A snowmelt runoff calculation method based on meteorological element advantage analysis

Abstract

The present invention discloses a snowmelt runoff calculation method based on meteorological factor advantage analysis, including the following specific steps: obtaining runoff data from a target hydrological station; constructing a target influencing monthly factor matrix based on the runoff data from the target hydrological station; constructing multiple models and calculating the contribution rate of each target influencing factor to runoff in different models; and calculating the snowmelt runoff volume. The present invention uses a partial correlation method to eliminate the influence of other influencing factors on the runoff data from the target hydrological station, and considers different influencing factors for runoff in different months; constructs a multi-model group of monthly regression equation sub-models and multiple target machine learning models, and uses the fitting effects of different models to propose a weighted average multi-model integrated contribution rate calculation method to obtain the contribution rate of each target influencing factor to runoff, effectively overcoming the non-universality of the contribution rate of the influencing factors on a single model, and enhancing the scientific nature and interpretability of the current snowmelt runoff calculation method.

Description

Technical Field

[0001] The present invention relates to the field of snowmelt runoff calculation, and in particular to a snowmelt runoff calculation method based on meteorological element advantage analysis. Background Art

[0002] Snowmelt runoff is a significant source of runoff in cold and arid regions. For example, snowmelt runoff from a source region is a significant component of runoff in a river basin, and in certain plateau regions, snowmelt runoff can account for 30% to 40% of the total annual runoff. In spring, rapid temperature rises and rapid snowmelt can easily trigger spring floods. Therefore, understanding the contribution and proportion of snowmelt to runoff is crucial for water resource management and flood prevention and mitigation. Current methods for calculating snowmelt runoff primarily rely on runoff segmentation methods, such as electronic filtering, isotope chemistry, hydrological models, graphical analysis, energy balance, and Eckhardt recursive digital filtering. However, due to the complex composition of runoff sources within a source region, spring runoff may include rainfall runoff, snowmelt runoff, and icemelt runoff. Furthermore, numerous factors influence runoff, including previous snowmelt, runoff, rainfall, evaporation, and soil moisture, all of which contribute to varying degrees. Currently, commonly used methods directly process runoff data, failing to account for the physical factors that influence runoff composition, such as meteorological and water resource factors. This lacks the mechanisms and interpretability of runoff variation, hindering the understanding and prediction of snowmelt runoff. To quantitatively understand the contribution of various factors to runoff, particularly the impact of snowmelt on runoff, and to achieve scientific management and efficient utilization of water resources, this paper proposes a method for estimating snowmelt runoff from a source region based on meteorological factor analysis. Summary of the Invention

[0003] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and provide a snowmelt runoff calculation method based on the analysis of meteorological element advantages.

[0004] The technical solution adopted by the present invention is:

[0005] A snowmelt runoff calculation method based on meteorological factor advantage analysis includes the following specific steps:

[0006] S1. Obtain runoff data of target hydrological station;

[0007] S2. Based on the runoff data of the target hydrological station, the target influencing monthly factor matrix is ​​constructed;

[0008] S3. Construct multiple models and calculate the contribution rate of each target influencing factor to runoff in different models:

[0009] S31. Based on the runoff data of the target hydrological station and the target influencing monthly factor matrix, a sub-model of the monthly regression equation group including the target influencing factors is established;

[0010] S32. Analyze the goodness of fit of each target influencing factor to the sub-model of the monthly regression equation group based on the relative importance method, and calculate the average contribution increment of the goodness of fit;

[0011] S33. Based on the average contribution increment of the goodness of fit, the contribution of each target influencing factor to runoff in the sub-model of the monthly regression equation group is obtained;

[0012] S34. Based on the target hydrological station runoff and the target influencing monthly factor matrix, multiple target machine learning models are constructed;

[0013] S35. Perform model evaluation on multiple target machine learning models to obtain the goodness of fit of the multiple target machine learning models;

[0014] S36. Based on the goodness of fit of the multi-objective machine learning model, the contribution of each target influencing factor to runoff in the multi-objective machine learning model is calculated using a model interpretability algorithm;

