the present application, the step S42 is further:

step S42a, taking the water condition of the research area, the total water demand of the drainage basin of the research area and the drainage flow demand under the important section as inputs, and driving a water resource optimization scheduling model;

And step S42b, respectively solving the water resource optimization scheduling model by adopting MOGA, MOPSO, MOSA and MOACO multi-objective optimization algorithms to obtain a non-inferior solution set.

According to an aspect of the application, the step S43 is further:

Step S43a, respectively constructing decision criteria of non-inferior solutions obtained by MOGA, MOPSO, MOSA and MOACO multi-objective optimization algorithms, wherein the decision criteria comprise efficiency, stability and flexibility;

Step S43b, calculating weights of four decision criteria for each non-inferior solution set, and obtaining Si value and Ri value of each non-inferior solution, and calculating comprehensive utility value of each non-inferior solution based on the Si value and the Ri value;

Step S43c, sorting the solutions in each non-inferior solution set according to Si and Ri values to obtain VIKOR solution sorting of each target optimization algorithm;

Step S43d, layering and aggregating VIKOR solution sequencing results of different algorithms by using an MLAA method, establishing a multi-layer structure model from the bottom layer to the top layer, and integrating solution sequencing of different algorithms layer by layer until the top layer obtains global optimal sequencing; the solution at the first place of the global optimal sequencing is the optimal compromise solution obtained by a plurality of multi-objective optimization algorithms together, and the optimal solution is used as the optimal solution, namely the river basin water resource scheduling plan.

According to another aspect of the present application, there is provided a river basin water resource scheduling planning system in a water network system, comprising:

at least one processor; and

A memory communicatively coupled to at least one of the processors; wherein,

The memory stores instructions executable by the processor for execution by the processor to implement the method for scheduling water resources in a river basin in a water network system according to any one of the above-described technical schemes.

The method has the beneficial effects that the method for planning the water resource scheduling of the river basin in the water network system is adopted, the advanced hydrologic model and the optimization algorithm are utilized, the reasonable water resource scheduling plan is formulated in combination with the actual situation, the water resource utilization efficiency is improved, and the flood risk is reduced. The related art effects will be described in detail in the detailed description.

Drawings

Fig. 1 is a flow chart of the present invention.

Fig. 2 is a flowchart of step S1 of the present invention.

Fig. 3 is a flow chart of step S2 of the present invention.

Fig. 4 is a flowchart of step S3 of the present invention.

Fig. 5 is a flowchart of step S4 of the present invention.

Detailed Description

In the embodiment, the accuracy and reliability of water demand prediction are greatly improved by constructing a weather-water condition mapping relation, adopting a method of combining a full-scale model year method with two-dimensional joint distribution, integrating multiple prediction models for cross verification and the like. Firstly, a solid foundation is laid for fine scheduling through detailed description of water demand of different industries and departments. Then RVMF method and generalized simulation model are introduced, system screening is carried out on the multi-level sections, and key control sections with obvious influence on the modulation result are screened. By focusing the important sections, the scale and the computational complexity of the scheduling problem are greatly reduced, and the scheduling effect can be analyzed more focused. The section screening and the water demand prediction are combined, so that the refinement of the scheduling boundary and the high efficiency of the scheduling process are realized. And finally, introducing an MLAA-VIKOR model to perform non-inferior solution set optimization, fully exploring the feasible region of the scheduling scheme, simultaneously considering a plurality of scheduling targets, and balancing the requirements of all aspects. Compared with the traditional algorithm, the method and the system have the advantage that the systematicness and the robustness of the scheduling scheme are greatly improved. By adopting the technologies of CNN-transducer, machine learning and the like and combining with a hydrologic physical model, the modeling and optimizing capabilities are enhanced, the calculation speed is improved, the problems of high calculation complexity, poor real-time performance and insufficient adaptability are solved, and the process interpretability is ensured.

In a word, in the scheme, a complete, coherent and efficient planning flow is constructed by fully considering data cooperation, process coupling and logic cooperation of different stages and modules. The method has the advantages of full data accumulation and mapping analysis in the early stage, compact section screening in the middle stage and demand prediction, and multi-objective optimization solution in the later stage, and is relatively independent and closely cooperated. The system design based on the flow and the coupling improves the practicability of the whole scheme. In addition, the relevant system and standard construction are realized in the aspects of data acquisition and sharing, experience is accumulated in the aspects of model integration and algorithm optimization, universal components are deposited, and an analysis platform, a visualization tool and the like are built in the aspect of cross-department coordination.

