Hydrological model parameter calibration method based on automatic session flood segmentation

Abstract

The invention discloses a hydrological model parameter calibration method based on automatic session flood segmentation, and belongs to the field of runoff forecasting. The method makes full use of the data characteristics of the drainage basin flow process, proposes a method for segmenting session flood according to the flow percentage, deduces a target function weighted by the volume of session flood on the basis of the method, and obtains a hydrological model parameter calibration result. Hydrological model parameter calibration with session flood as a target is realized; according to the method, on the basis of an original hydrological model parameter calibration method, the high flow proportion in the parameter calibration process is adjusted, the defect that the flood forecast error of the flood period session is large due to the fact that the low flow proportion in a traditional parameter calibration method is too large is overcome, and the flood forecast accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the field of runoff forecasting, and more specifically, relates to a method for calibrating hydrological model parameters based on automatic segmentation of flood events. Background Art

[0002] High-precision hydrological forecasting can provide decision-making basis for reservoir operation, flood control and disaster reduction, and optimal allocation of water resources. The flood peak forecasting accuracy of flood events during the flood season plays an important role in disaster prevention and reduction, flood forecasting and early warning, etc. There are a large number of parameters in the hydrological forecasting model, and some parameters cannot be estimated based on measured data. Selecting appropriate calibration samples and objective functions for calibration is the key to improving the forecasting accuracy of the hydrological model.

[0003] The application of traditional parameter calibration methods has certain limitations. The main problem is that the method steps of manual parameter calibration are complex and require high experience from hydrological workers. In this context, with the development of computer technology, the automatic optimization technology based on optimization algorithms is widely popular because of its high efficiency and simple operation. However, the existing parameter calibration methods have defects: considering the entire flow process during the calibration period, the low flow proportion in the dry season is too large, which affects the forecasting accuracy during the flood season. Adopting the way of manually segmenting the flow process during the flood season increases the workload of parameter calibration and cannot make full use of the convenience of computer programming technology.

[0004] Therefore, the existing methods have problems of large workload of manually selecting flood events and too high proportion of low flow in the calibration period data, resulting in low forecasting accuracy of flood events during the flood season. Summary of the Invention

[0005] In view of the above defects or improvement requirements of the prior art, the present invention provides a method for calibrating hydrological model parameters based on automatic segmentation of flood events, aiming to solve the technical problems of too high proportion of low flow in the flow sequence and large error of flood events during the flood season existing in the existing parameter calibration methods.

[0006] To achieve the above object, the technical solution of the present invention is as follows:

[0007] A method for calibrating hydrological model parameters based on automatic segmentation of flood events, comprising the following steps:

[0008] S1. Determine the peak flow threshold Q p , start and end flow thresholds Q b ;

[0009] S2. Peak flow derivation for the flow segment higher than the peak flow threshold Q p , including extracting the high flow process Q p higher than the peak flow threshold Q t,j from the historical data flow sequence, and from the extracted flow exceeding Q pDetermine the peak flow rate during the flood process And record the peak occurrence time t p,j , the expression is:

[0010]

[0011] In the formula, Is the peak flow rate of the j-th flood, m 3 / s; Q t,j Represents the j-th flow process with a flow rate exceeding Q p ;

[0012] S3. Deduce the start and end points of the flood events, including the coordinates of each peak obtained from step S2 Deduce the start point of the flood event backward in time from the coordinates of each peak, and record the start point coordinates of the flood event And deduce the end point of the flood event forward in time, and record the end point coordinates of the flood event And conduct a rationality check on the flood event division;

[0013] S4. Based on the peak flow Corresponding to the end time t of the flood event e,j , and the peak flow Corresponding to the start time t of the flood event s,j Calculate the flood volume V of each flood event i , using the flood volume V i And the peak flow as weights to form the objective function Obj for parameter calibration, seek the minimum value of the objective function Obj, and use the parameters when Obj takes the minimum value as the optimal parameters of the calibrated Xin'anjiang model;

[0014] S5. Use the calibrated model parameters and combine with the Xin'anjiang model to calculate the basin flood process.

