CN-121997761-A - Reservoir thermocline characteristic forecasting method considering climate hydrologic condition change

CN121997761ACN 121997761 ACN121997761 ACN 121997761ACN-121997761-A

Abstract

The invention provides a reservoir thermocline characteristic forecasting method considering climate hydrologic condition change, belongs to the technical field of hydrologic simulation and hydraulic engineering, and is suitable for a reservoir area without main tributary to collect. The method comprises the steps of collecting and preprocessing a multi-source data set, constructing a basin yield and confluence model and a vertical plane two-dimensional water temperature model, screening main influencing factors of vertical water temperature and constructing the data set, designing 27 working conditions based on the yield and confluence model, a reservoir water quantity balance equation model and a Copula model, considering weather hydrologic factors and reservoir scheduling coupling relation, generating an expansion data set by utilizing the vertical plane two-dimensional water temperature model, establishing an LSTM-BNN neural network model for vertical water temperature probability prediction, evaluating the performance of the model and identifying the appearance time, the position, the thickness and the strength of a quantized thermocline. The invention realizes effective expansion of the data set through physical reasonable working condition combination, avoids redundancy of the training set, and solves the problem of insufficient generalization capability of the model under the condition of short-series monitoring data.

Inventors

GAO YULEI
GUO YIJING
FAN XIANGJUN
CHEN GEN
ZHANG DIJI
XIA LIMING
MOU XIAOYU
Duan Zhengquan

Assignees

中国长江三峡集团有限公司

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (10)

1. A reservoir thermocline characteristic forecasting method considering climate hydrologic condition change is characterized by comprising the following steps: S1, collecting a physical mechanism model dataset WD, wherein the WD comprises a topographic dataset WDdx, a hydrological element actual measurement dataset WDsw, a meteorological element actual measurement dataset WDqx, a water environment element actual measurement dataset WDhj, a reservoir actual dispatching dataset WDtd and a vertical water temperature actual measurement dataset WDcxsw in front of a dam; S2, preprocessing a sub-data set of a time sequence and a space scatter type in the WD to form an effective WD; S3, constructing a drainage basin yield and confluence model based on drainage basin topographic raster data in WDsw and WDdx; s4, constructing a reservoir area elevation two-dimensional water temperature model based on WDdx, WDsw, WDqx and WDtd; S5, analyzing the correlation between the hydrologic element, the meteorological element, the water environment element, the reservoir scheduling factor and the vertical water temperature, and determining the main influencing factor of the vertical water temperature according to the multi-dimensional screening basis; S6, taking the main influencing factors as input and the preprocessed vertical water temperature in front of the dam as output, and constructing a data driving model data set SD1; s7, dividing the SD1 into a training set SD1-1, a verification set SD1-2 and a test set SD1-3; s8, calculating a daily precipitation sequence JS1, JS2 and JS0 of an extreme precipitation year and a years average precipitation year based on precipitation statistics data; s9, simulating warehousing flow sequences I0, I1 and I2 corresponding to JS0, JS1 and JS2 by using a drainage basin production and confluence model; S10, calculating ex-warehouse flow sequences O0, O1 and O2 corresponding to I0, I1 and I2 based on a water balance equation and a scheduling rule; s11, constructing three groups of daily scale air temperature sequences QW0, QW1 and QW2 based on ERA5 analysis data; s12, adopting a Copula joint distribution model to calculate inflow water temperature sequences SW0, SW1 and SW2 corresponding to QW0, QW1 and QW 2; s13, determining three types of stop-beam door scheduling situations DLM0, DLM1 and DLM2 based on WDtd; S14, constructing a combined simulation working condition of [ JS0-JS1-JS2, QW0-QW1-QW2 and DLM0-DLM1-DLM2 ]; S15, operating the working conditions by using a vertical two-dimensional water temperature model, collecting vertical water temperature data in front of the dam, and finishing to form a data set SD2; s16, establishing an LSTM-BNN neural network model, taking SD1-1 and SD2 as training sets and SD1-2 as verification set optimization parameters, and determining an input time step; S17, inputting the SD1-3 into the optimized model to obtain a forecast result; s18, evaluating the performance of the model by adopting preset indexes; S19, combining the dam front water level and the water inlet sill elevation, and identifying and quantifying the appearance time, the position, the thickness and the strength of the thermocline.
2. The reservoir thermocline characteristic forecasting method considering the climate and hydrologic condition change according to claim 1, wherein the preprocessing of the step S2 comprises the steps of detecting abnormal values by adopting an optimal information bipartition method, a variable bandwidth nuclear density estimation method and a visual inspection method, processing the missing values and the abnormal values by adopting linear interpolation, aligning all time sequence time periods and unifying time scales.
3. The reservoir thermocline characteristic forecasting method considering the climate and hydrologic condition change as claimed in claim 1, wherein the basin yield and confluence model in the step S3 is a three-water source Xinanjiang model, and the model comprises four layers of evaporation and distribution calculation, runoff calculation, water diversion source calculation and confluence calculation.
4. The method for predicting characteristics of a reservoir thermocline in consideration of a change in climatic hydrologic conditions according to claim 1, wherein the elevation two-dimensional water temperature model in step S4 is a CE-QUAL-W2 model, and the basic equations include fluid continuity equation, x and z momentum equation, state equation, water level fluctuation equation, and component migration equation.
5. The reservoir thermocline characteristic forecasting method considering the climate and hydrologic condition change as claimed in claim 1, wherein the screening in the step S5 is based on the physical reasons including the interference degree of the elements on the target elements in the neural network, the data quality, the difficulty in data acquisition, the water temperature layering of the reservoir area and the physical mechanism of heat conduction of the water body of the reservoir area.
6. The reservoir thermocline characteristic prediction method considering climate and hydrologic condition variation as claimed in claim 1, wherein the frequency calculation of extreme water fall years in step S8 adopts the formula p= (m) a)/(n+1 2A) Where a=0, n is the years of precipitation data, with 95% frequencies above rainy year and 5% frequencies below rainy year.
7. The reservoir thermocline characteristic prediction method according to claim 1, wherein in step S11, QW0 is a reference climate state air temperature sequence t_mean, QW1 is a warm phase air temperature sequence t_mean+σ, and QW2 is a cold phase air temperature sequence t_mean Σ, climate benchmark period of the ERA5 analysis data was 1991-2021.
8. The method for predicting characteristics of a thermocline of a reservoir in consideration of a change in climatic hydrologic conditions according to claim 1, wherein the number of combined simulation conditions in step S14 is 3×3×3=27.
9. The method for predicting reservoir thermocline characteristics in consideration of climate and hydrologic conditions according to claim 1, wherein the preset indexes in the step S18 comprise a prediction interval coverage PICP, a prediction interval average width PINAW and a continuous ranking probability score CRPS which characterize probability prediction performance, and an average absolute error MAE which characterizes deterministic prediction performance.
10. The reservoir thermocline characteristic forecasting method considering climate and hydrologic condition changes as claimed in claim 1, wherein in the step S19, the thermocline appearance time is the first appearance date when the daily average temperature difference of the surface water temperature and the water intake bottom plate elevation water temperature is more than 1 ℃ in three days on average, the thermocline position is determined by the upper water depth and the lower water depth of a water layer with the vertical temperature gradient of more than 0.2 ℃ per meter, the thermocline thickness is the difference value of the upper water depth and the lower water depth, and the thermocline strength is the vertical temperature gradient.

