Risk evaluation of urban rainwater system waterlogging based on neural network and dynamic hydraulic model

Abstract

With the rapid development of urban construction and the further improvement of the degree of urbanization, despite the intensification of the drainage system construction, the problem of urban waterlogging is still showing an increasingly significant trend. In this paper, the authors analyze the risk evaluation of urban rainwater system waterlogging based on neural network and dynamic hydraulic model. This article introduces the concept of risk into the study of urban waterlogging problems, combines advanced computer simulation methods to simulate different conditions of rainwater systems, and conducts urban waterlogging risk assessment. Because the phenomenon of urban waterlogging is vague, it is affected by a variety of factors and requires comprehensive evaluation. Therefore, the fuzzy comprehensive evaluation method is very suitable for solving the risk evaluation problem of urban waterlogging. In order to improve the scientificity of drainage and waterlogging prevention planning, sponge cities should gradually establish rainwater impact assessment and waterlogging risk evaluation systems, comprehensively evaluate the current capacity of urban drainage and waterlogging prevention facilities and waterlogging risks, draw a map of urban rainwater and waterlogging risks, and determine the risk level. At the same time, delineate drainage and waterlogging prevention zones and risk management zones to provide effective technical support for the formulation of drainage and storm waterlogging prevention plans and emergency management.

Keywords

Urban rainwater system waterlogging;risk assessment;neural network

1 Introduction

The emergence and development of rainwater systems in modern cities, from the early collection and transportation of rainwater to point source treatment, to the current stormwater management and integrated management of rainwater systems, rainwater systems are gradually improving their control and management methods to maximize their effectiveness [1]. As an indispensable and important municipal infrastructure, the drainage system integrates the drainage of rain and sewage, the improvement of the living environment, and the prevention of pollution of public waters, which promotes the healthy circulation of urban water [2, 3]. If the water supply system is compared to the arteries of the city, the drainage system is the vein of the city, and it is the lifeline project to restore the water environment and make the city develop healthily [4]. It protects the city from sewage and storm water, creates a safe and hygienic living and production environment for the city, enables the energy flow and hydrological cycle of the urban ecosystem to proceed normally, and ensures the sustainable development of the city [5].

With the rapid development of urban construction and the further improvement of the degree of urbanization, although the construction of municipal facilities, including drainage systems, has been intensified, the problem of urban waterlogging has still shown a significant trend, especially in coastal cities [6, 7]. Waterlogging disasters caused by natural factors such as wind, storm, and tide are even more serious. It causes water accumulation in houses, flooding of large farmland, paralysis of transportation hubs, and even makes the underground power equipment of cars and high-rise buildings ineffective [8]. And urban lifeline systems such as electricity, water, gas, and communications were damaged.

From the above, we can see that urban waterlogging disasters have continuously threatened urban security for many years. It brings risks to the normal operation of the city [9, 10]. Therefore, it is necessary to set up advanced computer simulation methods to simulate different working conditions of the rainwater system and carry out flood risk assessment and risk management, so as to reduce waterlogging losses. It is also necessary to provide decisions for the urban rainwater system manager stand by [11].

The urban rainwater pipe system is a pipe system that is laid below the ground to transport rainwater by gravity flow. It belongs to the outdoor rainwater pipe system [12]. The auxiliary structures of the pipeline system include inspection wells, stormwater outlets, overflow wells and water outlets, which are mainly used for connecting pipes and regulating the amount of water. From the material point of view, rainwater canals are usually concrete pipes, reinforced concrete pipes, plastic pipes and brickwork channels [13, 14]. Because the flooded areas are generally low-lying or flat terrain and long rainwater pipelines, in these areas, rainwater is difficult to drain into nearby water bodies after converging in the pipe network, and rainwater pumping stations need to be installed. In some cases, rainwater pumping stations have become a limiting factor for rainwater removal in cities due to their flow limitations and damage to pumps. They are also a major risk factor for waterlogging.

2 Related work

The urban drainage pipe network system involves a variety of structures and ancillary structures such as urban watersheds, water intakes, underground drainage pipe networks, natural drainage networks, land distribution, streets, road networks, terrain, and hydraulic facilities [15 –17]. It is a complex complex. Urban rainwater systems usually include house rainwater piping systems, outdoor rainwater ditch systems, rainwater pumping stations, pressure pipes, drainage ditches and stormwater outlets, but not necessarily all of them for each specific rainwater system [18].

The model needs to establish rainfall events to provide support for simulated rainfall. There are usually two methods: measured and synthetic rainfall. The input measured rainfall event can be one rainfall or multiple rainfalls over a period of time. Models are usually calibrated using measured rainfall data and based on flow monitoring data and historical flooding records [19, 20]. Synthetic rainfall events, also called design rainfall events, are statistical rainfall events that establish a clear rainfall time and recurrence period based on analysis of years of rainfall records. Synthetic rainfall events are often used for analysis and design of models that have already been calibrated, which can make the researched rainfall more representative [21, 22].