[0015] Furthermore, in step S1, select continuous multi-year historical data, and select the top p% of the flow data from the historical peak flow data according to the magnitude of the flow values. The top p% of the flow data is high-flow data, and the smallest flow value among the top p% of the flow data is used as the peak flow threshold Q p ;

[0016] Select the last b% of the flow data from the continuous historical peak flow data according to the magnitude of the flow values. The last b% of the flow data is low-flow data, and select the flow data with an exceedance probability of b% as the start and end flow threshold Q b .

[0017] Furthermore, for the peak flow derivation in the flow segment higher than the peak flow threshold in step S2, specifically, extract the flow sequence higher than the peak flow threshold Q pThe high-flow process Q t,j :

[0018] Q t,j ={Q t ≥Q p |t = 1, 2, 3, ..., T} (1)

[0019] Where: Q t,j represents the j-th flow process with a flow exceeding Q p , m 3 / s, j = 1, 2, …, n, where n is the number of high-flow processes segmented by Q p ; t is time, and T is the duration of the flow sequence.

[0020] Furthermore, in step S3, the starting point and ending point of the flood event are found according to the derived flood peak position, specifically including:

[0021] S31: Starting from the flood peak coordinates obtained in step S2 and pushing back in time to find the starting point of the flood event. For the flow process Q t , set the sliding window width L = T × c%, where c is the percentage for calculating the sliding window width. When the flow value is lower than the low-flow reference value Q b , and the slope k of the flow within the reference length L is less than the given threshold d, then it is determined that the flood starts to rise, and the starting point coordinates of the flood event are recorded

[0022]

[0023] Where k is the slope of the sliding window; m = 1, 2, … t p,j , representing the movement of the sliding window back in time; Δt is the time step of the flow sequence; T S is the coordinate set of the starting positions of all flood events; t s,j is the starting time of the flood event corresponding to the flood peak ; is the flow value at the starting point of the flood event, m 3 / s;

[0024] Continuous historical data forms a curve. The abscissa of a point on the curve is the time point, and the ordinate is the flow corresponding to the time point. The points on the curve that meet the requirements of formula (3) and formula (4) are the starting points of the flood events;

[0025] S32: Starting from the flood peak coordinates obtained in step S2 and pushing forward in time to find the ending point of the flood event. When the flow value is lower than the low-flow reference value Q bWhen the slope k of the flow rate within the reference length L is greater than the given threshold -d, it is determined that the flood has ended, and the coordinates of the end point of the flood event are recorded.

[0026]

[0027] In the formula, m = 1, 2, … T - t p,j , representing the movement of the sliding window in the positive time direction; T e is the coordinate set of the end positions of all flood events; t e,j is the flood peak corresponding to the end time of the flood event; is the flow rate value at the end point of the flood event, m 3 / s;

[0028] The continuous historical data forms a curve. The abscissa of a certain point on the curve is the time point, and the ordinate is the flow rate corresponding to the time point. The point on the curve that satisfies the requirements of formula (5) and formula (6) is the end point of the flood event;

[0029] S33: When the start and end times t s,j corresponding to the multiple flood peaks derived are the same as t e,j , the division of the flood events is reasonable, and they are merged into a multi-peak flood. When the start point of the next flood event is earlier than the end point of the previous flood event, the division of the flood events is unreasonable. The lowest flow rate value at the overlapping part of the two floods should be used as the splitting point to re-divide the start and end points, and the coordinates of the start and end points are recorded.

[0030] Furthermore, the calculation method of the flood volume V of each flood event and the objective function Obj for parameter calibration formed with the flood volume and flood peak as weights are respectively: i In the formula, V

[0031]

[0032] In the formula, V i is the flood volume of each flood event, Obj i is the objective function of each flood event, ω i is the weight of the objective function of each flood event, Obj is the objective function for parameter calibration, and n is the number of flood events.

