Dam water temperature response to air temperature change time lag analysis method and device, storage medium

Abstract

This invention discloses a method, apparatus, and storage medium for analyzing the time-delay response of downstream water temperature to air temperature changes. The method includes: first, collecting flow and water temperature data from upstream and downstream hydrological stations, as well as air temperature data from a nearby meteorological station, to establish a basic meteorological and hydrological dataset; second, establishing a heat balance equation for the downstream hydrological control section as a physical constraint condition for a TFT water temperature model; and finally, constructing a TFT water temperature model for the downstream hydrological control section, utilizing the multi-head attention mechanism of the TFT model to analyze the contribution weight of air temperature to the downstream water temperature simulation at different times, and analyzing the daily dynamic process of the time-delay response of downstream water temperature to air temperature changes. This invention overcomes the limitation of traditional cross-correlation analysis, which can only output a single static time delay, and can output daily, dynamically changing time delays, precisely depicting the dynamic evolution of the time delay of downstream water temperature response to air temperature changes over time.

Description

Technical Field

[0001] This invention belongs to the field of ecological and environmental information processing technology, specifically relating to a time-delay analysis method and device for downstream water temperature response to air temperature changes, and a storage medium. Background Technology

[0002] Water temperature is a crucial environmental factor in river ecosystems, influencing the physicochemical properties and biochemical processes of water bodies, as well as the habitat quality of aquatic organisms and the ecological balance of the water body. Due to the water volume and heat regulation effects of reservoirs, "temperature lag" and "cooling lag" occur downstream of the dam, causing downstream water temperature changes to differ from those of natural rivers, exhibiting a certain lag in response to air temperature changes. Furthermore, influenced by the water and heat regulation during different reservoir operation periods and seasonal changes in air-water interface heat exchange, the lag time of downstream water temperature response to air temperature changes—that is, the time lag—is dynamic and not a fixed value. Accurately analyzing the time lag of downstream water temperature response to air temperature changes at different times of the year is key to assessing the ecological and environmental effects of reservoirs, conducting downstream water temperature forecasting, and optimizing reservoir ecological operation.

[0003] There are relatively few methods specifically designed to analyze the time lag in response to changes in air temperature. Traditional time lag analysis methods are mainly based on statistical cross-correlation analysis. This involves calculating the Pearson correlation coefficient between water temperature and air temperature at different lag days, and taking the lag time corresponding to the highest correlation coefficient as the average lag time. However, traditional methods can only provide a uniform average lag number throughout the year and cannot reveal the dynamic changes in the lag time of water temperature response to changes in air temperature across seasons.

[0004] Existing technologies focus more on water temperature forecasting methods, and most of them consider air temperature as an input variable, but they do not adequately consider the time lag of water temperature response to air temperature changes. Among the statistical analysis-based water temperature forecasting methods, an hourly river water temperature forecasting method is proposed (application / patent number: 201811187721.2), which uses a composite cosine function to carry out probabilistic forecasting of hourly river water temperature; and a rolling forecasting method for northern river and canal water temperature that relies only on air temperature and flow rate (application / patent number: 202510338883.5), which establishes a rolling forecasting method for northern river and canal water temperature based on the empirical relationship between water temperature, air temperature, and flow rate. Among the physical mechanism-based water temperature forecasting methods, a joint water-heat balance calculation method (application / patent number: 201910178017.9) was proposed, establishing water balance equations and heat balance equations, defining unit accumulated heat, and using the joint water-heat balance calculation method to analyze reservoir discharge water temperature. A smart winter water temperature forecasting method for open water conveyance channels (application / patent number: 202210432897.X) was developed based on the linear, unidirectional, and sequential flow characteristics of open water conveyance channels. Based on the Lagrange mass tracking method and combined with the gradual heat loss physical process of water, and according to the principles of mass conservation and heat balance, a physical mechanism-based smart water temperature forecasting model was constructed using an initial cross-sectional water temperature + interval air temperature chain model. All of the above methods involve the correlation analysis between water temperature and air temperature, but they fail to reveal the time lag of water temperature response to climate change at different times of the year. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a method, device, system, and storage medium for analyzing the time-delay response of downstream water temperature to air temperature changes. It overcomes the limitation of traditional cross-correlation analysis, which can only output a single static time delay, and instead outputs daily, dynamically changing time delays. This allows for a precise depiction of the dynamic evolution of the time delay in downstream water temperature response to air temperature changes over time, improving the methodology of eco-hydrological time series correlation analysis, enhancing the utilization efficiency of meteorological and hydrological data, and providing technological support for downstream water temperature forecasting and reservoir ecological management. Furthermore, this invention is not limited to analyzing the time delay of downstream water temperature response to air temperature changes; it can be extended to other environmental factors affecting river water temperature, such as solar radiation and wind speed, depending on the actual river conditions. The heat balance equation and TFT water temperature model are adjusted based on the input variables to analyze the annual variation characteristics of the time delay of downstream water temperature in response to changes in other different influencing factors.