Info Works software uses the intensity (or depth) of a duration-frequency relationship to establish an artificial design storm, where the reciprocal of the frequency is the so-called recurrence period. Intensity is the opposite of diachronic growth, and decreases with time [23, 24]. At present, China has set up a flood control engineering system at the basin level to prevent passenger water from entering the city. There is also a urban pipe network drainage system, but there is no consideration of storm runoff that exceeds the drainage capacity of the rainwater pipe network. From the perspective of drainage planning system, flood prevention planning mainly focuses on river basins and rivers, and the drainage involved mainly focuses on regional flood disasters rather than urban waterlogging disasters; drainage pipe network planning focuses more on drainage facilities such as pipes and pumping stations [25]. The layout and size were determined.

The lack of connection between the two systems, and the lack of systematic planning for the connection of pipelines and rivers. Especially for the rainwater of the super pipe network design standard, there is neither a unified design method nor related technical standards and specifications, and of course there is no corresponding engineering facility and comprehensive means [26, 27]. The existing drainage planning system cannot meet the rapid development of the city. The planning and design standards are not perfect. The current drainage planning norms and design standards are mostly for rainwater facilities such as drainage pipes and pumping stations, but there are no clear requirements and technical standards for urban waterlogging prevention and control, and response to excessive rainwater. The technical standards of rainwater pollution and comprehensive utilization of rainwater in the early stages of the city are not complete [28]. As far as the urban rainwater pipe network system is concerned, due to historical reasons, the design recurrence period has been relatively low [29, 30]. Most of the urban rainwater drainage projects previously constructed are 1 year or less than the design standard of the 1 year recurrence period, even 1 year and 2 The design standards for encounters, three encounters in one year, and four encounters in one year are much lower than those of general developed countries. The planning and design methods need to be updated urgently. For a long time, China has continued to use traditional reasoning formulas to calculate rainwater flow to guide drainage network design. In fact, this method is only suitable for use in areas with a small confluence area, but there are no strict definitions of restrictions when using it [31, 32]. With the rapid expansion of the city scale, the rainfall characteristics of the city have undergone major changes, and the traditional reasoning formula method has been used in general for the rainwater system design to have certain deviations from the actual situation; in addition, the reasoning formula method cannot reflect the management. The actual situation of the operation of the Internet; In addition, the formula of heavy rain intensity in many cities has not been revised and updated for many years, and it has long been unsuited to the new situation changes [33].

At present, digital simulation technology has been applied in drainage planning and design, but only a few areas have been tried. The lack of corresponding professional technical reserves, especially the inaccuracy and incompleteness of basic data, has restricted the application of drainage system simulation technology. In addition, the research on rainwater simulation models in Chinese cities started late, and there is a big gap compared with other developed countries and regions. At present, there is no universally developed mature model, and foreign models are mainly used for calculation. Chinese cities have not yet established a rainwater impact assessment and waterlogging risk assessment system. Existing flood disaster risk distribution maps are mainly targeted at floods or tidal floods in the outer rivers, and there are few evaluations for the occurrence of internal waterlogging in cities. The planning process lacks quantitative analysis methods, and it is even more difficult to manage the risk of floods.

3 The proposed method

3.1 Confluence mode

The confluence model determines how quickly rainfall falls from the catchment into the drainage system. According to the principle of the model calculation engine, the confluence model can be divided into a bilinear reservoir model, a unilinear reservoir model, and a non-linear reservoir model. This is also the distinction method in hydrology. The unilinear reservoir model was proposed by Visa, and a conceptual reservoir was used to simulate the process of water storage and discharge. The water storage capacity S of the reservoir is: $S = KQ$ (1)

Continuity equation: $I - Q = \frac{dS}{dt}$ (2)

Assume that the effective rainfall stops at t1 after the start of the runoff, and the outflow of f > f1 is: $Q = Q_{0} e^{(t_{1} - t) / K}$ (3)

In the same way, the bilinear reservoir method uses two consecutive hypothetical linear reservoirs with the same reserves and first-order relationship to represent the storage capacity of the ground and drainage channels and the delay between peak rainfall and runoff. This method allows peak runoff to occur after peak rainfall. The amount of rainwater S in the subcatchment area is the product of the amount of rainwater discharged into the pipe network and the time unit coefficient K. Runoff coefficient is related to rainfall intensity, catchment area and slope. Six types of convergence models, as shown in Table 1:

Table 1

Comparison of convergence models

Model	Introduction	Scope of application
Wallingford Confluence Model	Based on the bilinear reservoir model, the confluence parameters depend on rainfall intensity, contribution area, and slope.	British drainage system with subcatchment less than 1 ha
Large Contribution Area Convergence Model	Based on the bilinear reservoir model, taking into account the flow characteristics of the catchment area and assuming the detention of the pipeline, the convergence model was modified using the convergence coefficient multiplier and runoff time lag factor to delay the outflow peak.	Convergence calculation for large contribution area below 100 tea
SPRINT Confluence Model	Based on a single linear reservoir model, developed for the European SPRINT project, the equation is independent of rainfall intensity.	Confluence calculation for large lumped catchment area
Desbordes Confluence Model	The French standard confluence model is based on a single linear reservoir model. It is assumed that the outlet flow of the catchment area is proportional to the rainwater volume of the catchment area, and the catchment area width needs to be defined.	Continuous simulation
SWMM Convergence Model	The non-linear reservoir model developed in the United States uses the non-linear reservoir and the motion wave equation to calculate the slope flow. It is necessary to define the width of the subcatchment and the surface Manning roughness coefficient.	Usually used with Harton or Greek-Ampt permeable area volume models
Unit Water Level Convergence Model	The user determines the parameters to determine the unit water level of the catchment area, which is then used to calibrate the runoff caused by the net rainfall from the catchment area.	Reliance on selection

Rainfall runoff enters the rainwater pipeline and begins the process of pipe network confluence. Info Works CS software requires predefined parameters before hydraulic calculations. For closed pipes or open channels, pre-defined cross-sectional shapes are needed, such as the diameter of a round pipe, the height of an open channel to define the lining of a channel, and the hydraulic roughness coefficient of a predefined channel. White et al. Calculated the hydraulic roughness. Urban rainwater management system as shows in Fig. 1.

Fig. 1

Urban rainwater management system.

The model uses fully solved Saint-Venant equations, which are dynamic wave methods, to simulate open channel flow in pipelines. The Saint Venant equations are a set of mass conservation and momentum conservation equations, that is:

$\frac{δ A}{δ t} + \frac{δ Q}{δ x} = 0$ (4) $\frac{δ Q}{δ t} + \frac{δ}{δ x} (\frac{Q^{2}}{A}) + gA (cos \frac{δ y}{δ x} - S_{0} + \frac{Q | Q |}{K^{2}}) = 0$ (5)

3.2 Neural algorithm improvement

Human brain is an extremely complex and huge system, and also a system with perfect functions. There are always two main lines in the study of machine simulation of human brain thinking activity, that is, symbolism and connectionism. Neural network is a representative of connectionism. It is widely connected by a large number of processing units with simple functions - artificial neurons. According to certain rules or algorithms, according to the external conditions or specified goals, it can achieve the purpose of memory, processing, adaptation and even learning information by constantly changing the connection strength weight between neurons. BP network as shows in Fig. 2.

Fig. 2

BP network.

In the decision-making process, it is necessary to have an accurate and timely judgment on the impact of current rainfall. However, due to the characteristics of flood calculation, it is difficult to get relevant timely information. First of all, there are many factors involved in flood calculation, which leads to a very complex impact of various factors on the results. It takes several hours or even longer to use the existing model for simulation calculation. Secondly, the duration of a rainstorm, especially a rainstorm with great impact on flood control, may be only a few hours, and the best decision-making time is fleeting. Therefore, it is very important to predict the future flood quickly and accurately according to the real-time observation data.

(1) Calculating the forward propagation of BP networks.

${net}_{ij} = \sum_{k = 1}^{N_{i - 1}} O_{(i - 1) k} K_{(i - 1) kj}$ (6) $O_{ij} = f_{s} ({net}_{ij}) = \frac{1}{1 + exp [- ({net}_{ij} - θ_{ij})]}$ (7)

(2) Derivation of error back propagation algorithm.

First, the error is defined: $e_{j} = d_{j} - y_{j}$ (8)

The structure of artificial neural network is a new information processing system which imitates the structure and function of brain cells, the structure of brain nerves and the brain function of thinking processing. The structure and working mechanism of artificial neural network are basically based on the organization structure (brain neural network) and activity law of human brain. It reflects some basic characteristics of human brain, but it is not to reproduce the real part of human brain, so it can be said that it is some abstract, simplified or imitated. $E = \frac{1}{2} \sum_{J} {(d_{j} - y_{j})}^{2}$ (9)

The main features are nondifferentiable and step type, which are often used in cellular neural networks, such as pattern recognition, character recognition or noise control. None of the above includes the delay effect and the refractory period of the neurons, that is, the output only corresponds to the current input, and the delay feature will be realized by adding a delay unit in the future. $Δ W_{ijk} = - η \frac{\partial E}{\partial W_{ijk}}$ (10)

The recursive relationship between ΔW_ijk and neuron output: $\begin{matrix} Δ W_{ijk} = - η \frac{\partial E}{\partial W_{ijk}} = - η \frac{\partial E}{\partial {net}_{(i + 1) k}} \cdot \\ \frac{\partial {net}_{(i + 1) k}}{\partial W_{ijk}} = η δ_{ik} \cdot \frac{\partial {net}_{(i + 1) k}}{\partial W_{ijk}} \end{matrix}$ (11)

