Concept of digital twin construction scheme for flood storage space in mid-lower Yangtze River

The integration of the digital twin with the water industry is a new generation of information technology and communication technology in the integrated application of the water industry, is an important implementation path to promote the new stage of high-quality water development. The current level of wisdom in the water conservancy industry is still low, and there is relatively little research and application for the digital twin. For this reason, the basic connotation and development concept of the digital twin is discussed in depth, and research is conducted focusing on the problems that may be faced in the construction process of the digital twin in the flood storage space. The study gives a general scheme of flood storage space construction with data base construction, digital scene construction, intelligent simulation and accurate decision support construction as the main axes. The related results can provide constructive planning ideas for future smart river basin construction.


Introduction
Digital twinning is an important supporting technology for watershed digital transformation. It mainly uses historical data and real-time monitoring data obtained by sensors. On the basis of algorithm model, it uses a variety of information and communication technologies such as distributed storage, digital mapping and cloud computing to digitally map multi-scale, multi variable, multi physical field and multi probability processes in the real world in the virtual world. It also carries out historical replay, realistic presentation and future prediction of the whole life cycle process of corresponding physical entities, realizes the twin interaction and integration of the physical world and the virtual digital world, comprehensively improves the application ability of information and data in the physical world, reduces costs and improves efficiency [1,2] . With the development of various sensing communication, information measurement and control decision-making technologies and the vigorous improvement of computer software and hardware level, digital twin technology also provides unlimited innovation opportunities for the development of all walks of life. Since 2016, Gartner Group has accelerated the development of high-speed technology and smart city, and has made a series of achievements in the field of industrial transformation and smart manufacturing, especially in the field of energy and smart city [3][4][5] .
In recent years, the Ministry of water resources has actively promoted the construction of informatization and intelligence in the water conservancy industry. With the deepening of information resources and business applications, digital twin technology has also been popularized and applied to the planning and management of river basin water resources. In November 2018, IBM first proposed the concept of smart earth, and based on this, it developed a set of intelligent water systems for the Colombian water authority, enabling the water authority to manage the water infrastructure prospectively [6] . Referring to the concept of "smart earth", Jiang et al. [7] analyzed the strategic needs and technological driving factors of "smart basin", and analyzed the key technologies and supporting platforms used therein. For example, Internet of Things technology, RFID tag, wireless sensor network, cloud computing and cloud storage, watershed virtual reality system platform, meteorological and hydrological information guarantee platform based on multi-source coupling, binary water cycle numerical simulation platform, water resources numerical regulation platform and watershed data assimilation system platform are discussed. Jiang et al. [8] analyzed the shortcomings existing in the current operation and management of water conservancy projects, focused on the operation mechanism of the application of digital twin technology in the operation and management of water conservancy projects, and proposed that the operation process of water conservancy projects can be monitored, diagnosed, analyzed, decided and predicted in real-time through digital twin technology, so as to realize the intelligent operation, accurate control and reliable operation and maintenance of water conservancy projects at the same time, the "digital twin basin" and "digital twin water network" integrating digital and water conservancy have become the symbols of high-quality development of the water conservancy industry, and a variety of intelligent water conservancy system platforms are emerging, including "smart Yangtze River", "Digital Yellow River", "one map of water conservancy", "smart dam" Digital Water Conservancy Internet of Things, etc.
However, the water conservancy industry is still in the initial stage of smart water conservancy, and there is still a large gap compared with smart industries such as urban transportation, electric power and meteorology. The existing water resources information system is constructed in many aspects to the specific water resources business needs, the lack of coordination and sharing between various subsystems and modules, prominent problems such as data barriers and information islands, low overall intelligence level and lack of ways to interact with the outside in real-time [9] . In particular, the digital and intelligent construction of river basin flood control systems still faces many problems, such as insufficient construction of monitoring station network systems and the inability to perceive various dynamic change processes of the river basin in real-time.
At the same time, due to the weak data foundation and the large gap between the model foundation and the actual application, the computing power of the existing model is difficult to support the actual dispatching application requirements, and the simulation accuracy is low, which can not reflect the mutual feedback relationship between the change of the real world and the response of the virtual world. Therefore, how to establish an intelligent service platform for watershed water resources management based on the new requirements of smart watersheds in the new era and digital twins has become an inevitable choice for the high-quality development of water conservancy in the new era.
In this context, this paper will focus on the understanding and thinking of the basic connotation and development concept of the watershed digital twin, systematically discuss the overall technical scheme and construction content of watershed flood storage space digital twin construction with the goal of watershed flood storage space digital twin construction, and preliminarily discuss its future development direction. With a view to realizing information digitization, control automation and intelligent decision-making through digital twin technology, strengthening the construction of "four pre" functions of prediction, early warning, rehearsal and plan, comprehensively promoting the development of water conservancy modernization and promoting the intensive utilization of water resources.