According to one aspect of the present application, the step S1 is further:

The method is used for realizing depth fusion of multi-source heterogeneous data, not only expands the breadth and depth of the data and provides more comprehensive and three-dimensional watershed hydrologic feature description, but also lays a data foundation for subsequent feature extraction, relation mapping and analytical modeling. The difference of the multisource data in the aspects of space-time scale, distribution characteristics and the like also puts higher requirements on data preprocessing, characteristic engineering and the like, and proper data cleaning, standardization and other methods are required to be coordinated and unified.

And S12 to S14, performing dimension reduction and feature extraction on the multidimensional historical meteorological factors and the water regime data by adopting factor analysis and a CNN-transducer model. The factor analysis realizes data reduction and redundancy elimination to a certain extent by constructing a quantitative index system, reveals the internal correlation of meteorological factors and water conditions on space-time distribution, and provides good priori knowledge and optimization direction for subsequent deep feature learning. Then, the CNN-transducer model further optimizes a feature space based on a factor analysis result, adaptively extracts deep features hidden in high-dimensional data, and enables the subsequent weather-water condition mapping to be more accurate and efficient while achieving dimension reduction. Meanwhile, the graph convolution neural network is good at extracting local features, and the transducer model is good at capturing long-range dependence, so that the CNN-transducer model can comprehensively describe complex correlations in weather-water condition space-time evolution rules. In a word, the coupling application of the factor analysis and the CNN-transducer model has complementary advantages and brings out the best in each other, so that the data dimension is greatly reduced, the calculation efficiency is improved, and the rich information of the original data is also reserved to the maximum extent.

In the embodiment, through a series of data processing and modeling flows, the cooperation of the multi-source heterogeneous data is well realized, the potential value of different types of data is fully explored, and the obvious data fusion effect is embodied.

In a word, the quantitative relation between the meteorological factors and the water conditions can be used for parameter calibration and simulation verification of various hydrologic models, and physical and mechanical properties and simulation accuracy of the models are improved. In addition, the quantitative mapping relation has important value for reasonably evaluating the influence of future climate situations on watershed water resources, flood disasters and the like on the hydrological process response under the background of deeply understanding climate change.

According to an aspect of the present application, the step S13 is further:

And step S13f, training the CNN-transducer and storing.

In the embodiment, the idea of introducing positive and negative ideal points is adopted to convert the original continuity index into the relative quantity of the degree of approach to the ideal state, so that the quality difference of different index combinations can be reflected essentially, and the subsequent clustering and evaluation can be facilitated. Then, the random simulation method avoids the limitation of subjectively determining the division points, can adaptively generate a large number of representative training samples in the positive and negative ideal point range, and is beneficial to improving the generalization capability and the robustness of the model. Finally, the random simulation sample distribution is more uniform, so that the problem of unbalance of the sample is solved, the influence of outliers is reduced, and the stability and the convergence rate of the model are greatly improved. Therefore, the discretization process not only well reserves the distribution characteristics of the original data, but also realizes the expansion and optimization of the sample space.

In this embodiment, compared with the conventional equidistant or equal-frequency discretization method, the method uses positive and negative ideal points and random simulation to perform sample discretization on the historical meteorological factors and the water regime index system, and generates a training sample of the CNN-transducer, so that the method has great technical advantages.

In addition, the traditional discretization method generally directly inputs the original index value as a sample characteristic to a model, and the fitness coefficient reflects the relative position and the good and bad ordering of the sample in the index space to a certain extent, and contains more abundant discrimination information. The fitness coefficient is output as a model, so that on one hand, CNN-transformers are guided to learn the relative relation and difference characteristics among samples, and on the other hand, a new idea is provided for subsequent analysis such as sample clustering, abnormal point recognition and the like. In addition, the value range of the fitness coefficient is between 0 and 1, so that the standardization effect is achieved, and the samples with different sources and scales can be conveniently compared and evaluated. Overall, the combined application of the fitness coefficients with CNN-transducers enhances the model's ability to understand and characterize complex samples. The CNN can efficiently extract the spatial characteristics of the sample part through the partial convolution and pooling operation, and capture the association mode between indexes. The transducer well mines context semantic information and long-range dependency relations inside the sample through a self-attention mechanism and position coding. The two models are used in series, so that CNN-transducer can further extract global semantic information of samples on the basis of local feature learning, and the comprehensive representation comprising multi-scale and multi-level features is constructed, thereby greatly improving the recognition and understanding capability of the models to samples in different modes. In addition, because meteorological-hydrological data often exhibit scale effects and long memory characteristics, a single CNN or transducer model has difficulty in fully characterizing its time-space evolution law. And the hybrid structure of the CNN-transducer is expected to comprehensively extract sample characteristics from two dimensions of space and time through alternating iteration of convolution and attention, so that a more comprehensive and accurate mapping relation is established. Fully plays the complementary and synergistic effect of the advantages of the two models.