Among them, $δ_{ik} = - \frac{\partial E}{\partial {net}_{(i + 1) k}}$ $\frac{\partial {net}_{(i + 1) k}}{\partial W_{ijk}} = \frac{\partial net}{\partial W_{ijk}} (\sum_{h = 1}^{N_{i}} o_{ih} W_{ihk}) = O_{ij}$ (12) $Δ W_{ijk} = - η δ_{ik} O_{ij} (8)$ $δ_{ik} = - \frac{\partial E}{\partial O_{(i + 1) k}} \frac{\partial O_{(i + 1) k}}{\partial {net}_{(i + 1) k}}$ (13) $\frac{\partial O_{(i + 1) k}}{\partial {net}_{(i + 1) k}} = f^{'} ({net}_{(i + 1) k})$ $f^{'} (x) = \frac{e^{- (x - θ)}}{{[1 + e^{- (x - θ)}]}^{2}} = f (x) [1 - f (x)]$ $\begin{matrix} \frac{\partial O_{(i + 1) k}}{\partial {net}_{(i + 1) k}} = f^{'} ({net}_{(i + 1) k}) = f ({net}_{(i + 1) k}) \\ (1 - f) = o_{(i + 1) k} (1 - o_{(i + 1) k}) \end{matrix}$ (14) $\frac{\partial E}{\partial O_{(i + 1) k}}$ The output node is o_(i+1)k $\frac{\partial E}{\partial O_{(i + 1) k}} = y_{k} - d_{k}$ (15) $\begin{matrix} δ_{ik} = - \frac{\partial E}{\partial {net}_{(i + 1) k}} = - \frac{\partial E}{\partial O_{(i + 1) k}} = \\ (d_{k} - y_{k}) o_{(i + 1) k} (1 - o_{(i + 1) k}) \end{matrix}$ (16)

Hidden layer node is o_(i+1)k $\begin{matrix} \frac{\partial E}{\partial O_{(i + 1) k}} = \sum_{h = 1}^{N_{i + 2}} \frac{\partial E}{\partial {net}_{(i + 2) h}} \cdot \frac{\partial {net}_{(i + 2) h}}{\partial O_{(i + 1) k}} \\ = - \sum_{h = 1}^{N_{i + 2}} δ_{(i + 1) h} w_{(i + 1) kh} \end{matrix}$ (17) ${net}_{ij} = - \sum_{k = 1}^{N_{i - 1}} o_{(i - 1) k \cdot} w_{(i - 1) kj}$ (18)

As far as the existing typical neural network models are concerned, although many details of biological neural network have been omitted, they fully retain the basic structure of brain neural network system and partly reflect the internal mechanism of biological neural system. According to the different topological structure of the connection between neurons, the neural network structure can be divided into two main categories, namely hierarchical network and interconnection network. $δ_{ik} = O_{(i + 1) k} (1 - O_{(i + 1) k}) \sum_{h = 1}^{N_{i + 2}} δ_{(i + 1) h} w_{(i + 1) kh}$ (19)

The weight formula of the BP algorithm is adjusted to: $Δ W_{ijk} = 〈 \begin{matrix} η (d_{k} - y_{k}) y_{k} (1 - y_{k}) o_{ij} \\ _{η o (i + 1) k} (1 - o_{(i + 1) k}) \cdot \sum_{h = 1}^{N_{i + 2}} δ_{(i + 1) h} w_{(i + 1) kh} o_{ij} \end{matrix}$ (20)i-th layer neuron is $δ_{ik} = 〈 \begin{matrix} (d_{k} - y_{k}) y_{k} (1 - y_{k}) \\ _{o (i + 1) k} (1 - o_{(i + 1) k}) | \cdot \sum_{h = 1}^{N_{i + 2}} δ_{(i + 1) h} w_{(i + 1) kh} \end{matrix}$ (21)

For a simple forward network, given a certain input mode, the network can produce a corresponding output mode and keep it unchanged. As the most basic working unit of neural network, the structure of neuron is very simple, and its processing ability is relatively simple. However, the neural network composed of a large number of neurons with simple structure and function has many excellent characteristics. $Δ W (t) = η δ O + α Δ W (t - 1)$ (22)

The processing of information by neural network is a kind of collective function, which is completed by a large number of neurons. What is suitable for it is the distributed storage of information and the access mode of associative memory. The speed and powerful learning and memory function of neural network for information processing are determined by its large-scale parallel working mode, nonlinear processing, variability of network structure and other inherent structural characteristics. $w (t + 1) = w (t) + Δ w (t) sgn (\frac{\partial E_{T}}{\partial w_{t}})$ (23)

Learning is the instinct of the nervous system. Imitating the learning process of human beings, people put forward a variety of learning methods of neural network, including three kinds: learning with tutor, learning without tutor and reinforcement learning. When the neural network model is classified according to the learning mode, it can be divided into three kinds: the learning network with tutor, the learning network without tutor and the reinforcement learning network. Mentored learning (or supervised learning) is conducted under the guidance and investigation. If the learning fails to meet the requirements, then it is necessary to continue learning (relearning). Unsupervised learning (or unsupervised learning) is done by learners or the nervous system itself. For example, if someone is interested in biological neurons, he will find many books to learn. No one supervises this kind of learning. The degree of learning depends on the ability of neural network in the brain (including his own interest, because interest is also a kind of neural reaction). Finally, he can master this kind of knowledge to a certain extent. Learning is a relatively long-lasting process of change. Learning is often a process of reasoning. For example, learning can also be achieved through experience. Learning is the most important ability of neural networks.