Basic connotation and characteristics of digital twin basin
Digital twin basin is the application of digital twin technology in the comprehensive management of water basin resources. In essence, it is to establish a twin basin system in the virtual digital space through information technology, which is one-to-one corresponding to different elements in the real world, mutual mapping of different physical processes, and real-time collaborative interaction with the external world. The target objects of digital twin basin include important projects such as reservoirs, embankments, flood storage and detention areas within the whole water basin, different elements such as key river sections, hydrological stations and coastal cities, as well as the physical change process of various elements at different scales, such as rainfall, runoff generation, confluence and prediction and dispatching of hydraulic projects.
Compared with the traditional water basin resources management system platform, the biggest advantage of the digital twin lies in the organic integration of mathematical model and Internet of Things monitoring system. Through the external Internet of Things monitoring and sensing system, it can provide good data support and more powerful modeling means for the simulation modeling of real basin. Through real-time monitoring of the information transmission chain of perception, communication transmission and initial boundary update of the background model, the background model can continuously obtain and update external information, and more perfect "digitization" and "interest" of the real world. At the same time, in the face of insufficient understanding of the mechanism of complex water resources systems, based on a large number of data samples collected, a "gray box" or "black box" model can be established through artificial intelligence and other technologies to simulate and model the complex process, which can effectively develop, improve and supplement the traditional modeling technology based on physical process and assist in accurate decision-making.

Construction route and development concept of digital twin basin
The digital twin basin can be developed and constructed according to the overall route of database construction, digital scene construction, intelligent simulation and accurate decision support construction.
(1) Database. Based on space-time big data, the database collects business data in different fields, real-time monitoring data from sensors, and forms a circular data flow integrating information collection, recording and decision-making in combination with 5G, Internet of Things and other new generation information and communication technologies, the database provides the foundation for the construction of digital twin basin. It is also the link between twin basin and real basin. It is mainly based on the digital identification of global elements and the network system of air, space and earth integrated perception and monitoring stations, and links the virtual world with the physical world through external data dynamic perception, data organization and management and image digital mapping.
(2) Digitize the scene. Digital scene construction is the depiction of the twin basin to the real basin. It mainly takes the digital twin model platform as the information integration display carrier for the comprehensive utilization of basin water resources, introduces GIS, VR and other three-dimensional visualization technologies, and constructs a multi-scale integrated digital twin basin theme scene according to different business decision-making needs.
(3) Intelligent simulation. The intelligent simulation focuses on the interactive response of twin basins to the real world. It mainly takes the global panoramic data resources, high-performance collaborative computing and deep learning machine intelligence platform as the "wisdom center" to accurately describe and simulate the whole process of the whole elements of the real basins in the virtual twin basins.
(4) Accurate decision-making. Precise decision-making refers to the role of twin basins on real basins. Combined with the intelligent simulation dispatching results, the dispatching scheme is refined and optimized from the perspectives of time, space, quantity and degree, so as to support the fine implementation of dispatching decision-making under different temporal and spatial granularity, and quantitatively evaluate the impact of the scheme on the real world.
Through the construction and improvement of digital twin basin, a universe of business objects can be created in the future, a complete set of new modes of water basin resources operation and management can be further formed, such as "global three-dimensional perception, trusted interconnection of all things, dynamic scene display, distributed ubiquitous computing, panoramic interactive regulation and data-driven decision-making", as shown in Figure 1.

Digital twin characteristics of flood discharge and storage space
As there are too many elements in the basin and the change process is complex, the existing technical means are difficult to carry out the overall digital twinning of the basin. It is necessary to carry out the digital twinning construction in sub-regions and subprojects according to the characteristics of different elements and business needs. Flood storage space is an important part of the digital twinning of the basin, and its twinning object is mainly for the construction of weak links in the existing flood control system, such as flood storage and detention areas, beaches and embankments. At this stage, due to the lack of control and lagging construction of flood storage space such as flood storage and detention area in the basin and embankment, the digital twin construction of flood storage space has the characteristics of large construction scope, weak data foundation and unclear application mechanism. Therefore, it is urgent to establish a more intelligent and accurate spatial intelligent twin simulation system for flood discharge and storage, so as to provide strong support for spatial dispatching and control of flood discharge and storage and flood control and disaster reduction in the basin.