In conclusion, positive and negative ideal points and random simulation are used for sample discretization, a fitness coefficient is used for enriching sample information, CNN-transducer is used for feature extraction and relation mapping, and the positive and negative ideal points, the random simulation and the CNN-transducer are combined clearly and organically, so that an effective sample analysis flow is formed. The integration is to follow the data characteristics and model mechanism, and carefully design and optimize each step so that the steps can be matched and promoted mutually. For example, the discretized samples meet the requirements of CNN-transformers on data format and distribution, the fitness coefficient provides a training target with more discrimination for the model, and the feature extraction result of the CNN-transformers lays a foundation for analysis of sample clustering, anomaly detection and the like. Through seamless connection and effective coupling, the efficacy of each step is exerted to the maximum extent, and the overall technical synergy is realized.

In summary, in step S13, through a series of innovative designs such as sample discretization, fitness coefficient construction, CNN-transform hybrid modeling, etc., optimal extraction of sample features and accurate construction of mapping relationships are realized. The positive and negative ideal points and the random simulation improve the objectivity and the representativeness of discretization, the fitness coefficient enriches the discrimination information of the sample, the CNN-transducer plays the complementary advantages of the model, and the integration innovation of multiple technologies generates remarkable synergy. The set of flow can accurately describe causal relation from complex weather-water condition data, enhance the reliability of subsequent prediction and scheduling, improve the technical level of the whole scheme, provide a new paradigm for the research of an intelligent hydrologic analysis method, and play an important role in promoting the development of hydrologic subjects.

According to one aspect of the present application, the step S2 is further:

According to an aspect of the application, the step S21 is further:

According to an aspect of the application, the step S22 is further:

In this embodiment, the Feng Ping model year method fully utilizes the similarity and regularity of water demand in different years, and obtains a more representative water demand predicted value through classification statistics and weighted average of the abundant, flat and dead water years. Compared with a general time sequence analysis method, the method better considers the influence of hydrologic cycle, and the prediction result is more robust and reliable. By introducing two-dimensional joint distribution verification, two-dimensional probability distribution of an actually measured water demand sequence and a simulated prediction sequence is constructed, and confidence intervals and uncertainties of predicted values are quantitatively described. Compared with the traditional single error index, the joint distribution can reflect the coincidence degree and the discrete degree of the predicted value and the measured value more comprehensively, and visual probability evidence is provided for the reliability of the prediction method. The combination of the two improves the interpretation and the credibility of the prediction while guaranteeing the prediction stability, and is an important improvement and supplement to the traditional water demand prediction technology.

By adopting a cross-validation mode, a plurality of prediction models (trend epitaxy, multiple regression and machine learning) are integrated to predict the water demand of different departments, and an optimal method is selected in a self-adaptive mode according to the model performance. The method has the remarkable technical advantages that firstly, through multiple rounds of training tests, limited sample data are fully utilized, randomness and uncertainty of sample division are reduced, and model performance evaluation is more objective and accurate. Meanwhile, through the design of K-fold intersection, the limitation of a single verification mode is avoided, and the reliability and the robustness of the performance index are improved. The multi-model integrated prediction exerts complementary advantages of different types of models. The trend epitaxial model is good at capturing the long-term variation trend of water demand, the multiple regression model can characterize the association relationship between water demand and social economic indexes, and the machine learning model has unique advantages in modeling nonlinear complex relationships. The integrated application of the three models enhances the adaptability of the prediction method to different influence factors and data modes, so that the prediction result is more comprehensive and reliable. Finally, dynamically selecting the optimal prediction method of each department according to the performance index, and embodying the thought of cutting clothes according to local conditions and sizes. The self-adaptive optimization mechanism avoids the influence of artificial subjective judgment, and can be matched with the most suitable prediction technology according to the data characteristics and model effects of different departments needing water, so that the prediction precision is further improved. Overall, the combination of cross-validation + multi-model integration + performance preference is both robust and flexible, a new paradigm for water demand prediction that is effective.