3.3 Waterlogging reason analysis

This article considers that climate factors, ground factors, and human activities are the three major causes of flooding. Heavy rain is the direct cause of urban waterlogging. From a meteorological point of view, under the combined influence of the seasonal migration of the south and north branches of the westerly current over eastern China, the seasonal migration of the subtropical high pressure zone, and the differences in the thermal properties of the East Asian continent and ocean waters, China The annual and monthly precipitation varies greatly during the year, and the summer precipitation is dominated by heavy rain.

The terrain affects the monsoon circulation, increases the probability of precipitation, and affects the distribution of heavy rain. The absolute height of the area and the slope of the ground affect the confluence. Due to the special terrain caused by natural reasons, some areas are more prone to urban flooding problems. For example, Tianjin is located in the northern part of the North China Plain, facing the Bohai Sea in the east, with low terrain and high groundwater levels. Urbanization brings the continuous increase of hardened pavement and building area, which increases the runoff coefficient, increases surface runoff, increases the output flow, shortens the confluence time, easily accumulates water on urban roads and low-lying areas, and forms a flood disaster. Change is the root cause of flood disasters. Table 2 shows the amount of water lost after rainfall on different surfaces, and Table 3 compares the rainfall runoff rates of different materials.

Table 2
Water loss after rainfall on different surfaces

Surface type New asphalt pavement Old asphalt road Concrete tile Red brick Land surface Grassland

Water loss /mm 3.1 6.2 10.1 15.9 19.7 37.5

Surface type	New asphalt pavement	Old asphalt road	Concrete tile	Red brick	Land surface	Grassland
Water loss /mm	3.1	6.2	10.1	15.9	19.7	37.5

Table 3

Comparison table of rainfall runoff rate of different materials

Underlying condition	Impervious area ratio (%)	Runoff rate (%)
Green space, permeable	0∼10	60
concrete brick
Sod brick paving	75	80
Glazed tile paving	85∼100	90∼100

Ground subsidence is a kind of slow-moving geological disaster caused by the compression of the earth’s surface due to the compression of the earth’s crust under the influence of natural and man-made factors. It is a kind of gradual geological disaster, which is an irreparable permanent environment and resource loss. The consequences of destruction. In these low-lying areas, the pumping station pipe network and the drainage capacity of the pumping station are restricted. Some low-lying transportation facilities, such as highway bridges and culverts, experience long periods of flooding and traffic disruption during heavy rains.

With the development of the city, although the drainage network system and pumping stations in the new and old districts have been transformed, they still cannot meet the needs of urban development. The number of drainage facilities is insufficient due to land factors and the layout is unreasonable; the pumping station facilities are obsolete due to financial factors and are operating with illness. Except for newly built rainwater drainage pipe networks, most of the old urban areas are rain and sewage confluence systems. Due to the aging of the pipe network, the sedimentation of the sediment reduces the drainage capacity. When the rainfall intensity is high, it will often result in poor drainage of rainwater and stoppage of traffic, which will seriously affect the production and life of residents.

4 Simulation

The risk evaluation considers the occurrence probability of the risk event and the comprehensive consequences of the loss, and involves the combined effects of various factors and various types of risk events. The evaluation is completed in three steps by determining evaluation indicators, evaluation standards, and comparing at specific risk levels. The evaluation indicators include risk rate, risk loss, risk amount, etc., and the forms are different depending on the application field. Evaluation criteria are usually given in combination with evaluation methods, and are mostly related to subjective factors. Risk assessments need to be compared at certain risk levels, such as a single risk level or an overall risk level.

The sample data is calculated by the Hydras model and SSFM model. In order to ensure that all variables receive equal attention in the training process and prevent some neurons from reaching the oversaturated state, it is better to convert the data in the sample into a unified scale (normalization) with the excitation function of the output layer. If it is not normalized, a very small weight factor may be needed to input a large flow value into ANN, which results in floating-point error and training time growth. $\frac{| x_{p} - x | 2}{\sum_{i = 1}^{n} {| x_{p} - x_{i} |}^{2}} \times 100$ (24)

When calculating the effect of rainwater utilization measures on urban flood reduction, due to the lack of design data of the actual reservoir, when dealing with the effect of the reservoir, the rainfall data is directly reduced, so the treatment in light rain is more practical. When the intensity of a single rainfall is large, not only the capacity of the reservoir, but also the design of the reservoir should be considered. For example, when the rain is strong at the water inlet of the reservoir, if the water flow capacity of the water inlet is exceeded, the water will be discarded when the reservoir is not full. In the future, when the data of the reservoir is complete, the design of the reservoir can be considered comprehensively to make the calculation more practical. Error comparison chart of algorithm operation as show in Fig. 3. Algorithm operation error as show in Table 4.

Fig. 3

Error comparison chart of algorithm operation.

