CN-122021445-A - Urban flood simulation prediction method

CN122021445ACN 122021445 ACN122021445 ACN 122021445ACN-122021445-A

Abstract

The invention discloses a city flood simulation prediction method which comprises the steps of generating rainfall forecast data based on a numerical weather forecast mode, obtaining live rainfall data, predicting future rainfall trend data according to the live rainfall data through a time sequence prediction method, carrying out data assimilation according to the rainfall forecast data, the live rainfall data and the future rainfall trend data through a set Kalman filtering algorithm to obtain assimilated rainfall forecast data, constructing a hydrologic model according to basic geographic data and historical flood data of a forecast area, respectively driving the hydrologic model according to the rainfall forecast data and the assimilated rainfall forecast data, carrying out flood development simulation by adopting a GPU parallel acceleration technology to obtain a simulation forecast result, and repeatedly executing the steps according to updating of the rainfall forecast data to realize rolling simulation forecast. The invention realizes the dynamic rolling simulation prediction of the urban flood evolution process.

Inventors

CHEN GUANGZHAO
HOU JINGMING
MA LIPING
HU YUAN
WANG TIAN
REN CHONG
WANG YANHONG
LV JIAHAO
PAN XINXIN
FAN CHENCHEN
LIU LIJUN

Assignees

西安理工大学
渭南水文水资源勘测中心

Dates

Publication Date: 20260512
Application Date: 20260204

Claims (10)

1. The city flood simulation prediction method is characterized by comprising the following steps of: generating rainfall forecast data based on a numerical weather forecast mode, and acquiring live rainfall data; predicting future rainfall trend data by a time sequence prediction method according to the live rainfall data; carrying out data assimilation by a set Kalman filtering algorithm according to the rainfall forecast data, the live rainfall data and the future rainfall trend data to obtain assimilated rainfall forecast data; constructing a hydrologic hydrodynamic model according to the basic geographic data and the historical flood data of the forecast area; Respectively driving the hydrologic hydrodynamic model according to the rainfall forecast data and the assimilated rainfall forecast data, and carrying out flood evolution simulation by adopting a GPU parallel acceleration technology to obtain a simulation forecast result; And repeatedly executing the steps according to the updating of the rainfall forecast data to realize rolling simulation forecast.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises, The process of generating rainfall forecast data based on the numerical weather forecast model includes: Acquiring the longitude and latitude range of a forecasting area, and extracting lattice rainfall forecasting data from a numerical weather forecasting mode according to the longitude and latitude range; And cutting the lattice point rainfall forecast data to obtain the rainfall forecast data.
3. The method of claim 1, wherein the step of determining the position of the substrate comprises, The process of predicting future rainfall trend data by a time sequence prediction method comprises the following steps: Respectively constructing an autoregressive model, a moving average model and a differential model according to the live rainfall data; constructing an ARIMA prediction model according to the autoregressive model, the moving average model and the differential model; And calculating the future rainfall trend data through the ARIMA prediction model.
4. The method of claim 3, wherein the step of, The process of constructing the autoregressive model includes: introducing a hysteresis operator to perform time sequence analysis on the live rainfall data, and determining an autoregressive order; And calculating an autoregressive coefficient according to the autoregressive order and the live rainfall data, and constructing and obtaining the autoregressive model.
5. The method of claim 1, wherein the step of determining the position of the substrate comprises, The process of data assimilation by the ensemble Kalman filtering algorithm comprises: carrying out Gaussian disturbance on the initial rainfall data sequence to generate a plurality of groups of aggregate data; calculating a state prediction value and a state prediction error covariance matrix according to the set data; calculating a Kalman gain matrix according to the state prediction error covariance matrix and the observation error covariance matrix; And updating a state estimation value according to the Kalman gain matrix, the observation data and the state prediction value to obtain the post-assimilation rainfall forecast data.
6. The method of claim 5, wherein the step of determining the position of the probe is performed, The process of calculating the Kalman gain matrix includes: mapping the state variable to an observation space according to an observation operator; Calculating the product of the state prediction error covariance matrix and the observation operator transposition; calculating a sum matrix of the observation operator, the state prediction error covariance matrix and the observation operator transposition; Inverting the sum matrix to obtain an inverse matrix; And calculating the Kalman gain matrix according to the state prediction error covariance matrix, the observation operator transposition and the inverse matrix.
7. The method of claim 1, wherein the step of determining the position of the substrate comprises, The process of constructing the hydrokinetic model includes: Obtaining topographic data, pipe network data, land utilization data and historical flood actual measurement data of a forecast area; Denoising and filling the topographic data, and correcting a water flow path; performing topology connectivity checking and repairing on the pipe network data; dividing subareas according to the land utilization data, and setting a Manning coefficient and a infiltration parameter; cross-verifying and optimizing model parameters through historical rainfall and flow data to determine a model parameter set; And constructing the hydrohydrodynamic model according to the model parameter set and the preprocessed data.
8. The method of claim 1, wherein the step of determining the position of the substrate comprises, The hydrologic hydrodynamic model adopts a shallow water equation as a control equation, and calculates a water flow state variable according to the topography elevation, rainfall intensity and infiltration rate; Describing soil infiltration characteristics by adopting a Horton infiltration model, and calculating time-varying infiltration quantity according to the initial infiltration rate, the stable infiltration rate and the infiltration attenuation index; describing the water flow of a pipe network by adopting a one-dimensional Saint View south process, and calculating the flow of the pipe according to the water depth, the flow rate, the hydraulic radius and the Manning coefficient; And calculating the water quantity exchange between the surface runoff and the rainwater well by adopting a weir flow or hole flow formula.
9. The method of claim 1, wherein the step of determining the position of the substrate comprises, The process for carrying out flood evolution simulation by adopting the GPU parallel acceleration technology comprises the following steps: Converting the rainfall forecast data and the assimilated rainfall forecast data into a model input format to obtain a rainfall time sequence; inputting the rainfall time sequence to the hydrologic hydrodynamic model; Starting a GPU parallel computing unit to solve a control equation of the hydrologic hydrodynamic model, and simulating the surface flow, confluence and pipe network drainage process; And outputting the simulation forecasting result, wherein the simulation forecasting result comprises water level, flow and submerged range data.
10. The method of claim 1, wherein the step of determining the position of the substrate comprises, The process for realizing rolling simulation forecasting according to the updating of the rainfall forecasting data comprises the following steps: Monitoring the data updating moment of a numerical weather forecast mode; When new rainfall forecast data is detected, re-executing the steps of time sequence prediction, data assimilation, model driving and flood evolution simulation; And outputting the updated simulation forecast result to form a rolling forecast data sequence.