[0015] S37. Based on the fitting results of the monthly regression equation group sub-models and multiple target machine learning models, assign corresponding weights to different models;

[0016] S38. Based on the assigned weights, the contribution rate of each target influencing factor to runoff in different models is calculated;

[0017] S4. Calculate snowmelt runoff:

[0018] S41. Based on the contribution rate of each target influencing factor to runoff in different models, a matrix of the contribution rate of the influencing factors of runoff at the target hydrological station is constructed;

[0019] S42. Screening the influencing factors related to snowmelt from the target influencing factors to obtain the influencing factors as the target snowmelt influencing factors;

[0020] S43. Multiply the sum of the contribution rates corresponding to the target snowmelt influencing factors by the runoff data of the target hydrological station to obtain runoff data related to snowmelt, i.e., snowmelt runoff volume.

[0021] Furthermore, S1 specifically includes the following steps:

[0022] S11. Determine the target area and the hydrological stations within the target area;

[0023] S12, selecting a hydrological station in the target area as a target hydrological station;

[0024] S13. Select the long-term monthly runoff data of the target hydrological station as the runoff data of the target hydrological station:

[0025] Assume that the year in the long series monthly runoff data is , the month is , then the runoff data of the target hydrological station is ,in , =1,2,3...,12, .

[0026] Furthermore, S2 specifically includes the following steps:

[0027] S21. Based on the runoff data of the target hydrological station, select the meteorological factors that affect the runoff of the target hydrological station as the target influencing factors;

[0028] S22, arranging the target impact factors in the order of year and month corresponding to the runoff data of the target hydrological station to construct a target impact factor data set;

[0029] S23. Based on the target influencing factor dataset, the partial correlation method is used to calculate the lead-lag partial correlation coefficient between the target hydrological station runoff data and each target influencing factor, and a partial correlation coefficient tensor is constructed;

[0030] S24. Based on the partial correlation coefficient tensor, find the month corresponding to the maximum partial correlation coefficient between each target influencing factor and the runoff of the target hydrological station to obtain the target influencing monthly factor of the runoff of the target hydrological station;

[0031] S25, obtaining the maximum partial correlation coefficient matrix of monthly runoff based on the months in S24;

[0032] S26. Based on the target influencing monthly factors and the monthly runoff maximum partial correlation coefficient matrix, a target influencing monthly factor matrix is ​​constructed.

[0033] Furthermore, the target impact factor dataset , as follows:

[0034] (1)

[0035] In formula (1), is the total number of target impact factor types; is the year corresponding to the target impact factor, is the month corresponding to the target impact factor, where 、 Keep corresponding to the value in S13;

[0036] The partial correlation coefficient tensor , as follows:

[0037] (2)

[0038] In formula (2), The monthly runoff data from January to December of the target hydrological station and The correlation between different months corresponding to the target impact factors;

[0039] The maximum partial correlation coefficient matrix , as follows:

[0040] (3)

[0041] In formula (3), To influence the Monthly runoff The maximum monthly partial correlation coefficient of the target influencing factors;

[0042] Based on the maximum partial correlation coefficient matrix of the target influencing monthly factors and monthly runoff , construct the target impact monthly factor matrix .

[0043] Furthermore, in the monthly regression equation group sub-model, the number of sub-models for each month is 2 n-1 A total of 2 n-1 × sub-models, where the regression equations for a certain month It can be expressed as:

[0044] (4)

[0045] The goodness of fit It can be expressed as:

[0046] (5)

[0047] In formula (5), TSS is the total sum of squares of deviations; RSS is the regression sum of squares; ESS is the residual sum of squares;

[0048] Contribution of each target influencing factor to runoff in the monthly regression equation sub-model , as follows:

[0049] (6).