Description

Reservoir thermocline characteristic forecasting method considering climate hydrologic condition change Technical Field The invention relates to the technical field of hydrologic simulation and hydraulic engineering, in particular to a reservoir thermocline characteristic forecasting method considering climate hydrologic condition change. Background The water temperature is a key factor affecting the benefits of the aquatic ecosystem, the water body bio-geochemical circulation and the water resource utilization. The water temperature structure of the large deep reservoir is commonly influenced by natural meteorological factors and human activities, and the accurate forecast of the vertical water temperature and thermocline characteristics of the reservoir is important for ecological dispatching. The existing reservoir water temperature forecasting method is divided into a mechanism model and a data driving model. The mechanism model has definite physical mechanism, high calculation cost and high calculation efficiency, is difficult to meet real-time scheduling requirements, is limited by short-series monitoring data, and has insufficient generalization capability. Most of the existing data expansion methods are simple combinations of input elements, the working conditions are huge in quantity and partially deviate from physical reality, and the training set is redundant. Meanwhile, the existing forecast multi-output deterministic water temperature value can not provide uncertainty information and thermocline key characteristic parameters, and cannot meet ecological scheduling decision requirements. Disclosure of Invention The invention provides a reservoir thermocline characteristic forecasting method considering climate hydrologic condition change, which solves the following technical problems in the prior art that under the condition of short series of monitoring data, a forecasting model is difficult to capture water temperature response rules under the coupling effect of different climate hydrologic conditions and reservoir scheduling, the generalization capability is insufficient, the working condition combination design is unreasonable during data expansion, the problems of physical reality separation, huge quantity, training set redundancy and the like exist, the forecasting result is disjointed with the management requirement, and the vertical water temperature probability forecasting information and thermocline key characteristic parameters are lacked. In order to achieve the above purpose, the present invention adopts the following technical scheme: A reservoir thermocline characteristic forecasting method considering climate hydrologic condition change is applicable to a reservoir area without main tributaries, and comprises the following specific steps: S1, collecting a physical mechanism model dataset WD, which comprises a topographic dataset WDdx (reservoir area large section data, river basin topographic raster data), a hydrologic element actual measurement dataset WDsw (daily inflow flow rate, daily inflow temperature, river basin daily scale evaporation data, river basin daily scale precipitation data), a meteorological element actual measurement dataset WDqx (daily scale air temperature, dew point temperature, wind speed, wind direction, cloud quantity, solar radiation), a water environment element actual measurement dataset WDhj (water transparency), a reservoir actual scheduling dataset WDtd (stack door use layer number, stack door top elevation, daily outflow flow rate, daily dam front water level), a dam front vertical water temperature actual measurement dataset WDcxsw, wherein data sources comprise a meteorological bureau, a comprehensive scheduling operation management platform, a hydrologic annual survey, a geospatial data cloud, ERA5 redistribution data and the like. S2, preprocessing a sub-data set of a time sequence and a space scatter type in the WD, namely adopting an optimal information bipartition method, a variable bandwidth kernel density estimation method and a visual inspection method to detect abnormal values, adopting linear interpolation to fill the missing values and correct the abnormal values, and aligning time periods of all the time sequences to be unified into a daily scale time scale to form the effective WD. S3, constructing a three-water source Xinanjiang model based on river basin terrain raster data in WDsw and WDdx, wherein the model is of a dispersive structure and comprises four layers of evaporation calculation (a three-layer evaporation model), runoff calculation (full runoff), water diversion calculation (free reservoir structure divides surface runoff, soil middling and underground runoff), confluence calculation (river basin confluence is carried out by a unit line or linear reservoir method, and river confluence is carried out by Ma Sijing methods and a hysteresis algorithm). S4, based on WDdx, WDsw, WDqx and WDtd, constructing a CE-QUAL-W2 elevati