[0033] Furthermore, the specific method for calibrating the parameters of the Xin'anjiang model in step S4 is to divide the basin rainfall-runoff process into a calibration period and a verification period based on historical data, perform calculations and verifications based on the data in the calibration period and the verification period, use Obj in step S4 as the objective function, seek the minimum value of the objective function Obj, and take the parameters when Obj takes the minimum value as the optimal parameters of the Xin'anjiang model.

[0034] The beneficial effects of the present invention are as follows:

[0035] 1. Aiming at the defect that the proportion of low-flow data is too large in the traditional method for calibrating hydrological model parameters, the present invention introduces a method for automatically segmenting flood events, and uses the segmented flood events in the flow process during the calibration period for calibrating the model parameters, which improves the attention of the optimization algorithm to the prediction accuracy of high flows during the flood season.

[0036] 2. The present invention makes full use of the characteristics of flow data itself, and all the parameters required for segmenting flood events are obtained from the flow data, which improves the universality of the method for segmenting flood events.

[0037] 3. The present invention proposes a method for calibrating hydrological model parameters based on automatically segmenting flood events, and derives the calculation formula for the start and end points of flood events based on flow data. Based on flood events, the hydrological model parameters are calibrated more precisely, which improves the accuracy of flood events during the flood season in flood forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Flowchart of the method for calibrating hydrological model parameters based on automatically segmenting flood events according to the present invention.

[0039] Figure 2 Map of the study area provided by the embodiment of the present invention.

[0040] Figure 3 Schematic diagram of segmenting flood events in the study area in 1990.

[0041] Figure 4 Comparison chart of the prediction results in the study area in 1990.

[0042] Figure 5 Comparison chart of the prediction results of each flood event in the study area in 1990. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to make the purpose, technical solutions and advantages of the invention clearer, the present invention will be further described below with reference to the accompanying drawings.

[0044] Refer to Figure 1 , the present invention provides a method for calibrating hydrological model parameters based on automatically segmenting flood events, including:

[0045] (1) Calculate the peak flow threshold Q p and the start and end flow thresholds Q b ;

[0046] The specific methods for calculating the peak flow threshold and the start and end flow thresholds in step (1) are as follows:

[0047] Select a high-flow reference percentage p% and a low-flow reference percentage b%, and obtain a peak flood flow threshold Q with an exceedance probability of p% p and a start-stop flow threshold Q with an exceedance probability of b% b . Here, p = 2 and b = 90 are taken.

[0048] The present invention selects historical data for consecutive years. According to the magnitude of the flow values, the top p% of the flow data is selected from the historical peak flood flow data. The top p% of the flow data is high-flow data, and the smallest flow value among the top p% of the flow data is used as the peak flood flow threshold Q p .

[0049] According to the magnitude of the flow values, the bottom b% of the flow data is selected from the consecutive historical peak flood flow data. The bottom b% of the flow data is low-flow data. As an implementation manner of the present invention, the largest flow value among the b% of the flow data is selected as the start-stop flow threshold Q b ; As another implementation manner of the present invention, the flow data with an exceedance probability of b% is selected as the start-stop flow threshold Q b . When the flood flow value is lower than Q b , the rationality determination of the start and end points of the flood event in step (3) is performed.

[0050] (2) Peak flood derivation for the flow segment above the peak flood flow threshold;

[0051] The method for deriving the peak flood for the flow segment above the peak flood flow threshold described in step (2) is as follows:

[0052] Extract the high-flow process Q above the peak flood flow threshold Q p from the historical data flow sequence: t,j :

[0053] Q t,j ={Q t ≥Q p |t = 1, 2, 3,..., T} (1)

[0054] In the formula: Q t,j represents the jth flow process with a flow exceeding Q p , m 3 / s, j = 1, 2,..., n, and n is the number of high-flow processes segmented by Q p ; t is time, and T is the duration of the flow sequence.