Overall scheme for digital twin construction of flood discharge and storage space
Combined with the overall requirements of "four pre-projects", carry out the construction of digital twin basin according to the mode of "large system design and subsystem construction". Firstly, based on historical water regime, rain regime, engineering condition, social condition data, natural geographical data and real-time observation data of station network, develop a database that can perceive mutual feed and update in real-time. Then, fully investigate the business needs, and package the professional models such as river hydrodynamic evolution model, flood storage space application model and risk assessment and decision-making model in a unified and standardized format to form a professional model library. On this basis, combined with artificial intelligence and other technologies, high-order models such as knowledge maps, intelligence maps, model error feedback correction models, risk group decision-making models and adaptive regulation are established. The main characteristics of such models are high integration and interpretability. The model mechanism is more in line with human learning and decision-making process, which can effectively improve the computing power of existing models and improve model functions. At the same time, the highorder knowledge model is also the key support for the twin basin to conduct "millisecond" intelligent simulation of the real basin panorama, all elements, all time and space and the whole process. According to the danger and disaster information obtained from real-time monitoring and intelligent simulation, combined with the group intelligent decision support system, accurately control the space-time state and evolution trend of risks under different scheduling schemes, and complete the construction of digital scene integrating global danger presentation, full situation awareness and early warning and disaster loss mutual feedback response through visualization technology. Realize real-time feedback optimization and accurate decision-making of dispatching scheme under the condition of human-computer interaction. The overall road map of digital twin construction of flood discharge and storage space is shown in Figure  2.

Digital twin construction objectives of flood discharge and storage space
The digital twin construction objectives of flood discharge and storage space can be divided into the following four objectives: (1) Establish a database that can be updated in real-time based on the monitoring system of the air space integrated IoT station network.
(2) Build a multi-scale digital scene, which can improve the presentation method of a traditional digital scene from the low-dimensional point, line and surface display method to the multi-dimensional wide area and panoramic display method.
(3) Promote the water engineering simulation dispatching from the experience-based simulation dispatching mode to the whole factor, all time and space intelligent map dispatching mode.
(4) Improve the decision support from the traditional decision-making method based on static risk indicators to the interactive dynamic risk group decision-making method.
Finally, an embedded digital twin basin decision support platform with deep autonomous learning ability and continuous adaptive optimization can be built, which integrates basin perceptual monitoring, prediction and scheduling, scheme rehearsal, intelligent decision-making, early warning release and plan generation. Improve the integrated dispatching system from the perspectives of scene display, computing power and decision-making interaction.

Data base construction
The basin data base is a new generation of infrastructure supporting the high-quality development of water conservancy modernization. The construction of data base should be combined with the needs of different business scenarios, consider the difficulty of data acquisition, construction cost, construction cycle and other factors, and divide the basin data base into different scenarios of large scale, meso scale and fine scale according to the overall construction idea of "multi-scale construction and multi-scale integration" (see Figure 3).
(1) Construction of basin level data base. At the basin level, it belongs to large-scale, and its data base construction focuses on the overall information of the basin. The basin level data base takes the 3-D or 2.5-D topographic map of the largescale basin as the base map, and depicts the key information that can assist decision-making into the 3-D base map. The specific data information includes: large-scale satellite image data of the basin, DEM data, administrative division data, road data, water system information, place name data and other natural geographic information; social and economic data information of regional administrative divisions, population, cultivated land, roads, units and enterprises; spatial location of different flood storage and detention areas, beaches and embankments, realtime engineering status information and other engineering information; spatial information of important buildings such as embankments and realtime engineering information; and real-time water and rain information at different control stations.
(2) Construction of database at the river section level. The river reach level belongs to meso scale, which focuses on the information in a specific river reach. The database at the river section level takes the high-precision topographic map of the river section as the base map, and carries out three-dimensional modeling for the riverway embankment, flood storage and detention area, beach civil embankment, important flood storage engineering facilities, as well as important river crossing buildings, important riverside buildings and important risk embankment sections in the nearby wading areas. The white model can be used in some areas. And according to the element geospatial layout information provided by satellite images. Superimpose the real-time water and rain information of different stations, water level information along the river, flow velocity information, design water level information of embankment, embankment level, the elevation of embankment top, internal attribute information of embankment section, information of key risk hidden danger points, location information of UAV and camera layout, real-time monitoring picture information, socio-economic information such as population and GDP of coastal cities, road information and other information in the river reach scale, digital three-dimensional model.