Table 4

Comparison table of algorithm operation error

	Improve algorithm	Traditional algorithm
1	11.7	14.6
2	15.3	13.7
3	21.2	18.6
4	14.6	11.8
5	17.9	21.0
6	31.3	15.2
7	9.7	11.3
8	8.1	10.9
9	5.4	8.7
10	31.5	17.6
11	18.5	11.3
12	16.8	15.3

The urban water problem includes two aspects: on the one hand, with the development of the city, the demand for water resources is growing, and the existing water resources can not meet the requirements. In many places, water shortage has become the bottleneck of urban development. On the other hand, with the development of urbanization, the underlying surface of the city has changed a lot, the hardened ground increases, and the runoff coefficient increases. Urban flood has also become a problem affecting people’s lives. Rainwater utilization is a very effective way to solve the contradiction. Through various rainwater utilization measures, not only can we use a large number of rainwater resources to alleviate the problem of water shortage in the city, but also effectively reduce the flood peak of the city flood and alleviate the situation of surface water accumulation. By comparing the training time of 12 items of data, the results are shown in Table 5 and Fig. 4.

Table 5

Comparison table of algorithm training time

	Improved algorithm training time	Traditional algorithm training time
1	0.00371	0.00632
2	0.00452	0.00842
3	0.00476	0.00541
4	0.00441	0.01187
5	0.00542	0.00672
6	0.00540	0.01237
7	0.00514	0.00687
8	0.00671	0.01127
9	0.00601	0.00846
10	0.00578	0.01037
11	0.00594	0.00806
12	0.00612	0.01048

Fig. 4

Statistical chart of algorithm training time.

As shown in Fig. 4 and Table 5, the improved algorithm can be stably controlled within 0.007 s in training time. The amount of rainwater that can be used in different water years can provide reference for the utilization and regulation of rainwater resources. At the same time, the influence of rainwater utilization measures on urban flood control, especially the reduction of urban flood peak and urban water accumulation, is also studied To study the utilization of rainwater in an area, we should first understand the rainwater utilization potential of the area, and then study the available rainwater under the existing rainwater utilization measures, and what percentage of the rainwater utilization potential is. Only knowing these contents can we have a comprehensive evaluation of the rainwater utilization in an area. Generally speaking, in the future, the research on urban hydrological model will gradually combine the actual situation of urban underlying surface and drainage system more closely, and consider the comprehensive utilization measures of rainwater in urban areas and comprehensive regulation measures of rivers and lakes, the model will be more accurate and reliable. The prediction results is shown in Fig. 5.

Fig. 5

Statistical chart of prediction results.

The rainfall time distribution in China is very uneven, mainly concentrated in the flood season. During this period, the number of rainfall is relatively intensive, and the single rainfall is relatively large. Due to the limitation of the reservoir capacity, it is impossible to collect all the incoming water from the catchment area by using the reservoir, so a certain amount of waste water is needed. In the non flood season, the rainfall interval is long, and the single rainfall is also small, all the incoming water can be collected and utilized. When calculating the amount of urban rain water conservancy, calculate the situation of flood season and non flood season respectively. Finally, sum the amount of rain water conservancy in flood season and non flood season to get the amount of rain water conservancy in the whole year.

As shown in Fig. 5, the traditional algorithm is 50% in the prediction of the result, and the prediction result has a certain randomness. However, the accuracy of the improved algorithm. Because the phenomenon of urban waterlogging is vague, it is affected by a variety of factors and requires comprehensive evaluation. Therefore, the fuzzy comprehensive evaluation method is very suitable for solving the risk evaluation problem of urban waterlogging.

It can be seen from the calculation results that, with the increase of daily water quota, that is, the water use rate of the reservoir is accelerated. In the flood season with concentrated rainfall, the abandoned water is gradually reduced, and the annual rainwater utilization rate of the reservoir is increased. After the water quota of the day reaches a certain degree, the rainwater utilization of the reservoir basically reaches a stable state. At this time, there is no waste water and all possible rainwater can be collected.

5 Planning for urban waterlogging disasters

The prevention and control of urban waterlogging is a complex systematic project, and it is not possible to consider the problem in isolation, and to discuss drainage on drainage. The overall, systematic, coordinated, and comprehensive nature of the plan itself determines that the drainage and waterlogging prevention plan is a comprehensive and strong plan.

(1) Reasonable drainage area

Cities are often divided into several drainage zones by natural elements such as water systems and mountains. This is a state of natural formation. However, for development and construction, the original terrain and height difference are often artificially changed to destroy the natural zones. Rainstorms cannot remove rainwater in a timely manner. This is caused by the unreasonable drainage zoning, so the planning must divide the drainage zoning according to the drainage conditions of each area.

(2) Scientifically determine the system plan

The formulation of drainage and waterlogging prevention planning shall be based on the local economic and social development situation, climatic characteristics, receiving water bodies, meteorological and hydrological conditions, and topographic conditions. On the basis of full research and analysis, a systematic planning plan and urban construction land layout suggestions are proposed, and the system construction method is comprehensively determined. It is necessary to rationally arrange the drainage channels for rain and flood, optimize the system scheme, and select a storage method.