[0055] Furthermore, determine the peak flood flow from the extracted over-quantified flood process and record the peak occurrence time t p,j . The expression is as follows:

[0056]

[0057] In the formula, is the peak flow of the j-th flood, m 3 / s.

[0058] (3) Determination of the start and end points of individual floods;

[0059] The method for determining the start and end points of individual floods described in step (3) is as follows:

[0060] 01. Determine the peak The start time t s,j of the corresponding individual flood and the start point flow value Specifically, starting from the peak coordinates obtained in step S2 and moving backward in time to determine the start point of the individual flood. For the flow process Q t , set the sliding window width L = T × c%, where c is the percentage for calculating the sliding window width. When the flow value is lower than the low flow reference value Q b , and the slope k of the flow within the reference length L is less than the given threshold d, then it is determined that the flood starts to rise, and the start point coordinates of the individual flood are recorded Here, c = 1 and d = 1 are taken.

[0061] A curve is formed by continuous historical data. The abscissa of a certain point on the curve is the time point, and the ordinate is the flow corresponding to the time point. The point on the curve that meets the requirements of formula (3) and formula (4) is the start point of the individual flood.

[0062]

[0063] In the formula, k is the slope of the sliding window; m = 1, 2,... t p,j , representing the movement of the sliding window backward in time; Δt is the time step of the flow sequence; T S is the coordinate set of the start positions of all individual floods; t s,j is the start time of the individual flood corresponding to the peak ; is the start point flow value of the individual flood, m 3 / s.

[0064] 02. Starting from the peak coordinates obtained in step S2 and moving forward in time to determine the end point of the individual flood. When the flow value is lower than the low flow reference value Q b , and the slope k of the flow within the reference length L is greater than the given threshold -d, then it is determined that the flood ends, and the end point coordinates of the individual flood are recorded

[0065] The continuous historical data form a curve. The abscissa of a certain point on the curve is the time point, and the ordinate is the flow rate corresponding to the time point. The points on the curve that meet the requirements of formula (5) and formula (6) are the end points of the flood events.

[0066]

[0067] where m = 1, 2, … T - t p,j , representing the movement of the sliding window in the positive time direction; T e is the coordinate set of the end positions of all flood events; t e,j is the peak flood corresponding to the end time of the flood event; is the flow rate value at the end point of the flood event, m 3 / s.

[0068] (4) Rationality check of flood events;

[0069] The specific method for the rationality check of the flood events described in step (4) is as follows:

[0070] When the start and end times t s,j and t e,j corresponding to the multiple peak floods derived in step (3) are the same, the division of the flood events is reasonable, and they are merged into a multi-peak flood. When the start point of the next flood event is earlier than the end point of the previous flood event, the division of the flood events is unreasonable. The lowest flow rate value at the overlapping part of the two floods should be used as the splitting point to re-divide the start and end points and record the coordinates of the start and end points.

[0071] (5) Form an objective function with the accuracy of flood events as the goal;

[0072] The specific method for forming an objective function with the accuracy of flood events as the goal described in step (5) is as follows:

[0073] Based on the end time t of the flood event corresponding to the peak flood obtained in step (3), e,j and the start time t of the flood event corresponding to the peak flood s,j calculate the flood volume V i of each flood event, and form an objective function Obj for parameter calibration with the flood volume and the peak flood as weights.

[0074]

[0075] where V i is the flood volume of each flood event, Q t is the flow process; Δt is the time step of the flow sequence; Obj i is the objective function of each flood event, ωi is the weight of the objective function for each flood event, Obj is the objective function for parameter calibration, and n is the number of flood events.

[0076] (6) Hydrological model parameter calibration and flood forecasting

[0077] For the hydrological model parameter calibration and flood forecasting described in step (6), the specific method is as follows:

[0078] Based on historical data, divide the basin rainfall-runoff process into a calibration period and a verification period. Calculate and verify according to the data in the calibration period and the verification period. Use the flood event segmentation method proposed in the present invention to derive flood events. Adopt the SCE-UA algorithm, with Obj in step (5) as the objective function, calculate the value of the objective function Obj using flood event data and different parameter values, seek the minimum value of the objective function Obj, and take the parameters when Obj takes the minimum value as the optimal parameters of the Xin'anjiang model. Combine with the Xin'anjiang model to calculate the basin flood process.