[0006] To achieve the above objectives, the present invention provides the following solution: A time-delay analysis method for the response of downstream water temperature to air temperature changes includes: Obtain basic meteorological and hydrological datasets; Establish the heat balance equation for the hydrological control section downstream of the dam; A TFT water temperature model for the downstream hydrological control section is constructed based on basic meteorological and hydrological datasets; the heat balance equation is used as the physical constraint condition for the TFT water temperature model. The contribution weight of air temperature to the simulation of downstream water temperature at different times is analyzed using the multi-head attention mechanism of the TFT model, and the daily dynamic process of the time lag of downstream water temperature response to air temperature changes is analyzed.

[0007] As a preferred option, the meteorological and hydrological basic dataset includes: flow and water temperature data from upstream and downstream hydrological stations of the reservoir, as well as air temperature data from meteorological stations near the downstream hydrological station.

[0008] As a preferred option, based on the principle of heat balance in the downstream water body, the heat balance equation for the hydrological control section downstream of the dam is established: in, This is the correction factor for the impact of water and heat entering the warehouse. The thermal inertia index is the flow rate. The water-air heat exchange coefficient is denoted as ; the downstream water temperature calculated from the heat balance equation is denoted as . This serves as a physical constraint term in the loss function of the TFT water temperature model.

[0009] The present invention also provides a time-delay analysis device for the response of downstream water temperature to air temperature changes, comprising: The first processing module is used to acquire basic meteorological and hydrological datasets; The second processing module is used to establish the heat balance equation of the hydrological control section downstream of the dam. The third processing module is used to construct a TFT water temperature model of the hydrological control section downstream of the dam based on the basic meteorological and hydrological dataset; wherein, the heat balance equation is used as the physical constraint condition of the TFT water temperature model. The fourth processing module is used to analyze the contribution weight of air temperature to the simulation of downstream water temperature at different times using the multi-head attention mechanism of the TFT model, and to analyze the daily dynamic process of the time lag of downstream water temperature response to air temperature changes.

[0010] As a preferred option, the meteorological and hydrological basic dataset includes: flow and water temperature data from upstream and downstream hydrological stations of the reservoir, as well as air temperature data from meteorological stations near the downstream hydrological station.

[0011] As a preferred option, based on the principle of heat balance in the downstream water body, the heat balance equation for the hydrological control section downstream of the dam is established: in, This is the correction factor for the impact of water and heat entering the warehouse. The thermal inertia index is the flow rate. The water-air heat exchange coefficient is denoted as ; the downstream water temperature calculated from the heat balance equation is denoted as . This serves as a physical constraint term in the loss function of the TFT water temperature model.

[0012] The present invention also provides a time-delay analysis system for downstream water temperature responding to air temperature changes, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a time-delay analysis method for downstream water temperature responding to air temperature changes when executed by the processor.

[0013] The present invention also provides a storage medium storing a computer program, which executes a time-delay analysis method for the response of downstream water temperature to air temperature changes during operation.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes the multi-head attention mechanism built into the Temporal Fusion Transformer (TFT) model, using the heat balance equation as a constraint condition for the TFT model. It automatically learns the contribution weight of air temperature at different times to the simulation of downstream water temperature, analyzes the daily dynamic time lag of downstream water temperature response to air temperature changes, and makes up for the shortcomings of traditional cross-correlation analysis in revealing the time-lag dynamic change process of downstream water temperature response to air temperature changes. It improves the methodological system of eco-hydrological time series correlation analysis, enhances the utilization efficiency of meteorological and hydrological data, and provides scientific and technological support for downstream water temperature forecasting and reservoir ecological scheduling. Attached Figure Description

[0015] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the time-delay analysis method for the response of downstream water temperature to air temperature changes according to an embodiment of the present invention; Figure 2 The data is daily temperature data from 2014 to 2024 from a meteorological station near the hydrological station downstream of the reservoir, which is an example of this invention.

