CN-121579941-B - River and lake regulation system flood evolution intelligent forecasting method based on dynamic accumulation and drainage relation

CN121579941BCN 121579941 BCN121579941 BCN 121579941BCN-121579941-B

Abstract

The invention discloses a river and lake regulation and storage system flood evolution intelligent forecasting method based on a dynamic storage and drainage relation, which comprises the steps of constructing a space-time dynamic association diagram structure reflecting the hydraulic connection of a river basin, and mapping multi-source hydrologic data to generate a model input set with uniform space-time scale; the method comprises the steps of establishing a parameterized structure model of an aggregate reservoir, constructing a flood regulating algorithm equation set, adopting an adaptive solving mechanism to switch among different algorithm channels according to the real-time lag state in the composite relationship so as to solve the equation set, correcting the model state by utilizing real-time observation data and executing rolling prediction, wherein the parameterized structure model comprises a dynamic water level-reservoir capacity relationship driven by inflow change characteristics and a composite water level-flow relationship integrating the time lag effect and return water jacking feedback. The method effectively solves the problems that the dynamic characteristics of the accumulation and release relation in the complex river and lake system are difficult to describe, the solution of the non-in-phase evolution process is unstable and the high-dimensional parameters are difficult to set, and improves the continuity and physical consistency of flood forecast.

Inventors

Zhong Pingan
LI WEI
Ben Mengxue
Cai Hehe
WANG BIN
HAN YU
ZHU FEILIN
CHEN JUAN
XU BIN
WAN XINYU

Assignees

河海大学

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (7)