(3) Do a good job of connecting large and small drainage systems

The formulation of drainage and waterlogging prevention planning must first ensure the rationality of its own systems such as pipe networks and rivers and lakes. On this basis, it is also necessary to connect pipes, rivers, and water systems, especially with urban flood control systems. It is necessary to pay attention to the process of river level change, and calculate the drainage capacity based on the cross section of the river and the existing and planned control facilities. On this basis, engineering measures to deal with urban waterlogging are proposed.

(4) Do a good job of linking drainage planning and other planning

Drainage and waterlogging prevention planning involves multiple professional planning. Therefore, the relationship between urban drainage planning and land use planning should be coordinated and multi-professional connections should be made. The layout of urban land and road planning should consider the way out for rainwater discharge; vertical urban design and vertical design of roads must ensure the smooth drainage channels and comprehensive utilization of rainwater; urban green space planning should consider accepting nearby rainwater. Multi-disciplinary coordination can make planning scientific and reasonable.

In order to improve the scientificity of drainage and flood prevention planning and the level of urban risk management, we should learn from advanced foreign experience and gradually establish a rainwater impact assessment and flood risk assessment system in China, comprehensively assess the current status of urban drainage and flood prevention facilities and flood risk, and map cities Map of risk of rainstorm and waterlogging, determine the risk level, scientifically delineate drainage and waterlogging prevention zones and risk management areas, and provide effective technical support for the formulation of drainage and rainstorm waterlogging prevention programs and emergency management.

6 Conclusion

This paper introduces the theory and method of risk analysis into urban waterlogging problems, and proposes a set of urban waterlogging risk analysis methods that are applicable to rainwater systems. The risk identification map of waterlogging risks is given, and the risk research for rainwater systems is given. Processes and methods lay the theoretical foundation for further risk analysis. Areas with conditions should gradually promote the application of mathematical model methods in planning. In the future rainwater planning design and management, the reference, absorption, digestion, and development of hydrological and hydraulic models are indispensable technical links, so some of them should be developed as soon as possible. The simple, feasible, and practical model software with independent intellectual property rights is convenient for everyone to use and can reduce costs. At the same time, we must increase the training of technical personnel and improve the application level of rainwater models to improve the scientificity and operability of planning.

Footnotes

Acknowledgments

This paper was supported by (1) Project funded by China Postdoctoral Science Foundation; (2)Project funded by the Project(017/2018/A) of FDCT; (3)Project funded by the Project of Macao Foundation.

References

Unterthiner

, Mayr

, Klambauer

, et al., Toxicity Prediction using Deep Learning Computer Science 3(8) (2015), 23–34.

Spencer

, Eickholt

and Cheng

, A Deep Learning Network Approach to ab initio Protein Secondary Structure Prediction IEEE/ACM transactions on computational biology and bioinformatics / IEEE, ACM PP(99) (2014), 102–122.

, Yu

, Wang

, et al., Large-Scale Transportation Network Congestion Evolution Prediction Using Deep Learning Theory, Plos One 10 (2015), 45–60.

Sheng

, Siqi

, Zhen

, et al., Accurate De Novo Prediction of Protein Contact Map by Ultra-Deep Learning Model, PLOS Computational Biology 13(1) (2017), e1005324–1005335.

Angermueller

, Lee

H.J.

, Reik

, et al., Deep CpG: accurate prediction of single-cell DNA methylation states using deep learning, Genome Biology 18(1) (2017), 67–73.

Shao

, Fu

, Li

, Altan

and Cheng

, Remote sensing monitoring of multi-scale watersheds impermeability for urban hydrological evaluation, Remote Sensing of Environment 232 (2019), 111–338.

, Wang

, He

and Shao

, User Identification across Asynchronous Mobility Trajectories, Sensors 19(9) (2019), 2102–2111.

Shao

Zhenfeng

, Wu

Wenjing

, Wang

Zhongyuan

, Du

Wan

, Li

Chengyuan

, SeaShips: A Large-Scale Precisely Anno-tated Dataset for Ship Detection, IEEE Transactions on Multimedia 20(10) (2018), 2593–2604.

Shao

, Yang

and Zhou

, Performance Evaluation of Single-Label and Multi-Label Remote Sensing Image Retrieval Using a Dense Labeling Dataset, Remote Sensing 10(6) (2018), 964–977.

10.

Shao

and Cai

, Remote Sensing Image Fusion with Deep Convolutional Neural Network, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 11(5) (2018), 1656–1669.

11.

Zhang

, Weng

and Shao

, An evaluation of monthly impervious surface dynamics by fusing Landsat and MODIS time series in the Pearl River Delta, China, from 2000 to 2015, Remote Sensing of Environment 201 (2017), 99–114.

12.

Kiral-Kornek

, Roy

, Nurse

, et al., Epileptic Seizure Prediction Using Big Data and Deep Learning: Toward a Mobile System, E Bio Medicine 27(C) (2017), 103–111.

13.

Lao

, Chen

, Li

Z.C.