[0017] Figure 3 The data is the daily inflow data of the upstream hydrological station of the reservoir from 2014 to 2024, which is an example of the present invention.

[0018] Figure 4 The data represents the daily inflow water temperature data from the upstream hydrological station of the reservoir in this invention, collected from 2014 to 2024.

[0019] Figure 5 The data provided are the daily downstream flow data of the reservoir's hydrological station from 2014 to 2024, as presented in this invention example.

[0020] Figure 6 The data provided are daily downstream water temperature data from 2014 to 2024 at the hydrological station downstream of the reservoir, as described in this invention example.

[0021] Figure 7 This is the annual variation process of the water temperature downstream of the dam in response to changes in air temperature, as described in this invention example. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Example 1 like Figure 1 As shown, this invention provides a time-delay analysis method for the response of downstream water temperature to air temperature changes, comprising: collecting flow and water temperature data from upstream and downstream hydrological stations of the reservoir, as well as air temperature data from meteorological stations near the downstream hydrological station, to establish a basic meteorological and hydrological dataset; establishing a heat balance equation for the downstream hydrological control section as a physical constraint condition for the TFT water temperature model; constructing a TFT water temperature model for the downstream hydrological control section, and using the multi-head attention mechanism of the TFT model to analyze the contribution weight of air temperature to the downstream water temperature simulation at different times, thereby analyzing the daily dynamic process of the time delay in the response of downstream water temperature to air temperature changes. Figure 1 This is a flowchart of the time-delay analysis method for downstream water temperature response to air temperature changes based on the attention mechanism of this invention. The specific steps are as follows: (1) Data collection and preprocessing Collect long-term daily data such as air temperature, flow rate, and water temperature to establish a basic meteorological and hydrological dataset. Specifically: ① Collect long-term daily temperature data from meteorological stations near the hydrological station downstream of the reservoir, and record it as follows: , in °C, where Indicates time.

[0025] ② Collect long-term daily inflow and inflow water temperature data from upstream hydrological stations. The inflow is denoted as... Unit m 3 / s; the inlet water temperature is recorded as , in °C.

[0026] ③ Collect long-term daily data on downstream flow and water temperature from hydrological stations downstream of the reservoir. Downstream flow is denoted as... Unit m 3 / s; the water temperature downstream of the dam is recorded as , in °C.

[0027] ④ The z-score method was used to standardize and preprocess the air temperature, flow rate, and water temperature data. The training set, validation set, and test set were divided into three parts according to the ratio of 60%, 20%, and 20% and in chronological order, respectively, to build the TFT water temperature model.

[0028] (2) Construction of heat balance equation Based on the principle of heat balance in the downstream water body, the heat balance equation for the hydrological control section downstream of the dam is established: In the formula, This is the correction factor for the impact of water and heat entering the warehouse. The thermal inertia index is the flow rate. The water-air heat exchange coefficient is used for calibration using the raw values ​​of meteorological and hydrological data such as air temperature, flow rate, and water temperature from step (1). , and Or determine based on the actual situation of the reservoir project. , and The downstream water temperature calculated using the heat balance equation is denoted as... This serves as a physical constraint term in the loss function of the TFT water temperature model.

[0029] (3) Construction of TFT water temperature model Construct a TFT water temperature model for the hydrological control section downstream of the dam, specifically: ① Establish time-series characteristic engineering Set the backtracking window length to , Positive integer values ​​such as 7 days, 15 days, 30 days, 60 days, and 90 days can be selected. The output variable of the TFT water temperature model is... The water temperature downstream of the dam at any given time, the input variable is Time passed The air temperature and the water temperature downstream of the dam. For every moment... Construct the input feature vector for the TFT water temperature model: [a] Temperature at any time 1 to The lag value of days, i.e. .

[0030] [b] Water temperature downstream of the dam at any time 1 to The lag value of days, i.e. .

[0031] ② Establish a TFT water temperature model constrained by physical mechanisms. A TFT-based downstream water temperature model is constructed, and the heat balance equation from step (2) is introduced as a physical constraint. The total loss function of the TFT water temperature model is defined. : In the formula, These are the physical constraint weighting coefficients. ; Let the mean squared error loss function be used. This is the measured value of the water temperature downstream of the dam. The values ​​are TFT model simulation values ​​of the water temperature downstream of the dam. The a priori value of the downstream water temperature calculated using the heat balance equation.