1. A river and lake regulation system flood evolution intelligent forecasting method based on a dynamic accumulation and discharge relation is characterized by comprising the following steps: Constructing a space-time dynamic association diagram structure reflecting the watershed hydraulic connection, mapping the acquired multi-source hydrologic data into the space-time dynamic association diagram structure, and generating a model input data set with uniform space-time scale through node characteristic aggregation and edge weight updating; Extracting inflow change characteristics based on a model input data set, and establishing a polymerization reservoir parameterized structure model representing integral accumulation and drainage characteristics of a main stream and a lake, wherein the polymerization reservoir parameterized structure model comprises a dynamic water level-reservoir capacity relationship driven by the inflow change characteristics in the model input data set and a composite water level-flow relationship integrating a time delay effect and return water jacking feedback; Constructing a flood regulating algorithm equation set based on the aggregation reservoir parameterized structure model, and solving the flood regulating algorithm equation set by adopting a self-adaptive solving mechanism according to the real-time delay state identified from the composite water level-flow relation to obtain algorithm state variables; Performing state correction on the aggregation reservoir parameterized structure model based on real-time observation data, performing rolling prediction by using the corrected model and algorithm state variables, and outputting water level and flow process information in a future period; establishing a polymerization reservoir parameterized structure model for representing the integral storage and release characteristics of a main stream and a lake, wherein the method comprises the following steps of: Dynamically polymerizing the dry-flow river channel and lake storage flood areas by adopting a multi-level equivalent unit method, determining an adaptive polymerization boundary and an equivalent dam site section, and summarizing the water body in the adaptive polymerization boundary into an aggregation reservoir unit; Based on a model input data set, extracting the outlet flow of an upstream water reservoir, the flow of tributary water and the interval confluence flow, and combining the three into a total inflow sequence of the aggregate reservoir; constructing a storage and drainage relationship of the aggregate reservoir unit by taking the aggregate reservoir total inflow sequence as a driving variable; the dynamic water level-reservoir capacity relationship is constructed by the following steps: fitting to generate a water level-reservoir capacity self-adaptive static structure based on historical observation data in a model input data set; Carrying out signal decomposition on the total inflow sequence of the polymerization reservoir to obtain a trend item reflecting inflow ramp characteristics and a disturbance item reflecting inflow mutation characteristics; constructing a time-varying driving factor for controlling the change rate of the reservoir capacity by utilizing a trend term, and constructing a nonlinear enhancement operator for controlling the shape correction of the curve by utilizing a disturbance term; The parameter weight of the water level-reservoir capacity self-adaptive static structure is adjusted in real time by utilizing a time-varying driving factor and a nonlinear enhancement operator, so that a dynamic water level-reservoir capacity relation which dynamically evolves along with the inflow process is generated; The composite water level-flow relationship is constructed by: Extracting water level and flow characteristics from an observation subset of a model input data set, and constructing a water level-flow main control relation based on a power law function through multi-scale nonlinear mapping; On the basis of the water level-flow main control relation, respectively constructing a time delay correction model for representing flood propagation time delay, a jacking feedback model for representing downstream water level to outflow inhibition effect and a dredging compensation model for representing river bed topography change; And superposing the time delay correction model, the jacking feedback model and the flushing compensation model into a water level-flow main control relation through a layered parameterized coupling mechanism, and establishing a composite water level-flow relation interactively driven by an integrated time domain, a space domain and a morphology domain.
2. The method of claim 1, wherein constructing a spatio-temporal dynamic associative map structure reflecting watershed hydropower connections and generating a model input dataset of uniform spatio-temporal dimensions comprises: Performing time synchronization and space registration on the acquired multi-source hydrological data, constructing a multi-mode hydrological time sequence matrix, and establishing an initial graph structure according to the physical communication relation of the river channels; Driving the initial graph structure to perform time-varying evolution by using a multi-mode hydrologic time sequence matrix, and updating edge weights by calculating dynamic similarity among nodes to obtain a space-time dynamic association graph structure; Extracting node embedding characteristics of the space-time dynamic associated graph structure, and executing node missing complementation, anomaly detection and residual feedback correction to obtain an optimized graph structure containing reconstruction characteristics; and converting the optimized graph structure into a high-dimensional embedded matrix, and performing tensor expansion and space-time standardization processing to generate a model input data set.
3. The method of claim 1, wherein the water level-reservoir capacity adaptive static structure is characterized by a third-order polynomial, and the specific calculation expression of the dynamic water level-reservoir capacity relationship is: ; wherein Z is the water level at the current moment t, V (Z) is the dynamic reservoir capacity, a 0 、a 1 、a 2 、a 3 is the static polynomial coefficient; Alpha (t) is a time-varying driving factor, defined as the ratio of the trend term to the historical maximum inflow; B 1 、B 2 、B 3 、B 4 is a nonlinear enhancement operator corresponding to each order item, and is defined as the product of the disturbance item and a preset enhancement operator coefficient.
4. The method of claim 1, wherein the dead time correction model is constructed by: calculating the time change rate of the total inflow sequence of the polymerization reservoir to obtain an inflow lag gradient; nonlinear dynamic amplification is carried out on a preset baseline time delay parameter based on the input flow lag gradient, and a time delay factor is generated; constructing a time sequence residual error feedback channel in a water level-flow main control relation by utilizing a time-varying time factor, and establishing a correction structure for driving the water level-flow main control relation to dynamically perform time delay mapping along with inflow variation; The specific calculation formula of the time-varying time factor is tau (t) =tau 0 *(1+k 1 *dQ in ); Wherein τ (t) is a time-varying factor, τ 0 is a baseline time-varying parameter to be calibrated, k 1 is a time-varying amplification factor to be calibrated, and dQ in is an inflow rate lag gradient; The dynamic delay mapping is implemented by introducing a time offset term of Q (t- τ (t)) in the water level-flow master relation.
5. The method of claim 1, wherein the jacking feedback model is constructed by: Acquiring real-time water level data of an equivalent dam site section, and calculating the water level change rate of the equivalent dam site section along with time; constructing a backwater feedback function for describing the backwater degree of the downstream river by taking the water level change rate as a driving variable; And superposing the backwater feedback function as a negative feedback item to a water level variable of the composite water level-flow relation so as to compensate the jacking influence of downstream water level lifting on the outflow process.
6. The method of claim 4, wherein the adaptive solution mechanism is constructed by: constructing a time delay state discriminator, monitoring a time delay factor in a compound water level-flow relation in real time, and outputting a real time delay state reflecting the current flood response characteristic; establishing a dynamic reservoir capacity synchronous iterative algorithm suitable for a water level flow synchronous response stage and a prediction-correction hybrid algorithm suitable for a water level flow outphasing response stage in parallel to form heterogeneous complementary double-algorithm channels; and automatically switching between the double algorithm channels based on the real-time lag state, and selecting an algorithm path matching the current response characteristic to solve the flood regulating algorithm equation set.
7. The method of claim 6, wherein the adaptive solving mechanism is specifically implemented by: Comparing the time-varying factor with a stable threshold and a fluctuation threshold by using a time-varying state discriminator, judging that the real-time-varying state is a static section when the time-varying factor is lower than the stable threshold, and judging that the real-time-varying state is a dynamic section when the time-varying factor is higher than the fluctuation threshold; When the dynamic reservoir capacity synchronous iteration algorithm is in the static interval, reservoir capacity, water level and flow variable are solved through simultaneous until convergence conditions are met; when the water level is in a dynamic interval, a prediction-correction hybrid algorithm is started, a future water level response trend is predicted by using a historical flow sequence containing time delay information, and correction solution is carried out according to a dynamic water level-reservoir capacity relation and a water quantity balance equation.

