Complex factor-driven Yangtze River upstream long-term inflow quantitative forecasting method
By utilizing factors such as sea surface temperature, sea ice, snow depth, and soil moisture through a multiple regression model, a long-term quantitative forecasting method for the inflow of water in the upper reaches of the Yangtze River was constructed. This method solves the problem of neglecting the spatial distribution details of factors in traditional methods and achieves a higher accuracy forecast for the inflow of water in the upper reaches of the Yangtze River.
Patent Information
- Application Number
- CN202511136508.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-14
- Publication Date
- 2025-11-21
AI Technical Summary
Existing technologies fail to fully utilize the spatial distribution details of complex factors in long-term hydrological forecasting, resulting in insufficient forecast accuracy. In particular, in the long-term inflow forecasting of the upper reaches of the Yangtze River, traditional methods may ignore the local characteristics of factors such as sea surface temperature, sea ice, soil moisture, and snow cover.
Using a multiple regression model, sea surface temperature, sea ice, snow depth and soil moisture as climate factors, and by constructing statistical models for different forecast periods and months, we can identify the correlations in key areas and establish a long-term quantitative forecasting method for water inflow in the upper reaches of the Yangtze River driven by complex factors.
It has improved the accuracy and predictive ability of long-term inflow forecasts in the upper reaches of the Yangtze River. Through the application of multiple regression models, it has achieved more accurate predictions of runoff, meeting operational needs.
Smart Images
Figure CN120995109A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of hydrological prediction, and in particular to a complex factor driven long-term inflow prediction method for the upper reaches of the Yangtze River. BACKGROUND
[0002] Long-term prediction research methods can be mainly divided into two categories. The first category is sequence evolution method, and the core is to seek time sequence evolution law from the prediction variable itself. The second category is factor relationship method, and the core is to mine and build the relationship between factors and prediction variables. In this chapter, SVD and other methods are used to analyze the possible relationship between Pacific sea surface temperature, Atlantic sea surface temperature, Arctic sea ice, Eurasian snow and the runoff of the upper reaches of the Yangtze River, Xiangjiaba and Three Gorges Reservoir, identify the key area of climate factors affecting precipitation and Three Gorges Reservoir inflow, and discuss the main influence mechanism, which provides a basis for building a runoff statistical prediction model in the later stage.
[0003] It can be found that in the factor selection, the grid data of sea surface temperature, sea ice, soil moisture and snow cover are introduced to find more accurate correlation. Traditionally, research usually relies on circulation indices or key area sea surface temperature indices. Although these methods are effective, they may ignore the details and local characteristics of spatial distribution. By using grid data, the spatial variation of each factor can be analyzed in more detail, and the specific area highly related to climate phenomena can be identified. The advantage of this method is that it not only improves the accuracy of identifying high correlation areas, but also enhances the prediction ability of statistical models. By capturing the high correlation area of complex factors, the runoff can be more accurately predicted. SUMMARY
[0004] The present application aims at the deficiencies of the prior art, and provides a complex factor driven long-term inflow prediction method for the upper reaches of the Yangtze River.
[0005] To achieve the above object, the technical scheme adopted by the present application is as follows: The present application provides a complex factor driven long-term inflow prediction method for the upper reaches of the Yangtze River, comprising: S1, analyzing the influence of climate factors on inflow runoff; S2, identifying climate factors affecting natural runoff; S3, using the climate factors, building statistical models for different prediction periods and months, inputting the climate factor data into a multiple regression prediction model, and dividing the data set according to the proportion of 70% for the training set and 30% for the validation set.
[0006] Further, in S1, the climate factors include sea surface temperature, sea ice, snow depth and soil moisture.
[0007] The sea surface temperature mentioned: With global warming, the sea surface temperature is rising, which leads to an increase in seawater evaporation, thereby changing the distribution and intensity of precipitation; precipitation is increasing in mid-to-high latitude regions and tropical regions, while precipitation is decreasing in subtropical regions, which will directly affect the generation and distribution of surface runoff; The sea ice: Changes in sea ice affect runoff by influencing ocean circulation and heat transfer; the reduction of seawater in the Arctic and Antarctic regions allows more freshwater to enter the ocean, altering the ocean's salinity and density structure, which in turn affects ocean currents and heat transfer; The snow depth: Snow has a high albedo, which can reflect a large amount of solar radiation; as global warming leads to a decrease in snow cover, the surface area exposed to solar radiation increases, and the absorbed heat rises accordingly, further accelerating the process of climate warming. Soil moisture affects the surface energy balance by altering surface albedo, soil heat capacity, and vegetation growth. When soil moisture is high, surface evaporation and vegetation transpiration are enhanced, consuming more energy and causing the surface temperature to decrease. When soil moisture is low, more surface energy is transferred to the atmosphere in the form of sensible heat flux, causing the surface temperature to increase.