[0032] The multi-head self-attention mechanism of the TFT model is used to capture the dependencies between different locations in the meteorological and hydrological series, and outputs an attention weight matrix. ,in This represents the number of attention heads in the TFT model.

[0033] Based on the meteorological and hydrological data such as air temperature and water temperature in step (1), the TFT water temperature model is trained using the Adam optimizer until the validation set loss converges. The root mean square error (RMSE) is used. RMSE ), correlation coefficient ( R Nash efficiency coefficient ( NSE ), Kling-Gupta efficiency coefficient ( KGE The simulation accuracy of the TFT water temperature model was evaluated on the training set, validation set, and test set.

[0034] (4) Target feature attention weight extraction The contribution weights of air temperature to the downstream water temperature simulation at different times are analyzed using the multi-head attention mechanism of the TFT model, and used to analyze the weighted average lag time. From the TFT water temperature model trained in step (3), the multi-head attention weights of the last step of the decoder of the TFT model are extracted to obtain the attention weight matrix. The corresponding last query position (current) (Time) for all historical locations (backtracking window) Attention distribution at historical moments : In the formula, Attention weight matrix Chinese correspondence Attention distribution at any given moment ; For the first Head attention mechanism Attention distribution at any given moment.

[0035] The TFT water temperature model was quantitatively characterized in simulating the current situation. When the water temperature downstream of the dam is constant, The water temperature downstream of the dam at different historical moments ( time,……, time, The degree of dependence on time information (i.e., temperature). Normalize the result so that the sum is 1. .

[0036] (5) Weighted average lag time analysis Based on the target feature attention weights extracted in step (4), the daily dynamic process of the water temperature response to air temperature changes downstream of the dam is analyzed with time lag. The lag day array is then used to analyze this process. With attention weight Perform a weighted summation to obtain Time lag of downstream water temperature response to air temperature changes : Based on the daily analysis of attention weights and weighted average lag time using the TFT water temperature model, a daily dynamic lag time series is obtained. The average value of the time lag between the water temperature downstream of the dam and the air temperature change is calculated on a yearly basis (Day of Year) to reveal the annual variation process of the time lag between the water temperature downstream of the dam and the air temperature change.

[0037] Example: The application research area of ​​this invention is the downstream section of a large reservoir on the upper reaches of the Yangtze River, and it analyzes the annual variation process of the water temperature downstream of the large reservoir in response to changes in air temperature with time lag.

[0038] Step 1: Data Collection and Preprocessing Collect daily data on air temperature, flow rate, and water temperature from 2014 to 2024 to establish a basic meteorological and hydrological dataset. Specifically: ① Collect daily temperature data from meteorological stations near the hydrological station downstream of the reservoir from 2014 to 2024. Figure 2 The data is daily temperature data from 2014 to 2024 from a meteorological station near the hydrological station downstream of the reservoir, which is an example of this invention.

[0039] ② Collect daily inflow and inflow water temperature data from upstream hydrological stations for the period from 2014 to 2024. Figure 3 The data is the daily inflow data of the upstream hydrological station of the reservoir from 2014 to 2024, which is an example of the present invention. Figure 4 The data represents the daily inflow water temperature data from the upstream hydrological station of the reservoir in this invention, collected from 2014 to 2024.

[0040] ③ Collect daily downstream flow and downstream water temperature data from hydrological stations downstream of the reservoir from 2014 to 2024. Figure 5 The data provided are the daily downstream flow data of the reservoir's hydrological station from 2014 to 2024, as presented in this invention example. Figure 6 The data provided are daily downstream water temperature data from 2014 to 2024 at the hydrological station downstream of the reservoir, as described in this invention example.

[0041] ④ The z-score method was used to standardize and preprocess the air temperature, flow rate, and water temperature data. Data from 2014 to 2020 were selected as the training set, data from 2021 to 2022 were selected as the validation set, and data from 2023 to 2024 were selected as the test set to build the TFT water temperature model.

[0042] Step 2: Constructing the Heat Balance Equation Based on the principle of heat balance in the downstream water body, a heat balance equation for the hydrological control section downstream of the dam is established, and the original values ​​of meteorological and hydrological data such as air temperature, flow rate, and water temperature from step (1) are used for calibration. , and : Step 3: Constructing the TFT Water Temperature Model Construct a TFT water temperature model for the hydrological control section downstream of the dam, specifically: ① Establish time-series characteristic engineering Set the backtracking window length to The output variable of the TFT water temperature model is The water temperature downstream of the dam at any given time, the input variable is The air temperature and water temperature downstream of the dam over the past 60 days. For each moment... Construct the input feature vector for the TFT water temperature model: [a] The lag value of temperature over 1 to 60 days, i.e. .