[0050] Furthermore, the process of constructing the multiple target machine learning models specifically includes the following steps:

[0051] 1) Use XGboost, LSTM or random forest intelligent prediction methods to build Machine learning models , as follows:

[0052] ① The runoff of the target hydrological station Matrix of monthly factors affecting the target Divide into training sets in proportion With the test set ;

[0053] ② Use the training set to train each machine learning model to obtain multiple target machine learning models , and input the test set into multiple target machine learning models to obtain Prediction results of runoff data using a machine learning model;

[0054] 2) Yes The model evaluation is performed on the target machine learning model to obtain the goodness of fit of multiple target machine learning models. ;

[0055] 3) Calculated by the model interpretability algorithm Contribution of each target influencing factor to runoff in a target learning model , as follows:

[0056] (7)

[0057] 4) Goodness of fit of the sub-model based on the monthly regression equation group and goodness of fit of machine learning models with multiple objectives ,get( ) Model fitting effect ;

[0058] 5) Based on ( ) Model fitting effect , giving ( ) models with different weights get corresponding weights :

[0059] (8)

[0060] In formula (7), Model type, Represents the proportion of the fitting effect of a certain model to the sum of the fitting effects of all models;

[0061] 6) Based on weight , calculate the contribution rate of each target influencing factor to runoff, that is, the weighted average final contribution rate of each target influencing factor , the specific calculation is as follows:

[0062] (9).

[0063] Furthermore, the contribution rate matrix of the factors affecting the runoff volume of the target hydrological station is , as follows:

[0064] (10).

[0065] Furthermore, the snowmelt runoff , the specific calculation is as follows:

[0066] (11)

[0067] In formula (11), is the year corresponding to the target impact factor, is the month corresponding to the target impact factor, is the number of target snowmelt influencing factors.

[0068] Compared with the prior art, the present invention has the following beneficial effects:

[0069] The present invention fully considers the complexity of runoff composition, uses the partial correlation method to eliminate the influence of other influencing factors on the runoff data of the target hydrological station, and considers different influencing factors for the runoff in different months; considers the fitting effect of different models and the contribution rate of each target influencing factor to runoff in different models, constructs a multi-model group of monthly regression equation sub-models and multiple target machine learning models, and proposes a weighted average multi-model integrated contribution rate calculation method based on the fitting effect of different models, which effectively overcomes the non-universality of the contribution rate of the influencing factors on a single model and enhances the scientificity and interpretability of the current snowmelt runoff calculation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0071] Figure 1 It is a flow chart of a snowmelt runoff calculation method based on meteorological factor advantage analysis of the present invention. DETAILED DESCRIPTION

[0072] Example

[0073] like Figure 1 As shown in FIG, a snowmelt runoff calculation method based on meteorological factor advantage analysis includes the following specific steps:

[0074] S1. Obtain runoff data of target hydrological station;

[0075] Specifically, S11, determining the target area and the hydrological stations in the target area;

[0076] S12, selecting a hydrological station in the target area as a target hydrological station;

[0077] S13. Select the long-term monthly runoff data of the target hydrological station as the runoff data of the target hydrological station:

[0078] Assume that the year in the long series monthly runoff data is , the month is , then the runoff data of the target hydrological station is ,in , =1, 2, 3..., 12; In order to ensure that the number of samples is sufficient, it is necessary to meet The sample size requirement is to obtain at least 30 years of runoff data from the target hydrological stations to calculate snowmelt runoff for all months of the year;

[0079] In this example, the target area is determined to be a river basin and multiple hydrological stations in the river basin. A station in the river basin is selected as the target hydrological station, and the monthly runoff data from 1979 to 2019 are selected as the runoff data of the station. The sample size is 41 years from 1979 to 2019, which meets the requirements.

[0080] S2. Based on the runoff data of the target hydrological station, the target influencing monthly factor matrix is ​​constructed:

[0081] Specifically, S21, based on the runoff data of the target hydrological station, selecting meteorological elements that affect the runoff of the target hydrological station as target influencing factors;

[0082] S22, arranging the target impact factors in the order of year and month corresponding to the runoff data of the target hydrological station to construct a target impact factor data set;

[0083] S23. Based on the target influencing factor dataset, the partial correlation method is used to calculate the lead-lag partial correlation coefficient between the target hydrological station runoff data and each target influencing factor, and a partial correlation coefficient tensor is constructed;