[0079] Taking the Muma River Basin as the research object, the research area is as Figure 2 shown. Select the historical flood data from 1980 to 1990 for model parameter calibration. Among them, 1980 - 1989 is the calibration period, 1990 is the verification period, the simulation time step is 24h, the high flow percentage p% is taken as 2%, the low flow percentage b% is taken as 90%, the sliding window length percentage c% is taken as 1%, and the slope threshold d is taken as 1. Among them, the division of flood events in 1990 is as Figure 3 shown. A total of three flood events, namely 19900509, 19900613, and 19900809, are divided. The traditional method and the method proposed in the present invention are respectively used for parameter calibration and flood forecasting. The parameters of the Xin'anjiang model to be calibrated are shown in the column of parameter names in Table 1.

[0080] The parameter calibration results are shown in Table 1. The comparison of the forecasting results in 1990 is as Figure 4 shown. The comparison of the flood event No. 19900509 is as Figure 5 (a) shown. The comparison of the flood event No. 19900613 is as Figure 5 (b) shown. The comparison of the flood event No. 19900809 is as Figure 5 (c) shown. The evaluation indexes of each flood event are shown in Table 2. It can be seen from the table that the relative errors of the flood peaks of the flood events No. 19900509 and No. 19900613 have been significantly reduced. Generally, the accuracy of the proposed forecasting method is better than that of the traditional parameter calibration method.

[0081] Table 1 Xin'anjiang model parameter calibration results

[0082]

[0083]

[0084] Table 2 Evaluation Indexes of Floods in Each Event in 1990

[0085]

[0086]

[0087] Finally, it should be noted that the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art. The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements on some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for calibrating hydrological model parameters based on automatically segmented flood events, characterized in that, It includes the following steps: S1. Determine the peak flood flow threshold Q p and the start and end flow thresholds Q b ; S2. Higher than the peak flow threshold Q p Derivation of the peak flow for the flow segment, including extracting the high-flow process Q higher than the peak flow threshold Q from the historical data flow sequence p Determining the peak flow from the flood process with the extracted flow exceeding Q t,j and recording the peak occurrence time t p , the expression is: p,j ​​ In the formula, is the peak discharge of the j-th flood, m 3 / s; Q t,j represents the j-th flow process with a flow exceeding Q p ; S3. Determination of the start and end points of the flood events for each flood peak, including the coordinates of each flood peak obtained from step S2 Backwardly deduce the start point of the flood event in the reverse direction of time, and record the coordinates of the start point of the flood event And forwardly deduce the end point of the flood event in the forward direction of time, and record the coordinates of the end point of the flood event And perform a rationality check on the division of flood events; S4. According to the flood peak obtained in step S3 The ending time t of the corresponding flood event e,j and the flood peak The starting time t of the corresponding flood event s,j Calculate the flood volume V of each flood event i Take the flood volume V i and the flood peak as weights to form the objective function Obj for parameter calibration, seek the minimum value of the objective function Obj, and take the parameters when Obj takes the minimum value as the optimal parameters of the calibrated Xin'anjiang model; S5. Using the calibrated model parameters and combining with the Xin'anjiang model, calculate the flood process of the basin.

2. A method for calibrating hydrological model parameters based on automatically segmented flood events according to claim 1, characterized in that : In the step S1, historical data of multiple consecutive years is selected. According to the magnitude of the flow values, the top p% of the flow data is selected from the historical peak flow data. The top p% of the flow data is high-flow data, and the smallest flow value among the top p% of the flow data is used as the peak flow threshold Q p ; Select the flow data of the last b% in the continuous historical peak flow data according to the magnitude of the flow value. The flow data of the last b% is low flow data, and select the flow data with an exceedance probability of b% as the start and end flow threshold Q b .