[0008] Furthermore, S2 specifically refers to the relationship between pre-holiday climate factors and runoff in April, May, and April. The high correlation area in January is located in the tropical North Atlantic, showing a significant negative correlation. From February to March, the North Atlantic exhibits a tripolar distribution. By subtracting the sea surface temperature from the positively correlated and negatively correlated areas to form a new index for prediction, Atlantic sea surface temperature can serve as a predictor of early-stage climate when forecasting the natural flow of reservoirs in April. The high correlation between the reservoir's natural flow in April and the previous snow depth from January to March is found in the area north of the Caspian Sea and north of Lake Baikal, showing a positive correlation. That is, the increase in the previous snow depth in the key areas corresponds to the increase in natural flow. Moreover, this positive correlation can pass the 0.1 confidence level test. Therefore, the snow depth in the mid-to-high latitudes of Eurasia can be used as a climate prediction factor for the previous period. The high correlation between the reservoir's natural flow in April and the previous soil moisture from January to March is located in the Qinghai-Tibet Plateau and the Indochina Peninsula. The correlation coefficient shows a dipole-shaped change, and all correlations can pass the 0.1 confidence level test. Therefore, soil moisture can be used as a previous climate prediction factor.
[0009] Furthermore, S3 specifically refers to: Multiple linear regression is used to model the linear relationship between multiple independent variables and a dependent variable, finding a linear model such that a linear combination of the independent variables can predict the dependent variable. The expression is: ; in, The dependent variable; For the first One independent variable; as independent variable The regression coefficients; This represents the total number of independent variables; For the intercept term; This is the random error term.
[0010] Furthermore, S3 also includes: Anomaly sign consistency rate: In hydrometeorology, runoff anomalies are used to reflect runoff anomalies. The anomaly sign consistency rate is used to quantitatively evaluate the prediction accuracy of runoff anomalies, ranging from 0 to 100. The higher the score, the better the prediction. The formula is: ; in, This represents the total number of forecasts; when the forecast and actual anomalies have the same sign, The value is 1, and the opposite is true; the value is 0.
[0011] Furthermore, S3 also includes: Root Mean Square Error (RMSE) is a commonly used metric to measure the difference between predicted and observed values. By squaring the error, it amplifies the impact of large errors and is particularly sensitive to large errors in prediction models. A smaller RMSE value indicates a higher accuracy. The formula is: ; in, Indicates the first One predicted value; Indicates the first One actual observation value; This represents the total number of samples.
[0012] Furthermore, S3 also includes: Mean Absolute Error (MAE): A metric that measures the difference between predicted and observed values. It assesses the accuracy of a prediction model by calculating the average of the absolute errors between the predicted and observed values. Unlike RMSE, MAE treats all errors equally and does not overemphasize the impact of large errors. A smaller MAE value indicates higher accuracy of the prediction model. The formula is: .
[0013] The beneficial effects of this invention are as follows: by identifying sea temperature, sea ice, snow depth and soil moisture in key areas related to reservoir inflow as predictive factors, a statistical prediction model based on multiple regression is constructed to forecast monthly inflow, thereby achieving long-term quantitative forecasting of the Yangtze River's upper reaches driven by complex factors. Attached Figure Description
[0014] Figure 1 A flow chart of a complex factor driven long-term inflow prediction method for the upper reaches of the Yangtze River; Figure 2 A prediction result of a multiple regression model in April; Figure 3 A prediction result of a multiple regression model in May; Figure 4 A prediction result of a multiple regression model in June; Figure 5 A prediction result of a multiple regression model in July; Figure 6 A prediction result of a multiple regression model in August; Figure 7 A prediction result of a multiple regression model in September; Figure 8 A prediction result of a multiple regression model in October. DETAILED DESCRIPTION
[0015] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the drawings. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.
[0016] A complex factor driven long-term inflow prediction method for the upper reaches of the Yangtze River, comprising: S1, analyzing the influence of climate factors on inflow runoff; S2, identifying climate factors affecting natural runoff; S3, using the climate factors, constructing statistical models for different prediction periods and months, inputting the climate factor data into a multiple regression prediction model, and dividing the data set according to the proportion of 70% of the training set and 30% of the verification set.
[0017] In the S1, the climate factors include sea temperature, sea ice, snow depth and soil moisture.