[0043] [b] The lag value of the water temperature downstream of the dam from 1 to 60 days, i.e. .

[0044] ② Establish a TFT water temperature model constrained by physical mechanisms. A TFT-based downstream water temperature model is constructed, and the heat balance equation from step (2) is introduced as a physical constraint. The total loss function of the TFT water temperature model is defined. ( )for: The multi-head self-attention mechanism of the TFT model is used to capture the dependencies between different locations in the meteorological and hydrological series, and outputs an attention weight matrix. ,in The number of attention heads is determined. Based on the meteorological and hydrological data such as air temperature and water temperature in step (1), the TFT water temperature model is trained using the Adam optimizer until the validation set loss converges. The root mean square error (RMSE) is used. RMSE ), correlation coefficient ( R Nash efficiency coefficient ( NSE ), Kling-Gupta efficiency coefficient ( KGE The simulation accuracy of the TFT water temperature model on the training set, validation set, and test set was evaluated, as shown in Table 1.

[0045] Table 1

[0046] Step 4: Target Feature Attention Weight Extraction The contribution weights of air temperature to the downstream water temperature simulation at different times are analyzed using the multi-head attention mechanism of the TFT model, and used to analyze the weighted average lag time. From the TFT water temperature model trained in step (3), the multi-head attention weights of the last step of the decoder of the TFT model are extracted to obtain the attention weight matrix. The corresponding last query position (current) Attention distribution across all historical locations (60 historical moments in the backtracking window) .right Normalize the result so that the sum is 1. .

[0047] Step 5: Weighted Average Lag Time Analysis Based on the target feature attention weights extracted in step (4), the daily dynamic process of the water temperature response to air temperature changes downstream of the dam is analyzed with time lag. The lag day array is then used to analyze this process. With attention weight Perform a weighted summation to obtain Time lag of downstream water temperature response to air temperature changes : Based on the daily analysis of attention weights and weighted average lag time using the TFT water temperature model, a daily dynamic lag time series is obtained. The average value of the time lag between the water temperature downstream of the dam and the air temperature change is calculated on a yearly basis (Day of Year) to reveal the annual variation process of the time lag between the water temperature downstream of the dam and the air temperature change. Figure 7This is the annual variation process of the water temperature downstream of the dam in response to changes in air temperature, as described in this invention example.

[0048] The application of this invention's time-delay analysis method for downstream water temperature response to air temperature changes, based on an attention mechanism, is as follows: It can be applied to constructing a TFT-based downstream water temperature model constrained by physical mechanisms. Utilizing the multi-head attention mechanism built into the TFT model, it automatically learns the contribution weights of air temperature at different times to the simulation of downstream water temperature, analyzing the daily dynamic time delay of downstream water temperature response to air temperature changes. This overcomes the limitation of traditional cross-correlation analysis, which can only output a single static time delay, thus improving the methodology of eco-hydrological time series correlation analysis and enhancing the utilization efficiency of meteorological and hydrological data. This provides scientific and technological support for downstream water temperature forecasting and reservoir ecological management. This invention is not limited to analyzing the time delay of downstream water temperature response to air temperature changes; it can be extended to other environmental factors affecting river water temperature, such as solar radiation and wind speed, depending on the actual river conditions. The heat balance equation and TFT water temperature model are adjusted according to the input variables to analyze the annual variation characteristics of downstream water temperature in response to changes in other different influencing factors.

[0049] Example 2 The present invention also provides a time-delay analysis device for the response of downstream water temperature to air temperature changes, comprising: The first processing module is used to acquire basic meteorological and hydrological datasets; The second processing module is used to establish the heat balance equation of the hydrological control section downstream of the dam. The third processing module is used to construct a TFT water temperature model of the hydrological control section downstream of the dam based on the basic meteorological and hydrological dataset; wherein, the heat balance equation is used as the physical constraint condition of the TFT water temperature model. The fourth processing module is used to analyze the contribution weight of air temperature to the simulation of downstream water temperature at different times using the multi-head attention mechanism of the TFT model, and to analyze the daily dynamic process of the time lag of downstream water temperature response to air temperature changes.