[0084] S24. Based on the partial correlation coefficient tensor, find the month corresponding to the maximum partial correlation coefficient between each target influencing factor and the runoff of the target hydrological station to obtain the target influencing monthly factor of the runoff of the target hydrological station;

[0085] S25, based on the months in S24, obtain the monthly runoff maximum partial correlation coefficient matrix;

[0086] S26. Based on the target influencing monthly factors and the monthly runoff maximum partial correlation coefficient matrix, a target influencing monthly factor matrix is ​​constructed;

[0087] In this embodiment, based on the 41-year runoff data of the target hydrological station, , select the meteorological factors that affect the runoff of the target hydrological station as the target influencing factors, and the specific operations are as follows:

[0088] The study examined 130 circulation indices from a climate center, including 88 atmospheric circulation indices, 26 sea temperature indices, and 16 other types of climate indices. Among them, rainfall, runoff, snow depth, evaporation, and soil moisture were selected as target influencing factors affecting runoff at a station.

[0089] In order to facilitate the calculation of the lead-lag partial correlation coefficient between the runoff data of the target hydrological station and each target influencing factor, one year is taken forward and one year backward based on the runoff data of a certain station, that is, the monthly runoff data of 43 years from 1978 to 2020 are selected for the above target influencing factors;

[0090] The monthly runoff data corresponding to the target impact factor are arranged in chronological order to construct the target impact factor dataset. , as follows:

[0091] (1)

[0092] In formula (1), is the total number of target impact factor types; is the year corresponding to the target impact factor, is the month corresponding to the target impact factor, where 、 Keep corresponding to the value in S13;

[0093] Based on the target influencing factor dataset, the partial correlation method is used to calculate the lead-lag partial correlation coefficient between the target hydrological station runoff data and each target influencing factor, and the partial correlation coefficient tensor is constructed. , as follows:

[0094] (2)

[0095] In formula (2), The monthly runoff data from January to December of the target hydrological station and The correlation between different months corresponding to the target influencing factors; the partial correlation method is used to avoid the influence of the interaction between multiple factors on the correlation; in addition, the calculated partial correlation coefficient can be tested for significance to ensure the reliability and validity of the results;

[0096] Based on the partial correlation coefficient tensor , find the month corresponding to the maximum partial correlation coefficient between each target influencing factor and the runoff of the target hydrological station to obtain the target influencing monthly factor of the runoff of the target hydrological station;

[0097] Through the lead-lag partial correlation coefficient results of the target hydrological station runoff data and each target influencing factor, it is found that the month corresponding to the largest partial correlation coefficient between the target influencing factor selected in this embodiment and the target hydrological station runoff is the same period or the previous month, that is, these target influencing factors have a significant impact on the runoff of a certain station during the same period or one month in advance. Therefore, the target influencing factors corresponding to the same period and the previous month are considered to be the target influencing month factors of the runoff of a certain station, as shown in Table 1; the target influencing factors of other months are omitted here because they have relatively little impact on the runoff of a certain station.

[0098] Table 1 shows the partial correlation coefficients between the monthly runoff of a station and the target influencing factors, specifically the partial correlation coefficients between the monthly runoff of a station and the target influencing factors of the same period or one month in advance:

[0099] Table 1 Partial correlation coefficients between monthly runoff at a station and target influencing factors

[0100]

[0101] In Table 1, ✽✽ means passing the 99% significance test, and ✽ means passing the 95% significance test, indicating that the results are reliable and valid;

[0102] Based on the months mentioned above, the maximum partial correlation coefficient matrix of monthly runoff is obtained , as follows:

[0103] (3)

[0104] In formula (3), To influence the Monthly runoff The maximum monthly partial correlation coefficient of the target influencing factors;

[0105] Based on the maximum partial correlation coefficient matrix of the target influencing monthly factors and monthly runoff , construct the target impact monthly factor matrix .