Description

River and lake regulation system flood evolution intelligent forecasting method based on dynamic accumulation and drainage relation Technical Field The invention relates to the field of flood forecast, in particular to a river and lake regulation and storage system flood evolution intelligent forecast method based on a dynamic storage and release relation. Background Along with the deep development of large-scale river basin water resources, a complex river and lake combined regulation system is formed by a main river channel and a river-lake, the flood evolution process is subjected to multiple constraints of river channel along-path accumulation, lake diversion and downstream backwater jacking, and the characteristics of high nonlinearity and time variation are presented. Accurately analyzing the water exchange and propagation rule of the system under the dynamic boundary, constructing a high-fidelity flood evolution model, and having important engineering technical value for improving the scientificity and timeliness of flood control scheduling decisions of the river basin. In the prior art, a one-dimensional and two-dimensional hydrodynamic model is generally constructed by adopting a Saint View path group, or generalized simulation is carried out by utilizing a Ma Sijing-root model and other hydrologic methods. The method generally regards river channels and lakes as static units with fixed geometric boundaries, describes the energy regulating capacity according to a preset fixed water level-reservoir capacity curve of topographic data, and characterizes the water level-flow relation of a section based on single valued processing or empirical rope loop curves. In the computational solution layer, a finite difference method or a fixed iterative algorithm is often adopted to conduct time-step deduction on a discretized control equation. However, the conventional scheme has the problem that the storage and release structure characterization rigidifies and the solving mechanism has insufficient adaptability when facing to the scene of severe flood fluctuation or complex hydraulic connection. The method comprises the steps of firstly neglecting the real-time influence of flood wave dynamic effect on the accumulation capacity by a traditional model, enabling a fixed water level-reservoir capacity function not to reflect water level drop adjustment caused by inflow mutation, so that systematic deviation exists in dynamic reservoir capacity calculation, secondly, enabling water level and flow response of a control section to have obvious out-of-phase (hysteresis) characteristics, and enabling the water level and flow response to be influenced by backwater jacking and propagation time delay coupling, wherein stable convergence of a water balance equation is difficult to ensure in a large-amplitude hysteresis interval by a traditional single solving algorithm, and calculation concussion or divergence is easy to generate. Disclosure of Invention The invention aims to provide a river and lake regulation system flood evolution intelligent forecasting method based on a dynamic accumulation and discharge relation, so as to solve at least one of the problems in the prior art. According to one aspect of the application, a river and lake regulation system flood evolution intelligent forecasting method based on a dynamic accumulation and discharge relationship comprises the following steps: Constructing a space-time dynamic association diagram structure reflecting the watershed hydraulic connection, mapping the acquired multi-source hydrologic data into the space-time dynamic association diagram structure, and generating a model input data set with uniform space-time scale through node characteristic aggregation and edge weight updating; Extracting inflow change characteristics based on a model input data set, and establishing a polymerization reservoir parameterized structure model representing integral accumulation and drainage characteristics of a main stream and a lake, wherein the polymerization reservoir parameterized structure model comprises a dynamic water level-reservoir capacity relationship driven by the inflow change characteristics in the model input data set and a composite water level-flow relationship integrating a time delay effect and return water jacking feedback; Constructing a flood regulating algorithm equation set based on the aggregation reservoir parameterized structure model, and solving the flood regulating algorithm equation set by adopting a self-adaptive solving mechanism according to the real-time delay state identified from the composite water level-flow relation to obtain algorithm state variables; And carrying out state correction on the aggregation reservoir parameterized structure model based on real-time observation data, executing rolling prediction by using the corrected model and algorithm state variables, and outputting water level and flow process information of a future period.