Aiming at different departments (second, third industry, ecology and life) requiring water, a differential prediction model combination strategy is adopted. Due to the difference of water demand influencing factors of various departments and uneven availability and quality of data, the optimal effect is often difficult to achieve by adopting a unified prediction method. For example, the second and third industrial water requirements are closely related to the industrial scale and structure, and multiple regression, machine learning and the like can establish the association relationship between the water requirements and the multiple elements: the ecological water demand is more dependent on the ecological environment condition and the management policy, and the trend extension and regression method may be more applicable: the living water demand is related to factors such as population, town level and the like, and the advantages of the regression and machine learning methods are complementary. The differentiated model selection takes care of the characteristics and rules of different departments requiring water, and takes advantage of the modeling advantages of each method and is a key for improving the overall prediction level. In addition, even for the same department, the water requirement law can be described from different angles by adopting the cross verification of the combination of multiple models, so that the limitation of a single model is reduced, and the prediction result is more stable and reliable. Therefore, the differentiation and multi-element prediction strategy is a necessary way for the trend of water demand prediction to be refined and intelligent.

The method has the remarkable advantages that firstly, the formation mechanism of the water demand of each department and the difference of influencing factors are large, the modeling is respectively helpful to highlight main contradiction, the modeling key points are found out, and the defect of 'one-cut' is avoided. Meanwhile, the subsection gate prediction also provides possibility for differential model selection and parameter optimization, and can be accurately planned according to local conditions, so that the prediction precision is further improved. In addition, the subsection door prediction enables each water demand to be independently evaluated, and water resource management departments can conveniently and pertinently formulate demand response and water-saving measures. On the other hand, the water demand of each department is summarized to obtain the total amount, and the integrity and the relevance of the water resource system are reflected. The thought of scattered prediction and unified analysis not only plays the prediction advantages of all departments, but also strengthens the transverse connection of water requirements among the departments, and is convenient for viewing the balance of supply and demand and the water resource allocation from the global view. The prediction results of all departments are mutually verified, and the total amount and the components are mutually balanced, so that the prediction scheme is more scientific and reasonable, and a solid foundation is laid for the follow-up multi-target scheduling. In general, the method of subsection gate prediction and integration summarization not only refines the granularity of the description of the demand side, but also gives consideration to the overall cooperativity of the drainage basin, and the method is technically matched with each other and supported by each other, thus being an important innovation in the field of water demand prediction.

In conclusion, through a series of innovative designs such as a full-scale method, two-dimensional distribution, cross validation, multi-model integration, division gate differentiation prediction, dispersion aggregation and the like, the refinement level and the overall effect of water demand prediction are greatly improved. The technologies are mutually matched and have complementary advantages, and on the basis of accurately describing the water demand law and scientifically estimating the total water demand, high-quality demand side information is provided for subsequent section screening, dead water analysis, multi-target scheduling and other works, so that the applicability and reliability of the whole river basin water resource optimizing scheduling scheme are enhanced. Meanwhile, the multi-department differentiated water demand analysis can also provide important data support and scientific basis for river basin industry structure optimization, water saving policy formulation and the like, and the fine management and efficient utilization of the river basin water resources are assisted on a higher level.

According to one aspect of the present application, the step S3 is further:

According to an aspect of the present application, the step S33 is further:

S33a, constructing a generalized simulation model of a water resource system;

In this embodiment, when the initial section is selected, various factors such as watershed hydrology, water resources, and administrative management are fully considered. The selection of the initial section reflects the overall grasp of the natural condition and the social requirement of the flow field. On one hand, the selection ensures the comprehensiveness of the broken surface layout, and covers key nodes of water circulation and water resource utilization of the river basin: on the other hand, the problem guiding and management requirements are also highlighted, the water injection accident contradiction frequent areas are emphasized, and a grip is provided for developing fine management. Compared with the method for laying the section based on the terrain or administrative boundary, the method is closer to reality, the point selection is more targeted, and more valuable information can be provided for subsequent analysis. Meanwhile, in the mass alternative sections, the representative initial section is selected in multiple dimensions, and the data scale and the calculation load are greatly reduced. Overall, the comprehensive and typical and repeated initial section selection strategy lays a high-quality data foundation for section screening while reducing the workload.