[0106] S3. Calculate the contribution rate of the target influencing factor to runoff:

[0107] Specifically, S31, based on the runoff data of the target hydrological station and the target influencing monthly factor matrix, a sub-model of the monthly regression equation group including the target influencing factors is established;

[0108] S32. Analyze the goodness of fit of each target influencing factor to the sub-model of the monthly regression equation group based on the relative importance method, and calculate the average contribution increment of the goodness of fit;

[0109] S33. Based on the average contribution increment of the goodness of fit, the contribution of each target influencing factor to runoff in the sub-model of the monthly regression equation group is obtained;

[0110] S34. Based on the target hydrological station runoff and the target influencing monthly factor matrix, multiple target machine learning models are constructed;

[0111] S35. Perform model evaluation on multiple target machine learning models to obtain the goodness of fit of the multiple target machine learning models;

[0112] S36. Based on the goodness of fit of the multi-objective machine learning model, the contribution of each target influencing factor to runoff in the multi-objective machine learning model is calculated using a model interpretability algorithm;

[0113] S37. Based on the fitting results of the monthly regression equation group sub-models and multiple target machine learning models, assign corresponding weights to different models;

[0114] S38. Based on the assigned weights, the contribution rate of each target influencing factor to runoff in different models is calculated;

[0115] In this embodiment, in the monthly regression equation group sub-model, the number of sub-models for each month is 2. n-1 A total of 2 n-1 × sub-models, where the regression equations for a certain month It can be expressed as:

[0116] (4)

[0117] Analyze the goodness of fit of each target influencing factor to the sub-model of the monthly regression equation group based on the relative importance method , calculate the average contribution increment of the goodness of fit , where the goodness of fit It can be expressed as:

[0118] (5)

[0119] In formula (5), TSS is the total sum of squares of deviations; RSS is the regression sum of squares; ESS is the residual sum of squares;

[0120] Based on goodness of fit The average contribution increase , and obtain the contribution of each target influencing factor to runoff in the monthly regression equation group sub-model , as follows:

[0121] (6)

[0122] Build using XGboost, LSTM or random forest intelligent prediction methods Machine learning models , as follows:

[0123] ① The runoff of the target hydrological station Matrix of monthly factors affecting the target Divide into training sets in proportion With the test set ;

[0124] ② Use the training set to train each machine learning model to obtain multiple target machine learning models , and input the test set into multiple target machine learning models to obtain Prediction results of runoff data using a machine learning model;

[0125] right The model evaluation is performed on the target machine learning model to obtain the goodness of fit of multiple target machine learning models. ;

[0126] The model interpretability algorithm is used to calculate the Contribution of each target influencing factor to runoff in a target learning model , as follows:

[0127] (7)

[0128] The basic principle of the model interpretability algorithm is to calculate the The Shapely values ​​of the target impact factors are calculated, that is, a target impact factor is calculated. The weighted average marginal contribution when added to multiple target learning models is obtained. The contribution of the target influencing factor in the target machine learning model;

[0129] Goodness of fit of sub-models based on monthly regression equations and goodness of fit of machine learning models with multiple objectives ,get( ) Model fitting effect ;

[0130] based on( ) Model fitting effect , giving ( ) models with different weights get corresponding weights :

[0131] (8)

[0132] In formula (7), Model type, Represents the proportion of the fitting effect of a certain model to the sum of the fitting effects of all models;

[0133] Based on weight , calculate the contribution rate of each target influencing factor to runoff, that is, the weighted average final contribution rate of each target influencing factor , the specific calculation is as follows:

[0134] (9);

[0135] S4. Calculate snowmelt runoff:

[0136] Specifically, S41, based on the contribution rate of each target influencing factor to runoff in different models, a contribution rate matrix of the influencing factors of runoff at the target hydrological station is constructed;

[0137] S42. Screening the influencing factors related to snowmelt from the target influencing factors to obtain the influencing factors as the target snowmelt influencing factors;

[0138] S43, multiplying the sum of the contribution rates corresponding to the target snowmelt influencing factors by the runoff data of the target hydrological station to obtain runoff data related to snowmelt, i.e., snowmelt runoff volume;

[0139] In this embodiment, based on ( ) The weighted average final contribution rate of each target impact factor , constructed months Contribution rate matrix of target hydrological station runoff influencing factors of target influencing factors , as follows:

[0140] (10)

