CN-121980799-A - Drainage basin ecological hydrologic process simulation and prediction system based on mechanism hydrologic model and AI coupling

CN121980799ACN 121980799 ACN121980799 ACN 121980799ACN-121980799-A

Abstract

The invention discloses a system for simulating and predicting a watershed ecological hydrologic process based on a mechanism hydrologic model and AI coupling, which relates to the technical field of watershed ecological hydrologic simulation and prediction, and comprises the steps of integrating and preprocessing multi-source data through a data input layer, outputting a standardized data set, providing the standardized data set for a model coupling layer, directly receiving the standardized data set of the data input layer by the model coupling layer, outputting an optimized simulation result and a predicted value to a simulation predicting layer, receiving the simulation result and the predicted value of the model coupling layer by the simulation predicting layer, performing multi-dimensional ecological hydrologic element collaborative simulation and short-medium-long-term scene self-adaptive prediction, outputting the simulation predicting result to a result output layer, receiving the simulation predicting result of the simulation predicting layer by the result output layer, and providing a visual interaction and decision suggestion generating function. The invention provides an innovative scheme of a depth fusion mechanism model and an AI technology, which covers the comprehensive requirements of physical rationality, data fitting precision, multi-element coverage and high-efficiency response.

Inventors

ZHANG HANDAN
HU JIANYONG
LU HAO
WANG JUN
Zhang Hancui
HU CHENXIA
SHAN CHENGJU
MO YUCHANG
Ma Delian
WANG HUI

Assignees

浙江水利水电学院

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (6)

1. A system for simulating and predicting a river basin ecological hydrologic process based on a mechanism hydrologic model and AI coupling is characterized by comprising a data input layer, a model coupling layer, a simulation predicting layer and a result output layer, wherein the data input layer is used for carrying out multi-source data integration and preprocessing and outputting a standardized data set to be provided for the model coupling layer, the model coupling layer is used for directly receiving the standardized data set of the data input layer and outputting an optimized simulation result and a predicted value to the simulation predicting layer, the simulation predicting layer is used for receiving the simulation result and the predicted value of the model coupling layer and carrying out multi-dimensional ecological hydrologic element collaborative simulation and short-medium-long-term scene self-adaptive prediction and outputting the simulation predicting result to the result output layer, and the result output layer is used for receiving the simulation predicting result of the simulation predicting layer and providing a visual interaction and decision suggestion generating function.
2. The system for simulating and predicting a drainage basin ecological hydrologic process based on coupling of a mechanism hydrologic model and an AI according to claim 1, wherein the data integrated by the data input layer comprises basic geographic data and multi-source observation data.
3. The system for simulating and predicting a watershed ecological hydrologic process based on coupling of a mechanism hydrologic model and an AI according to claim 2, wherein the data input layer further comprises weather data of sparse sites in multisource observation data based on Python Pandas base for data cleaning, GDAL base for geographic data format conversion and kriging interpolation method for completion, and a standardized data set is output.
4. The system for simulating and predicting the river basin ecological hydrologic process based on the coupling of the mechanism hydrologic model and the AI is characterized in that the model coupling layer comprises a mechanism constraint module, an AI parameter inversion module and a dynamic feedback module, the mechanism hydrologic model of the mechanism constraint module is a SWAT model, standardized data are input, initial difficult-to-observe parameters are set as default values, multidimensional ecological hydrologic intermediate variables are output, the initial difficult-to-observe parameters comprise soil saturated water conductivity, a runoff curve number and a Manning coefficient, the AI model of the AI parameter inversion module adopts an LSTM model combined with an attention mechanism, the standardized data set of the data input layer and the intermediate variables of the mechanism hydrologic model are input, a difficult-to-observe parameter inversion value and a short-term hydrologic element predicted value are output, the dynamic feedback module directly inputs the multi-dimensional ecological hydrologic intermediate variables of the mechanism model and the difficult-to-observe parameter inversion value and the short-term hydrologic element predicted value of the AI model, and when the coupling deviation is more than 5%, the multi-source data are collected again to update parameters of the constraint module and the AI parameter inversion module.
5. The system for simulating and predicting a basin ecological hydrologic process based on coupling of a mechanism hydrologic model and AI of claim 4, wherein the simulated predictive layer comprises: the multi-element collaborative simulation module is used for directly inputting an optimized coupling result of a model coupling layer and a standardized data set of a data input layer, and directly outputting multi-dimensional ecological hydrologic simulation data comprising radial depth space distribution data, actual evapotranspiration data, pollutant migration concentration data and daily water consumption data of different vegetation types; The scene adaptation module outputs a prediction result according to the following logic based on real-time meteorological data: when the precipitation amount is more than or equal to 50mm in 24 hours, an AI model of an AI parameter inversion module of the model coupling layer is applied, an hour-level calculation step length and a 30-day training window are adopted, and a short-term prediction result is directly output; when the precipitation amount is less than 50mm in 24 hours, applying the SWAT model of the AI parameter inversion module of the model coupling layer, adopting a daily calculation step length and a training window of 1 year, and directly outputting a medium-long term prediction result.
6. The system for simulating and predicting the river basin ecological hydrologic process based on the coupling of the mechanism hydrologic model and the AI according to claim 1, wherein the result output layer comprises a visual interaction module, a short-term or medium-long-term prediction result of the simulation prediction layer, multidimensional ecological hydrologic simulation data and actual measurement data of the data input layer are directly input, an interaction interface supporting PC end or mobile end access is directly output, the system has the functions of space-time distribution display, data query and result comparison, and the decision suggestion module directly inputs the short-term or medium-long-term prediction result of the simulation prediction layer and directly outputs a decision report, wherein the decision report comprises a risk early warning and optimizing scheme.