Description

Urban flood simulation prediction method Technical Field The invention belongs to the technical field of urban flood early warning and forecasting, and particularly relates to an urban flood simulation and prediction method. Background The occurrence risk of extreme hydrologic events is obviously increased under the comprehensive influence of global climate change, human activity exacerbation and urban process acceleration, and the occurrence of extreme rainfall and flood disasters in urban areas frequently causes great social and economic damage. The observation data show that the extreme value of the hydrological weather in China continuously breaks through the history record, and the key indexes such as rainfall intensity and accumulated rainfall are particularly outstanding. In a highly urban area, flood disasters have more, frequent, strong and wide situations, and higher requirements are put on urban flood control and disaster reduction work. The existing urban flood simulation forecasting method mainly faces three technical bottlenecks, namely, the forecasting accuracy is insufficient, single-value-dependent weather forecasting mode or traditional observed data are difficult to effectively fuse multi-source data information, rainfall forecasting accuracy is limited, the forecasting period is short, the depth mining and trend extrapolation capability of live observed data are lacked, the existing information cannot be fully utilized to prolong the effective forecasting time, and the calculating efficiency and the forecasting accuracy are difficult to consider, so that the high-accuracy hydrologic hydrodynamic model is large in calculated amount and long in time consumption, and the timeliness requirement of real-time rolling forecasting is difficult to meet. These problems restrict the practical application effect of the urban flood warning system. Disclosure of Invention The invention aims to provide a high-precision simulation prediction method for urban flood, which solves the problems of low prediction accuracy, insufficient prediction period, and difficult compromise between calculation efficiency and prediction precision in the prior art. In order to achieve the above object, the present invention provides a method for simulating and predicting urban flood, comprising: generating rainfall forecast data based on a numerical weather forecast mode, and acquiring live rainfall data; predicting future rainfall trend data by a time sequence prediction method according to the live rainfall data; carrying out data assimilation by a set Kalman filtering algorithm according to the rainfall forecast data, the live rainfall data and the future rainfall trend data to obtain assimilated rainfall forecast data; constructing a hydrologic hydrodynamic model according to the basic geographic data and the historical flood data of the forecast area; Respectively driving the hydrologic hydrodynamic model according to the rainfall forecast data and the assimilated rainfall forecast data, and carrying out flood evolution simulation by adopting a GPU parallel acceleration technology to obtain a simulation forecast result; And repeatedly executing the steps according to the updating of the rainfall forecast data to realize rolling simulation forecast. Preferably, the process of generating rainfall forecast data based on the numerical weather forecast model includes: Acquiring the longitude and latitude range of a forecasting area, and extracting lattice rainfall forecasting data from a numerical weather forecasting mode according to the longitude and latitude range; And cutting the lattice point rainfall forecast data to obtain the rainfall forecast data. Preferably, the process of predicting the future rainfall trend data by the time series prediction method includes: Respectively constructing an autoregressive model, a moving average model and a differential model according to the live rainfall data; constructing an ARIMA prediction model according to the autoregressive model, the moving average model and the differential model; And calculating the future rainfall trend data through the ARIMA prediction model. Preferably, the process of constructing the autoregressive model comprises: introducing a hysteresis operator to perform time sequence analysis on the live rainfall data, and determining an autoregressive order; And calculating an autoregressive coefficient according to the autoregressive order and the live rainfall data, and constructing and obtaining the autoregressive model. Preferably, the process of data assimilation by the ensemble kalman filter algorithm comprises: carrying out Gaussian disturbance on the initial rainfall data sequence to generate a plurality of groups of aggregate data; calculating a state prediction value and a state prediction error covariance matrix according to the set data; calculating a Kalman gain matrix according to the state prediction error covariance matrix and the observation