And calculating ecological flow in the river channel by adopting RVMF method (variation range method), and superposing the ecological flow and water demand of each downstream department to obtain the drainage flow demand under the section. The ecological environment and the social and economic water flow requirements are considered, and the sustainable idea of watershed water resource management is embodied. On the one hand, the RVMF method fully considers the requirements of the ecological functions in the river channel on the variation of the flow, and determines the proper flow range required by maintaining the health of the ecological system of the river channel by analyzing the variation range of the flow in the same period for many years. Compared with the traditional Tennant method and the like, the method is more focused on the coupling of the ecological process and the hydrologic process, and the calculation result is more suitable for the dynamic property and the diversity of ecological water. On the other hand, ecological flow is overlapped with water requirements of downstream life, industry, agriculture and the like, comprehensive leakage requirements are obtained, interaction of human-water relationship is fully reflected, and multi-target attribute of river basin water resource allocation is revealed. The method ensures the basic requirement of the downstream, and simultaneously properly smoothes the fluctuation of water required, reduces the sudden increase and decrease of the section flow, and is more beneficial to the stability and controllability of the dispatching. In general, the method for determining the section leakage requirement by combining ecological flow and downstream water is a key measure for finely managing the water resources of the river basin, and comprehensively considers the social and economic water of the river basin on the basis of preferentially guaranteeing the ecological water. The construction of the generalized simulation model requires macroscopic overall grasp of space configuration, development and utilization modes, scheduling operation rules and the like of the watershed water resources. The scattered sections are integrated into a system framework, the influence of local regulation and control on the full-drainage basin is analyzed, the intrinsic mechanism of the influence of the sections is helped to clear, and key nodes of drainage basin response are grasped. Secondly, the model adopts a generalized structural form, and a drainage basin water resource operation mechanism is highly summarized and integrated through upstream and downstream section control conditions, inter-section water taking rules and the like, so that the complexity of the model is greatly simplified and the threshold of model application is reduced while the logic is kept clear. The method for revealing the river basin water quantity regulation effect through a small number of key nodes has important significance for supporting real-time scheduling decisions. And thirdly, importance diagnosis is carried out on the initial section by using a generalized model, and screening guidance of the representativeness and the criticality is embodied. Through setting up the section influence threshold value, the control effect of multiple needs water combination sight of operation ration aassessment section, and then the key section that the influence is showing to full river basin is screened out, can be when reducing management object, more accurately balance the water yield balance of accuse river basin. The multi-scenario analysis also reflects the consideration of uncertainty to a certain extent, and enhances the robustness of the screening result. In general, the section screening method based on generalized simulation performs beneficial exploration in the aspects of basin integrity, conciseness, representativeness and the like, and is a powerful tool for finely managing basin water resources.

In a word, the multi-source data and the multi-disciplinary method are subjected to deep fusion, so that tight logic connection and functional coupling are formed among links of section screening, and a remarkable forward effect is generated. Specifically, step S31 provides a solid data base for section screening, and lays a starting point for full-drainage-basin analysis: step S32 is based on the overall system, further focuses on key factors, and leads to the identification of the subsequent key sections through the key factor of the leakage requirement: step S33 quantitatively depicts the influence of the section on the basis of the first two steps in a model simulation mode, and an operable screening standard and an operable implementation path are formed. Therefore, links such as section primary selection, demand determination, model diagnosis and the like are buckled and pushed layer by layer, so that a complete technical scheme of the system is formed. The organic connection of each link not only improves the pertinence and the effectiveness of the section screening, but also lays a solid foundation for the subsequent water quantity scheduling. From a higher level, the cross fusion of multidisciplinary technology, the gradient progression of data, models and decisions also represents a new thought method for the water resource management of the watershed, and has important demonstration significance for realizing the modernization of the watershed treatment system and the treatment capacity. The important control section playing a key role in the whole-river basin water resource space allocation is screened by the system, so that the water quantity scheduling can be guided to optimize in the direction of 'force on point and effect on surface', and the management objects are reduced and the result is doubled. The identification of key sections is also beneficial to optimizing the layout of a monitoring station network and enhancing the management and control of hot spot areas, and the focus of the river basin management to the key areas, key periods and key points is promoted at a higher level. In addition, by quantitatively defining the influence of the section, important reference basis can be provided for water weight distribution, ecological compensation and the like in the river basin, and a novel river basin ecological protection compensation mechanism combining transverse ecological compensation and longitudinal fund transfer is promoted to be established. In a word, the central awareness of the river basin formed by section screening has profound and positive effects on the management system of the river basin, the improvement of a long-acting management and control mechanism and the promotion of the sustainable development of the river basin.