[0018] The sea temperature: with global warming, the sea surface temperature rises, leading to an increase in sea evaporation, and thus changing the distribution and intensity of precipitation; precipitation increases in middle and high latitude areas and tropical areas, while precipitation decreases in subtropical areas, which will directly affect the generation and distribution of surface runoff; The sea ice: the change of sea ice affects the runoff by affecting the ocean circulation and heat transfer; the reduction of sea water in the Arctic and Antarctic regions makes more fresh water enter the ocean, changing the salinity and density structure of the ocean, and thus affecting the ocean current movement and heat transfer; The snow depth: Snow has a high albedo, which can reflect a large amount of solar radiation; as global warming leads to a decrease in snow cover, the surface area exposed to solar radiation increases, and the absorbed heat rises accordingly, further accelerating the process of climate warming. Soil moisture affects the surface energy balance by altering surface albedo, soil heat capacity, and vegetation growth. When soil moisture is high, surface evaporation and vegetation transpiration are enhanced, consuming more energy and causing the surface temperature to decrease. When soil moisture is low, more surface energy is transferred to the atmosphere in the form of sensible heat flux, causing the surface temperature to increase.
[0019] Specifically, S2 refers to the relationship between pre-holiday climate factors and runoff in April, May, and April. The high correlation area in January is located in the tropical North Atlantic, showing a significant negative correlation. From February to March, the North Atlantic exhibits a tripolar distribution. By subtracting the sea surface temperature from the positively correlated and negatively correlated areas to form a new index for prediction, Atlantic sea surface temperature can serve as a predictor of early-stage climate when forecasting the natural flow of reservoirs in April. The high correlation between the reservoir's natural flow in April and the previous snow depth from January to March is found in the area north of the Caspian Sea and north of Lake Baikal, showing a positive correlation. That is, the increase in the previous snow depth in the key areas corresponds to the increase in natural flow. Moreover, this positive correlation can pass the 0.1 confidence level test. Therefore, the snow depth in the mid-to-high latitudes of Eurasia can be used as a climate prediction factor for the previous period. The high correlation between the reservoir's natural flow in April and the previous soil moisture from January to March is located in the Qinghai-Tibet Plateau and the Indochina Peninsula. The correlation coefficient shows a dipole-shaped change, and all correlations can pass the 0.1 confidence level test. Therefore, soil moisture can be used as a previous climate prediction factor.
[0020] Specifically, S3 is: Multiple linear regression is used to model the linear relationship between multiple independent variables and a dependent variable, finding a linear model such that a linear combination of the independent variables can predict the dependent variable. The expression is: ; in, The dependent variable; For the first One independent variable; as independent variable The regression coefficients; This represents the total number of independent variables; For the intercept term; This is the random error term.
[0021] S3 further includes: Anomaly sign consistency rate: In hydrometeorology, runoff anomalies are used to reflect runoff anomalies. The anomaly sign consistency rate is used to quantitatively evaluate the prediction accuracy of runoff anomalies, ranging from 0 to 100. The higher the score, the better the prediction. The formula is: ; in, This represents the total number of forecasts; when the forecast and actual anomalies have the same sign, The value is 1, and the opposite is 0.
[0022] S3 further includes: Root Mean Square Error (RMSE) is a commonly used metric to measure the difference between predicted and observed values. By squaring the error, it amplifies the impact of large errors and is particularly sensitive to large errors in prediction models. A smaller RMSE value indicates a higher accuracy. The formula is: ; in, Indicates the first One predicted value; Indicates the first One actual observation value; This represents the total number of samples.
[0023] S3 further includes: Mean Absolute Error (MAE): A metric that measures the difference between predicted and observed values. It assesses the accuracy of a prediction model by calculating the average of the absolute errors between the predicted and observed values. Unlike RMSE, MAE treats all errors equally and does not overemphasize the impact of large errors. A smaller MAE value indicates higher accuracy of the prediction model. The formula is: .
[0024] Taking the Xiangjiaba model construction and prediction results as an example, Tables 1 and 2 present the simulation and prediction performance of the model during the training and validation periods. Based on the test data from the Xiangjiaba model's commissioning and validation periods, the model exhibits strong predictive ability in the short-term forecast period (1-3 months), with the PS score generally exceeding 70% in the first 3 months of the commissioning period. The error index decreases relatively during the validation period. However, in the longer forecast period (4-6 months), limited by the number of factors, the prediction performance decreases with the increase of the forecast period, and the PS score in the validation period decreases by a relatively larger average compared to the commissioning period. For the Xiangjiaba station, the predicted PS score for monthly flow is mostly above 60%, which meets operational requirements.
[0025] Figures 2 to 8 The results visually demonstrate the predictive performance of the constructed multiple regression model during the training and validation periods. Overall, the multiple regression model is able to capture the degree of over- or under-progression in future traffic trends and the interannual fluctuation trend quite well.
[0026] Table 1 Model Effect Test of Xiangjiaba Rating Period
[0027] Table 2 Model Effect Test of Xiangjiaba Rating Period
[0028] The above-described embodiments only express the implementation of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.