, et al., A Deep Learning-Based Radiomics Model for Prediction of Survival in Glioblastoma Multiforme Scientific Reports, 7(1) (2017), 10353.

14.

, Shao

, Fu

and Cheng

, The Dynamic Analysis between Urban Nighttime Economy and Urbanization Using the DMSP/OLS Nighttime Light Data in China from 1992 to 2012, Remote Sensing 9(5) (2017), 416–420.

15.

Wang

, Hu

, Jiang

, Han

and Shao

, Heteroskedasticity tuned mixed-norm sparse regularization for face hallucination, Multimedia Tools and Applications 75(24) (2016), 17273–17301.

16.

Ding

, Shao

, Zhang

, Xu

and Wu

, A Comprehensive Evaluation of Urban Sustainable Development in China Based on the TOPSIS-Entropy Method, Sustainability 8(8) (2016), 746–758.

17.

Yin

, Cheng

, Wang

and Shao.

, ‘Big data’ for pedestrian volume: Exploring the use of Google Street View images for pedestrian counts, Applied Geography 63 (2015), 337–345.

18.

Stahl

, Schneider

and Brock

, EPSILON-CP: using deep learning to combine information from multiple sources for protein contact prediction, BMC Bioinformatics 18(1) (2017).

19.

Das

and Ghosh

S.K.

, Deep-STEP: A Deep Learning Approach for Spatiotemporal Prediction of Remote Sensing Data, IEEE Geoscience & Remote Sensing Letters 13(12) (2016), 1984–1988.

20.

Niu

, Zhang

, Shao

and Cheng

, Web-based geological hazard monitoring in the three gorges area of China, Photogrammetric Engineering and Remote Sensing 73(6) (2007), 709–719.

21.

, Yao

, Shao

and Wang

, From digital Earth to smart Earth, Chinese Science Bulletin 59(8) (2014), 722–733.

22.

Shao

and Li

, Image City sharing platform and its typical applications, Science China-Information Sciences 54(8) (2011), 1738–1746.

23.

and Shao

, The new era for geo-information, Science in China Series F-Information Sci-ences 52(7) (2009), 1233–1242.

24.

Polson

N.G.

and Sokolov

V.O.

, Deep learning for short-term traffic flow prediction, Transportation Research Part C: Emerging Technologies 79 (2017), 1–17.

25.

Cheng

, Guo

, Wang

, et al., MiRTDL: a deep learning approach for miRNA target prediction, IEEE/ACM Transactions on Computational Biology & Bioinformatics 13(6) (2016), 1161–1169.

26.

Huang

, Li

, Zhang

, et al., Deep Learning Driven Visual Path Prediction from a Single Image, IEEE Transactions on Image Processing 2016, 1–1.

27.

Wen

, Zhang

, Niu

, et al., Deep-Learning-Based Drug-Target Interaction Prediction Journal of Proteome Research, 16(4) (2017), 1401.

28.

Fan

, Xiao

and Zhao

, A short-term building cooling load prediction method using deep learning algorithms, Applied Energy 195 (2017), 222–233.

29.

Kuo

R.J.

, Hung

S.Y.

and Cheng

W.C.

, Application of an optimization artificial immune network and particle swarm optimization-based fuzzy neural network to an RFID-based positioning system, Information Sciences 262(3) (2014), 78–98.

30.

, Qian

, Li

, et al., Resource allocation based on quantum particle swarm optimization and RBF neural network for overlay cognitive, OFDM System Neurocomputing 2015 S092523121501303X.

31.

, Hu

and Bai

, Generalized radial basis function neural network based on an improved dynamic particle swarm optimization and Ada Boost algorithm, Neurocomputing 152(C) (2015), 305–315.

32.

Ghaedi

, Ghaedi

A.M.

, Ansari

, et al., Artificial neural network and particle swarm optimization for removal of methyl orange by gold nanoparticles loaded on activated carbon and Tamarisk Spectrochimica, Acta Part A: Molecular and Biomolecular Spectroscopy 132 (2014), 639–654.

33.

Khajeh

, Kaykhaii

, Hashemi

S.H.

, et al., Particle swarm optimization–artificial neural network modeling and optimization of leachable zinc from flour samples by miniaturized homogenous liquid–liquid microextraction, Journal of Food Composition and Analysis 33(1) (2014), 32–38.

Risk evaluation of urban rainwater system waterlogging based on neural network and dynamic hydraulic model

Abstract

Keywords

1 Introduction

2 Related work

3 The proposed method

3.1 Confluence mode

Table 2 Water loss after rainfall on different surfaces Surface type New asphalt pavement Old asphalt road Concrete tile Red brick Land surface Grassland Water loss /mm 3.1 6.2 10.1 15.9 19.7 37.5

6 Conclusion

Footnotes

Acknowledgments

References

Table 2
Water loss after rainfall on different surfaces

Surface type New asphalt pavement Old asphalt road Concrete tile Red brick Land surface Grassland

Water loss /mm 3.1 6.2 10.1 15.9 19.7 37.5