In summary, through links such as comprehensive screening, scientific determination, system diagnosis and the like, on the basis of river basin multi-source heterogeneous data, the system screens key control sections which play a decisive role in the configuration of the water resources of the whole river basin. All links are mutually connected and deepened layer by layer, the overall efficiency is highlighted on the basis of technology integration, and the method has important effects on the fine management of water resources in supporting watercourses, the improvement of scheduling efficiency, the optimization of space configuration and the like. The section screening system, strategy and scientificity create a new situation of river basin water resource management, and also indicate the direction for the subsequent river basin treatment works such as water volume scheduling, water quality control, ecological protection and the like.

According to one aspect of the present application, the step S4 is further:

According to an aspect of the present application, the step S42 is further:

According to an aspect of the application, the step S43 is further:

In the embodiment, on the basis of objective function setting, the water demand of the sub-river basin and the drainage flow under the section are juxtaposed, so that the overall consideration of all links of the water circulation of the river basin such as upstream, downstream, left and right banks, earth surface, underground, industry, ecology and the like is embodied, and the aim of maximizing the overall benefit of the river basin is achieved. Meanwhile, through the design of multiple objective functions, the requirements of different stakeholders are effectively coordinated, and the reasonable water use requirement of regional socioeconomic is met to the maximum extent while the ecological safety of the river basin is ensured. In the processing of constraint conditions, the introduction of engineering capacity constraint ensures the physical feasibility and engineering accessibility of a scheduling scheme, so that the scheme is always optimized in the scope of engineering bearing capacity. In addition, the model can further incorporate targets and constraints such as water quality and the like, and realize the coupling regulation and control of water quantity, water quality and sediment, so that the model has important reference significance for the comprehensive management of the river basin. In general, the constructed multi-objective scheduling optimization model is based on the practical river basin, gives consideration to multiple requirements of society, economy, ecology and the like, performs beneficial exploration on the 'diversification' of targets and the 'refinement' of constraints, and is a key technical support for optimizing configuration of water resources of the river basin.

Solving the scheduling model by adopting a plurality of multi-objective optimization algorithms, and generating a non-inferior solution set. Through algorithm integration application, the advantages of different algorithms are effectively fused, the advantages are improved, the advantages are avoided, the advantages are mutually complemented, and the solving precision and the solving efficiency are greatly improved. More importantly, due to the high complexity and uncertainty of the water resource scheduling problem, a single algorithm is often difficult to solve stably. And by means of multi-algorithm parallel solving, problems can be analyzed from different angles and different strategies, a 'comfortable area' of the algorithm is jumped out, and solving robustness is enhanced. In addition, the multi-algorithm parallel design also provides a richer sample space for subsequent solution set analysis and preference screening. In general, the integrated application of the multi-objective optimization algorithm is black technology for solving the problem of complex water resource scheduling, and provides more choices for subsequent decisions while improving the calculation efficiency.

And (3) optimizing the non-inferior solution set generated by the multiple algorithms by adopting an MLAA-VIKOR model, so as to obtain a final scheduling scheme. The step is creatively explored in the field of multi-objective optimization post-processing, and the defects of the prior researches are well overcome.