Digital scene construction
Based on the construction of data base, it is necessary to combine the actual business needs of different stages and carry out according to the concept of "large-scale decision-making and small-scale fine", that is, facing the application requirements of different scenarios such as planning and design and local fine simulation scheduling, build a digital theme scene that can deeply describe the dynamic process and characteristics of the system at the corresponding scale. Among them, the construction goal of large-scale and medium-scale scenes is mainly to enable decision makers to conveniently and quickly grasp the overall information and make overall planning and decision. The construction of local refinement scale scene is mainly to enable the decisionmakers to finely grasp the risk changes of key projects. The digital scene construction of flood discharge and storage space can be divided into largescale scene construction of river basin, mediumscale scene construction of river reach and local scale scene construction of flood discharge and storage space.

Construction of large-scale panoramic decision-making scene in the basin
Large-scale scene construction focuses on the macro-overall information of the basin. The scene contents displayed are different in different flood control operation stages. In the prediction and early warning stage, the monitoring information and prediction information of rainfall-runoff at different stations, as well as the engineering information such as water level and flow process of key river sections are displayed intuitively and stereoscopically in various ways such as 3D animation and 3D thematic map. According to the predicted flow exceeding the limit and water level rise, the length of dike dangerous section with different types of risks such as piping, seepage and collapse is calculated in real time, and the comprehensive statistical indicators such as the range of the maximum dangerous section and the maximum disaster bearing time of the dike in the prediction period are displayed on the 3-dimensional bottom map.
At the same time, the large-scale twin scene of the basin can also provide convenient and fast human-computer interaction function of the dispatching scheme. With the support of the background high-level knowledge model, through the fine modification of the dispatching scheme, the simulation results of different flood storage space application schemes can be quickly obtained while meeting the simulation accuracy In addition, the digital twin system will provide more dimensional dangerous and disaster description information. For example, the size of different types of comprehensive risk indicators such as total over alarm time, total over alarm length and total inundation loss GDP can be given in the form of benefit-risk index decision-making warehouse. Through the intelligent decision support system, the evaluation results of each expert on different schemes can be recorded and counted, and the dispatching schemes can be comprehensively sorted and optimized through model calculation.
Under the digital twin background, the largescale decision-making scenario of the basin will greatly improve the early warning risk description, scheme calculation efficiency, decision-making indicators and decision-making methods compared with the traditional dispatching methods.

Digital scene construction of river reach scale
The perspective of digital scene of river reach scale mainly focuses on the river reach, focusing on the effect of the use of beaches, civil embankments and flood storage and detention areas near the key river reach on the risk reduction of the river reach. According to the hydrological prediction results, combined with the flood routing model driven by knowledge map, the evolution characteristics of water flow in the main stream, and lake area in the forecast period can be quickly simulated. By selecting or checking different flood storage and detention areas and embankments near the river section, combined with flood storage space dispatching, the knowledge map is used to conduct flood storage space dispatching rehearsal, judge the effect of different flood storage space put into use combination on the water level lowering of the embankment along the control station, and mark and highlight the water level lowering space. On this basis, the changes of risk along the river embankment are finely displayed and quantified through thermal map and statistical table. At the same time, the time axis can be displayed at the bottom of the interface to dynamically display the dynamic changes of factors such as the embankment water level, safety factor, flood storage space water level, flood storage volume, flood diversion flow, inundation loss of coastal cities and towns at different times.
In addition, according to the analysis of the historical dangerous situations, multiple historical dangerous sections can be set near the key control river sections as risk points. Click the risk point to enter the panoramic view, conduct cloud patrol and monitoring of the risk point in the roaming state, and depict the statistical data such as the water level along the river near the risk point, the design flood level of the embankment, the real-time dangerous situation of the embankment, and the inundation loss in the three-dimensional panoramic view, so as to assist the decision-makers to intuitively grasp the dangerous situation and disaster situation in the important risk area from a close distance.
The application of twin scenes can enable decision makers to more intuitively and finely grasp the dynamic changes of different sections and types of risks in key river sections. Figure 4 shows the digital scene construction of the river reach scale.

Digital flood storage space construction
The construction of a spatial scale digital scene for flood discharge and storage focuses on the construction of flood diversion application scene and risk avoidance and transfer scene. After flood diversion is enabled. Through monitoring equipment such as drones and cameras, make fine decisions on the number of flood diversion gates, the opening degree In addition, it can display the real-time status information of risk aversion and transfer people in the form of a thermal diagram in combination with the function of "three operators" (see Figure 5).