Claims
1. A method for quantitatively predicting long-term water inflow in the upper reaches of the Yangtze River driven by complex factors, characterized in that, The application relates to a method for predicting natural runoff of a reservoir, comprising the following steps: S1, analyzing the influence of climate factors on reservoir inflow; S2, identifying climate factors affecting natural runoff; S3, using the climate factors, constructing statistical models for different prediction periods and months, inputting the climate factor data into a multiple regression prediction model, and dividing the data set according to a proportion of 70% for a training set and 30% for a verification set.
2. The method according to claim 1, wherein the method is characterized by: In the S1, the climate factors include sea temperature, sea ice, snow depth and soil moisture; The sea temperature: with global warming, the sea surface temperature rises, leading to an increase in sea water evaporation, and further changing the distribution and intensity of precipitation; precipitation increases in middle and high latitude regions and tropical regions, while precipitation decreases in subtropical regions, thereby directly affecting the generation and distribution of surface runoff; The sea ice: the change of sea ice affects runoff by influencing ocean circulation and heat transfer; the reduction of sea water in the Arctic and Antarctic regions makes more fresh water enter the ocean, changing the salinity and density structure of the ocean, and further affecting the movement of ocean currents and heat transfer; The snow depth: snow has the characteristics of high albedo and can reflect a large amount of solar radiation; with the reduction of snow caused by global warming, the area of the earth's surface exposed to solar radiation increases, and the absorbed heat rises, further accelerating the process of climate warming; The soil moisture: by changing the surface albedo, soil heat capacity and vegetation growth conditions, the soil moisture affects the surface energy balance; when the soil moisture is high, the surface evaporation and vegetation transpiration are enhanced, consuming more energy, leading to a decrease in the surface temperature; when the soil moisture is low, the surface energy is more in the form of sensible heat flux to the atmosphere, making the surface temperature rise.
3. The method according to claim 2, wherein, The S2 specifically is: the relationship between the previous climate factors and the natural runoff of the reservoir in April, May and June: The high correlation area of January in the tropical North Atlantic is a significant negative correlation. From February to March, the North Atlantic shows a north-south three-pole type distribution, and the difference between the positive correlation area of sea temperature and the negative correlation area of sea temperature forms a new index for prediction. When predicting the natural runoff of the reservoir in April, the Atlantic sea temperature can be used as a previous climate prediction factor; The high correlation area of the natural runoff of the reservoir in April and the previous snow depth in January to March is located in the north of the Caspian Sea and the north side of Lake Baikal, showing a positive correlation, that is, the increase of the previous snow depth in the key area corresponds to the increase of the natural runoff; and the positive correlation can pass the 0.1 reliability test, so the snow depth of the Eurasian middle and high latitudes is used as a previous climate prediction factor; The high correlation area of the natural runoff of the reservoir in April and the previous soil moisture in January to March is located in the Qinghai-Tibet Plateau and the Indochina Peninsula, and the correlation coefficient shows a dipole change, and the correlation can pass the 0.1 reliability test, so the soil moisture is used as a previous climate prediction factor.
4. The method according to claim 3, wherein, The S3 specifically is: Multiple linear regression is used to model the linear relationship between multiple independent variables and a dependent variable, and a linear model is found to make the linear combination of independent variables be able to predict the dependent variable, and the expression is: ； where, is the dependent variable; is the first independent variable; is the regression coefficient for the independent variable; is the total number of independent variables; is the intercept term; is the random error term.
5. The complex factor driven long-term inflow forecast method for the upper reaches of the Yangtze River according to claim 4, characterized in that, The S3 further comprises: Consistency of anomaly symbol: In hydro-meteorology, runoff anomaly is used to reflect the abnormality of runoff, and the consistency of anomaly symbol is used to quantitatively evaluate the prediction accuracy of runoff anomaly, ranging from 0 to 100, the higher the score, the better the effect, the formula is: ； wherein, is the total number of predictions; when the forecast is the same sign as the live anomaly, takes the value 1 and the opposite value 0.
6. The complex factor driven long-term inflow quantitative prediction method of the upper reaches of the Yangtze River according to claim 5, characterized in that, The S3 further includes: Root mean square error: a commonly used index to measure the difference between predicted values and actual observations. By squaring the error, it amplifies the impact of large errors, making it particularly sensitive to large errors in the prediction model. The smaller the value of RMSE, the better the prediction model. The formula is: ； wherein, represents the first predicted value; represents the first actual observed value; represents the total number of samples.
7. The complex factor driven long-term inflow forecasting method for the upper reaches of the Yangtze River according to claim 6, characterized in that, The S3 further includes: Mean absolute error: an index to measure the difference between predicted values and actual observations. By calculating the average of the absolute error between predicted values and actual values, it evaluates the accuracy of the prediction model. Unlike RMSE, MAE treats all errors equally and does not amplify the impact of large errors. The smaller the value of MAE, the higher the accuracy of the prediction model. The formula is: 。