Conventional practice is to choose from a number of non-inferior solutions at random, or to determine the final solution based on some criteria, often with subjective randomness. Aiming at the problem, the scheme creatively introduces a VIKOR method and an MLAA technology, and constructs a set of objective and efficient non-inferior solution set optimization mechanism. The VIKOR method starts from decision attributes of each alternative solution, and obtains a comprehensive utility value of the solution by calculating the proximity degree of each solution to positive and negative ideal solutions, thereby providing a quantization basis for rational screening. The thought based on attribute value difference can be reasonably balanced among complex influence factors, find out 'the satisfaction solution with harm but minimum loss', and has important significance for balancing the appeal of a plurality of benefit groups. And the MLAA is further combined with the solving results of different optimization algorithms under the framework of multi-layer hierarchical, and finally locks the global optimal solution through the aggregation analysis of hierarchical iteration, so that the problems of algorithm preference, initial value dependence and the like are avoided. The microscopic-to-macroscopic and local-to-whole analysis logic effectively solves the evaluation cracking problem possibly caused by multi-algorithm parallelism, and maximally exerts the integration advantages of different algorithms. In general, the MLAA-VIKOR model provides a brand-new thought and method for multi-target water resource optimization scheduling, and creates a new situation of river basin fine management.

In a word, seamless connection and deep combination of links such as model construction, algorithm solving, decision optimization and the like are realized in the technical level, a complete closed loop flow of information input-model analysis-scheme output is formed, and scientificity and effectiveness of scheduling decisions are greatly improved. Based on the key control section of the river basin, a scheduling model is constructed on the basis of comprehensively analyzing the supply and demand conditions of the region, and a plurality of algorithms are adopted for solving, so that the key and key of the river basin management are fully embodied in the technical links. Ensuring the technical realization to be highly matched with the management requirement. The multi-algorithm non-inferior solution set is directly connected, and the optimal scheduling scheme is rapidly locked through system evaluation, so that the result conversion efficiency is improved. In addition, each step is buckled and organically unified, and the dispatching scheme is grounded on the basis of multi-source data aggregation, multi-target weighing and multi-scenario analysis. The effective coupling of each link also promotes the communication collaboration of different fields of students and different management departments, and provides a platform support for refined management.

at least one processor; and

A memory communicatively coupled to at least one of the processors; wherein,

In the application, the river basin water resource scheduling index mainly comprises: in the embodiment, by constructing a quantitative index system of the space-time distribution characteristics of the historical weather factors and the water conditions and a mapping relation of the historical weather factors to the water conditions, the rain conditions are predicted by weather factor prediction, more accurate prediction results can be obtained compared with the traditional method, meanwhile, the ecological flow of the river channel is calculated by a RVMF method considering the water conditions, the section leakage rate is determined, meanwhile, in order to solve the problem that the calculation targets are too many and difficult to solve due to the fact that the number of sections is too many, the sections which are important for influencing the water quantity of the river basin are selected for analysis by screening the sections, the calculation engineering is reduced, and the data required by the water resource scheduling plan is obtained more quickly.

The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to the specific details of the above embodiments, and various equivalent changes can be made to the technical solution of the present invention within the scope of the technical concept of the present invention, and all the equivalent changes belong to the protection scope of the present invention.

Claims

1. The method for planning the river basin water resource scheduling in the water network system is characterized by comprising the following steps:

2. The method for scheduling water resources in a river basin in a water network system according to claim 1, wherein the step S1 is further:

3. The method for scheduling water resources in a river basin in a water network system according to claim 2, wherein the step S13 is further:

And step S13f, training the CNN-transducer and storing.

4. The method for scheduling water resources in a river basin in a water network system according to claim 1, wherein the step S2 is further:

5. The method for scheduling water resources in a river basin in a water network system according to claim 4, wherein the step S21 further comprises:

6. The method for scheduling water resources in a river basin in a water network system according to claim 4, wherein the step S22 is further:

7. The method for scheduling water resources in a river basin in a water network system according to claim 1, wherein the step S3 is further:

8. The method for scheduling water resources in a river basin in a water network system according to claim 7, wherein the step S33 further comprises:

S33a, constructing a generalized simulation model of a water resource system;

9. The method for scheduling water resources in a river basin in a water network system according to claim 1, wherein the step S4 is further:

10. The method for scheduling water resources in a river basin in a water network system according to claim 9, wherein the step S42 is further:

11. The method for scheduling water resources in a river basin in a water network system according to claim 9, wherein the step S43 is further:

12. The river basin water resource scheduling planning system in the water network system is characterized by comprising:

at least one processor; and

A memory communicatively coupled to at least one of the processors; wherein,

The memory stores instructions executable by the processor for performing the river basin water resource scheduling method in the water network system of any one of claims 1-11.