Intelligent simulation support construction
The traditional flood diversion application and dispatching effect analysis of flood storage space is often implemented based on the one-dimensional and two-dimensional hydrodynamic model, with high calculation time cost, which greatly limits the application of flood storage space in real-time joint flood control dispatching, and cannot quickly and accurately reflect the information feedback of the real disturbance of the physical world to the twin world and the influence of the interactive decision-making of the twin world on the real physical world. Under the background of data twinning, due to the significant increase in data dimension and data volume, including the real-time update of a large number of monitoring data, the data-driven model has become a new way to solve the difficulty of using flood storage space in real-time dispatching simulation analysis. Modeling and simulation of downstream flood storage space dispatching of emerging technology system is also an important research direction for the construction of digital twin intelligent simulation support of flood storage space. Therefore, based on the existing river flood hydrodynamic evolution model, flood storage and detention area, embankment flood diversion application model and other professional models, based on big data analysis and artificial intelligence methods, and according to the characteristics of different river sections. It is necessary to extract the characteristic factors of the sample data of river flood historical evolution process, and establish the mapping relationship between historical incoming water and predicted incoming water information and the future water level and discharge process of the section. Develop an intelligent evolution model that can realize fast and accurate river flood evolution simulation on this basis, further explore the action mechanism of flood control engineering groups on the hydrological situation of rivers and lakes, extract the entity combination of flood control engineering with obvious flood control compensation effect for downstream protection objects, introduce a variety of machine learning methods to build an interpretable artificial intelligence dynamic learning model, and build an analysis model of the application effect of flood control engineering with "time, space and quantity" cross-dimensional utility. Support fast and accurate simulation calculation of full state perception and scheduling scheme in human-computer interaction mode.

Construction of accurate decision support
The intelligent empowerment of the database through the high-order knowledge model improves the calculation efficiency and accuracy of the scheduling simulation model, which can further assist in improving the decision-making efficiency. However, since the construction of flood discharge and storage space twin system faces the business requirements of large-scale, full-time, multi-scale and strong coupling process, the operation simulation process of twin system will produce complex decision-making information. Under the background of wide area and complex decision-making information flow, how to combine the advantages of information diversity, integration and high interaction of digital twin system is the top priority to support the accurate decisionmaking of flood control dispatching. Therefore, the construction of accurate decision support for flood storage spatial dispatching should focus on the functional development of multi-attribute risk assessment and group intelligent decision-making of dispatching scheme. Based on the intelligent simulation function, the loss assessment model of the spatio-temporal situation map should be established according to the change characteristics of the different river and lake elements before and after dispatch. According to the changes of water and rain conditions before and after dispatching, the disaster information such as inundation affecting houses and population under different dispatching schemes should be evaluated and obtained, so as to realize the refined dispatching scheme and accurately grasp the risk change scenario of the whole region. On this basis, combined with the expert group evaluation information, establish a group intelligent decision-making model, comprehensively consider a variety of benefit and risk indicators, coordinate and consider the different suggestions of the expert group, and select a comprehensive and optimal scheduling scheme. Finally, a set of "space-time measurement" three-dimensional fine dispatching decision-making system will be built to realize the accurate use of flood storage space, accurate characterization of risk loss, fine distribution of flood volume and accurate transfer of personnel to avoid flood, as shown in Figure 6.

Conclusions
In the future, the smart basin will have a higher level of informatization and digitization, so as to truly realize the concept of a digital twin basin as a whole. As a key component, the spatial digital twin of flood discharge and storage will play an important role in supporting flood control and disaster reduction in the basin. In this context, this paper discusses the basic connotation and development concept of digital twin basin in detail. Taking flood storage spatial data twinning as an example, this paper puts forward the overall scheme of flood storage spatial digital twinning construction with database construction, digital scene construction, intelligent simulation support construction and fine decision support construction as the main axis, and discusses the functional significance of different construction contents and the function improvement relative to the existing technical system. Considering the realistic conditions of intelligent construction of the basin, the research focus at this stage should be the twin of accurate, omni-directional and dynamic mapping of the current situation of the basin through the data base. In addition, under the background of a large number of measured and historical data, the twin basins can learn by themselves through learning, analyzing, identifying and summarizing the operation law of basins through high-order knowledge model. On this basis, combined with expert experience, under the condition of human-computer interaction, realize the simulation twins of future development scenarios of the basin under different environments and different dispatching schemes, and finally make accurate decisions. In the future, we can further integrate and use technologies such as artificial intelligence and machine learning to expand knowledge beyond human cognition, deconstruct the water basin resources management mode from the perspective of pure mathematics, analyze and refine the possible problems of water basin resources management, and automatically derive corresponding technical methods to solve the problems, so as to realize the autonomous twin basin.