[0141] Screening the influencing factors related to snowmelt from the target influencing factors to obtain the influencing factors as the target snowmelt influencing factors;

[0142] Multiply the sum of the contribution rates corresponding to the target snowmelt impact factors by the runoff data of the target hydrological station. , get the runoff data related to snowmelt, that is, snowmelt runoff , the specific calculation is as follows:

[0143] (11)

[0144] In formula (11), is the year corresponding to the target impact factor, is the month corresponding to the target impact factor, is the number of target snowmelt influencing factors;

[0145] As shown in Table 2, it shows the contribution rate of the factors affecting runoff in each month at a station. When calculating the snowmelt runoff, the sum of the contribution rate of the previous month's snow accumulation and the current month's snow accumulation is selected as the snowmelt contribution rate, as follows:

[0146] Table 2 Contribution rate of factors affecting runoff in each month at a station

[0147]

[0148] The present invention fully considers the complexity of runoff composition, uses the partial correlation method to eliminate the influence of other influencing factors on the runoff data of the target hydrological station, and considers different influencing factors for the runoff in different months; considers the fitting effect of different models and the contribution rate of each target influencing factor to runoff in different models, constructs a multi-model group of monthly regression equation sub-models and multiple target machine learning models, and proposes a weighted average multi-model integrated contribution rate calculation method based on the fitting effect of different models, which effectively overcomes the non-universality of the contribution rate of the influencing factors on a single model and enhances the scientificity and interpretability of the current snowmelt runoff calculation method.

[0149] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the principles and essence of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A snowmelt runoff calculation method based on meteorological factor advantage analysis, characterized in that: The specific steps include: S1. Obtain runoff data of target hydrological station; S2. Based on the runoff data of the target hydrological station, the target influencing monthly factor matrix is ​​constructed; S3. Construct multiple models and calculate the contribution rate of each target influencing factor to runoff in different models: S31. Based on the runoff data of the target hydrological station and the target influencing monthly factor matrix, a sub-model of the monthly regression equation group including the target influencing factors is established; S32. Analyze the goodness of fit of each target influencing factor to the sub-model of the monthly regression equation group based on the relative importance method, and calculate the average contribution increment of the goodness of fit; S33. Based on the average contribution increment of the goodness of fit, the contribution of each target influencing factor to runoff in the sub-model of the monthly regression equation group is obtained; S34. Based on the target hydrological station runoff and the target influencing monthly factor matrix, multiple target machine learning models are constructed; S35. Perform model evaluation on multiple target machine learning models to obtain the goodness of fit of the multiple target machine learning models; S36. Based on the goodness of fit of the multi-objective machine learning model, the contribution of each target influencing factor to runoff in the multi-objective machine learning model is calculated using a model interpretability algorithm; S37. Based on the fitting results of the monthly regression equation group sub-models and multiple target machine learning models, assign corresponding weights to different models; S38. Based on the assigned weights, the contribution rate of each target influencing factor to runoff in different models is calculated; S4. Calculate snowmelt runoff: S41. Based on the contribution rate of each target influencing factor to runoff in different models, a matrix of the contribution rate of the influencing factors of runoff at the target hydrological station is constructed; S42. Screening the influencing factors related to snowmelt from the target influencing factors to obtain the influencing factors as the target snowmelt influencing factors; S43. Multiply the sum of the contribution rates corresponding to the target snowmelt influencing factors by the runoff data of the target hydrological station to obtain runoff data related to snowmelt, i.e., snowmelt runoff volume.

2. The snowmelt runoff calculation method based on meteorological factor advantage analysis according to claim 1 is characterized in that: S1 specifically includes the following steps: S11. Determine the target area and the hydrological stations within the target area; S12, selecting a hydrological station in the target area as a target hydrological station; S13. Select the long-term monthly runoff data of the target hydrological station as the runoff data of the target hydrological station: Assume that the year in the long series monthly runoff data is , the month is , then the runoff data of the target hydrological station is ,in , =1,2,3...,12, .