3. A method for calibrating hydrological model parameters based on automatically segmented flood events according to claim 1, characterized in that: The calculation of the flood peak in the flow segment higher than the flood peak flow threshold in the S2 step is specifically as follows: extract the high-flow process Q in the flow sequence that is higher than the flood peak flow threshold Q p : t,j : Q t,j = {Q t ≥ Q p | t = 1, 2, 3, ..., T} (1) Where: Q t,j represents the j-th flow process with a flow rate exceeding Q p in m 3 / s, j = 1, 2, …, n, where n is the number of high-flow processes segmented by Q p ; t is the time and T is the duration of the flow rate sequence.

4. A method for calibrating hydrological model parameters based on automatically segmented flood events according to claim 1, characterized in that: The S3 step finds the starting point and the ending point of the flood event according to the derived flood peak position, specifically including: S31: Backwardly deduce the starting point of the storm flood from the peak coordinates obtained in step S2 for the flow process Q Set the sliding window width L = T × c%, where c is the percentage for calculating the sliding window width. When the flow value is lower than the low flow reference value Q t , and the slope k of the flow within the reference length L is less than the given threshold d, then it is determined that the flood starts to rise, and the starting point coordinates of the storm flood are recorded b ​ where k is the slope of the sliding window; m = 1, 2, … t p,j , representing the backward movement of the sliding window in time; Δt is the time step of the flow rate sequence; T S is the coordinate set of the starting positions of all flood events; t s,j is the peak flood corresponding to the starting time of the flood event; is the flow rate value at the starting point of the flood event, m 3 / s; The continuous historical data forms a curve. The abscissa of a certain point on the curve is the time point, and the ordinate is the flow corresponding to the time point. The point on the curve that meets the requirements of formula (3) and formula (4) is the starting point of the flood event; S32: The peak coordinates obtained from step S2 Push forward to the positive direction of time to find the end point of the storm flood. When the flow value is lower than the low flow reference value Q b , and the slope k of the flow within the reference length L is greater than the given threshold -d, it is determined that the flood has ended, and the end point coordinates of the storm flood are recorded where \(m = 1, 2, \cdots T - t\) p,j , representing the forward movement of the sliding window in time; \(T\) e is the coordinate set of the end positions of all flood events; \(t\) e,j is the peak flood corresponding to the end time of the flood event; is the flow value at the end point of the flood event, \(m\) 3 / s; The continuous historical data forms a curve. The abscissa of a certain point on the curve is the time point, and the ordinate is the flow corresponding to the time point. The point on the curve that meets the requirements of formula (5) and formula (6) is the ending point of the flood event; S33: When the start and end times t corresponding to multiple peak floods derived s,j and t e,j are the same, the division of flood events is reasonable, and they are merged into a multi-peak flood. When the start point of the next flood event is earlier than the end point of the previous flood event, the division of flood events is unreasonable. The lowest flow value at the overlapping part of the two floods should be used as the splitting point to re-divide the start and end points, and record the coordinates of the start and end points.

5. A method for calibrating hydrological model parameters based on automatically segmented flood events according to claim 1, characterized in that: The flood volume V of each flood event i The calculation method and the objective function Obj for parameter calibration formed with flood volume and flood peak as weights are respectively as follows: Wherein, V i is the flood volume of each flood event, Obj i is the objective function of each flood event, ω i is the weight of the objective function of each flood event, Obj is the objective function for parameter calibration, and n is the number of flood events.

6. A method for calibrating hydrological model parameters based on automatically segmented flood events according to claim 1, characterized in that: The specific method for calibrating the Xin'anjiang model parameters in the S4 step is to divide the basin rainfall-runoff process into a calibration period and a verification period based on historical data, perform calculations and verifications based on the data in the calibration period and the verification period, take Obj in step S4 as the objective function, seek the minimum value of the objective function Obj, and use the parameters when Obj takes the minimum value as the optimal parameters of the Xin'anjiang model.

Images