Description

Drainage basin ecological hydrologic process simulation and prediction system based on mechanism hydrologic model and AI coupling Technical Field The invention relates to the technical field of watershed ecological hydrologic simulation and prediction, in particular to a watershed ecological hydrologic process simulation and prediction system based on coupling of a mechanism hydrologic model and an AI. Background The watershed ecological hydrologic process is a complex system of climate-earth surface interaction, and simulation and prediction are core technical bases of water resource management and ecological protection. Currently, main stream technical schemes in the industry are mainly divided into three types, but all the main stream technical schemes have the defect of difficult breakthrough, and the requirements of complex watershed and diversified scenes cannot be met: pure mechanism hydrologic model Represented by SWAT, HEC-HMS and VIC, the core is to construct a mathematical equation based on the laws of physics, describing the internal law of the hydrologic process. Defects: The model has strong parameter dependence, a large number of physical parameters are required to be input, wherein key parameters such as hydraulic characteristics of deep soil are difficult to obtain through field observation, experience estimation or regional mean value assignment is required, so that simulation errors are obvious in complex watercourses, and runoff simulation errors are larger; the calculation efficiency is low, the mechanism model needs to solve a physical equation grid by grid and time period by time period, and the calculation time can be up to hours or even days in a large-river basin or medium-long-term prediction scene, so that the real-time decision requirement can not be met; The adaptability is poor, the response to the sudden change of the underlying surface of the flow field or the extreme hydrologic event is lagged, and the prediction result deviates from the reality because the physical equation is difficult to cover the nonlinear change of the 'non-conventional' scene. (II) pure AI model Represented by LSTM, random forest and BP neural network, the core is to learn the statistical rule of hydrologic process through historical observation data, realize data-driven simulation and prediction. Defects: The black box has obvious characteristics that the model only pays attention to the mapping relation between input and output, the physical meaning of a predicted result cannot be explained, and causal analysis in scientific decision is difficult to support; the generalization capability is weak, the prediction reliability of the scene which is not experienced is low, and samples outside the historical data distribution are easy to cause prediction drift; The data dependence is high, continuous and complete long-sequence observation data is required, the data cannot be effectively deployed in a data scarcity river basin, the data noise is sensitive, and the model precision is seriously affected by a small amount of abnormal values. (III) simple coupling model The current industry coupling is mostly in a unidirectional call mode, for example, an AI model is used for predicting a certain key parameter of a mechanism model and then substituting the key parameter into the mechanism model for calculation, or the AI model is used for correcting the output result of the mechanism model. Defects: the coupling level is shallow, AI and the mechanism model are mutually independent, a dynamic feedback mechanism is not available, complex nonlinear association of hydrologic process cannot be cooperatively dealt with, and the problem of 'physical logic deficiency' or 'data fitting deviation' still exists; most of the functions are only focused on a single hydrologic element, multidimensional ecological hydrologic processes such as precipitation, evaporation and emission, pollutant migration and the like are not integrated, and comprehensive data support for the river basin 'water resource-ecology-environment' collaborative management cannot be provided; The method has poor scene suitability, namely the problem that the model strategy cannot be dynamically adjusted according to the hydrological scene, short-term prediction precision is insufficient or the medium-long term prediction stability is poor is still not solved. In summary, the prior art cannot simultaneously meet the comprehensive requirements of physical rationality, data fitting precision, multi-element coverage and efficient response, and an innovative scheme of a depth fusion mechanism model and an AI technology is needed to break through the current technical bottleneck. Disclosure of Invention In view of the above, the invention provides a drainage basin ecological hydrologic process simulation and prediction system based on a mechanism hydrologic model and AI coupling, which is used for solving the problems in the background technology. In order to