[0050] As one embodiment of the present invention, the meteorological and hydrological basic dataset includes: flow and water temperature data of upstream and downstream hydrological stations of the reservoir, and air temperature data of meteorological stations near the downstream hydrological station.

[0051] As one embodiment of the present invention, a heat balance equation for the hydrological control section downstream of the dam is established based on the principle of heat balance in the downstream water body: in, This is the correction factor for the impact of water and heat entering the warehouse. The thermal inertia index is the flow rate. The water-air heat exchange coefficient is denoted as ; the downstream water temperature calculated from the heat balance equation is denoted as . This serves as a physical constraint term in the loss function of the TFT water temperature model.

[0052] Example 3 The present invention also provides a time-delay analysis system for downstream water temperature responding to air temperature changes, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a time-delay analysis method for downstream water temperature responding to air temperature changes when executed by the processor.

[0053] Example 4 The present invention also provides a storage medium storing a computer program, which executes a time-delay analysis method for the response of downstream water temperature to air temperature changes during operation.

[0054] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A time-delay analytical method for the response of downstream water temperature to air temperature changes, characterized in that, include: Obtain basic meteorological and hydrological datasets; Establish the heat balance equation for the hydrological control section downstream of the dam; A TFT water temperature model for the downstream hydrological control section is constructed based on basic meteorological and hydrological datasets; the heat balance equation is used as the physical constraint condition for the TFT water temperature model. The contribution weight of air temperature to the simulation of downstream water temperature at different times is analyzed using the multi-head attention mechanism of the TFT model, and the daily dynamic process of the time lag of downstream water temperature response to air temperature changes is analyzed.

2. The time-delay analysis method for the response of downstream water temperature to air temperature changes as described in claim 1, characterized in that, The basic meteorological and hydrological dataset includes: flow and water temperature data from upstream and downstream hydrological stations of the reservoir, as well as air temperature data from meteorological stations near the downstream hydrological station.

3. The time-delay analysis method for the response of downstream water temperature to air temperature changes as described in claim 2, characterized in that, Based on the principle of heat balance in the downstream water body, the heat balance equation for the hydrological control section downstream of the dam is established: in, This is the correction factor for the impact of water and heat entering the warehouse. The thermal inertia index is the flow rate. The water-air heat exchange coefficient is denoted as ; the downstream water temperature calculated from the heat balance equation is denoted as . This serves as a physical constraint term in the loss function of the TFT water temperature model.

4. A time-delay analysis device for the response of downstream water temperature to air temperature changes, characterized in that, include: The first processing module is used to acquire basic meteorological and hydrological datasets; The second processing module is used to establish the heat balance equation of the hydrological control section downstream of the dam. The third processing module is used to construct a TFT water temperature model of the hydrological control section downstream of the dam based on the basic meteorological and hydrological dataset; wherein, the heat balance equation is used as the physical constraint condition of the TFT water temperature model. The fourth processing module is used to analyze the contribution weight of air temperature to the simulation of downstream water temperature at different times using the multi-head attention mechanism of the TFT model, and to analyze the daily dynamic process of the time lag of downstream water temperature response to air temperature changes.

5. The time-delay analysis device for downstream water temperature response to air temperature changes as described in claim 4, characterized in that, The basic meteorological and hydrological dataset includes: flow and water temperature data from upstream and downstream hydrological stations of the reservoir, as well as air temperature data from meteorological stations near the downstream hydrological station.

6. The time-delay analysis device for downstream water temperature response to air temperature changes as described in claim 5, characterized in that, Based on the principle of heat balance in the downstream water body, the heat balance equation for the hydrological control section downstream of the dam is established: in, This is the correction factor for the impact of water and heat entering the warehouse. The thermal inertia index is the flow rate. The water-air heat exchange coefficient is denoted as ; the downstream water temperature calculated from the heat balance equation is denoted as . This serves as a physical constraint term in the loss function of the TFT water temperature model.

7. A time-delay analysis system for the response of downstream water temperature to air temperature changes, characterized in that, include: The system includes a memory and a processor, wherein the memory stores a computer program that is executed by the processor, and the computer program, when executed by the processor, performs the time-delay analysis method for downstream water temperature response to air temperature changes as described in any one of claims 1-3.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when running, executes the time-delay analysis method for the response of downstream water temperature to air temperature changes as described in any one of claims 1-3.

Images