3. The snowmelt runoff calculation method based on meteorological factor advantage analysis according to claim 1 is characterized in that: S2 specifically includes the following steps: S21. Based on the runoff data of the target hydrological station, select the meteorological factors that affect the runoff of the target hydrological station as the target influencing factors; S22, arranging the target impact factors in the order of year and month corresponding to the runoff data of the target hydrological station to construct a target impact factor data set; S23. Based on the target influencing factor dataset, the partial correlation method is used to calculate the lead-lag partial correlation coefficient between the target hydrological station runoff data and each target influencing factor, and a partial correlation coefficient tensor is constructed; S24. Based on the partial correlation coefficient tensor, find the month corresponding to the maximum partial correlation coefficient between each target influencing factor and the runoff of the target hydrological station to obtain the target influencing monthly factor of the runoff of the target hydrological station; S25, obtaining the maximum partial correlation coefficient matrix of monthly runoff based on the months in S24; S26. Based on the target influencing monthly factors and the monthly runoff maximum partial correlation coefficient matrix, a target influencing monthly factor matrix is ​​constructed.

4. The snowmelt runoff calculation method based on meteorological element advantage analysis according to claim 3 is characterized in that: The target impact factor dataset , as follows: （1） In formula (1), is the total number of target impact factor types; is the year corresponding to the target impact factor, is the month corresponding to the target impact factor, where 、 Keep corresponding to the value in S13; The partial correlation coefficient tensor , as follows: （2） In formula (2), The monthly runoff data from January to December of the target hydrological station and The correlation between different months corresponding to the target impact factors; The maximum partial correlation coefficient matrix , as follows: （3） In formula (3), To influence the Monthly runoff The maximum monthly partial correlation coefficient of the target influencing factors; Based on the maximum partial correlation coefficient matrix of the target influencing monthly factors and monthly runoff , construct the target impact monthly factor matrix .

5. The snowmelt runoff calculation method based on meteorological element advantage analysis according to claim 1 is characterized in that: In the monthly regression equation group sub-model, the number of sub-models for each month is 2 n-1 A total of 2 n-1 × sub-models, where the regression equations for a certain month It can be expressed as: （4） The goodness of fit It can be expressed as: （5） In formula (5), TSS is the total sum of squares of deviations; RSS is the regression sum of squares; ESS is the residual sum of squares; Contribution of each target influencing factor to runoff in the monthly regression equation sub-model , as follows: （6）。 6. The snowmelt runoff calculation method based on meteorological element advantage analysis according to claim 1 is characterized in that: The process of constructing the multiple target machine learning models specifically includes the following steps: 1) Use XGboost, LSTM or random forest intelligent prediction methods to build Machine learning models , as follows: ① The runoff of the target hydrological station Matrix of monthly factors affecting the target Divide into training sets in proportion With the test set ; ② Use the training set to train each machine learning model to obtain multiple target machine learning models , and input the test set into multiple target machine learning models to obtain Prediction results of runoff data using a machine learning model; 2) Yes The model evaluation is performed on the target machine learning model to obtain the goodness of fit of multiple target machine learning models. ; 3) Calculated by the model interpretability algorithm Contribution of each target influencing factor to runoff in a target learning model , as follows: （7） 4) Goodness of fit of the sub-model based on the monthly regression equation group and goodness of fit of machine learning models with multiple objectives ,get( ) Model fitting effect ; 5) Based on ( ) Model fitting effect , giving ( ) models with different weights get corresponding weights : （8） In formula (7), Model type, Represents the proportion of the fitting effect of a certain model to the sum of the fitting effects of all models; 6) Based on weight , calculate the contribution rate of each target influencing factor to runoff, that is, the weighted average final contribution rate of each target influencing factor , the specific calculation is as follows: （9）。 7. The snowmelt runoff calculation method based on meteorological element advantage analysis according to claim 1 is characterized in that: The contribution rate matrix of the factors affecting the runoff of the target hydrological station , as follows: （10）。 8. The snowmelt runoff calculation method based on meteorological element advantage analysis according to claim 1 is characterized in that: The snowmelt runoff , the specific calculation is as follows: （11） In formula (11), is the year corresponding to the target impact factor, is the month corresponding to the target impact factor, is the number of target snowmelt influencing factors.

Images