CN-122022025-A - Method, system and medium for forecasting flow-domain sand production in consideration of rainfall accumulation effect

Abstract

The invention discloses a method, a system and a medium for forecasting flow-domain sand production, which take rainfall accumulation effect into consideration, wherein the method comprises data acquisition, early-stage influence rainfall calculation, exponential decay type early-stage rainfall index API t calculation, threshold value set construction, determination of a short window rainfall high threshold value P 3,high and a short window rainfall low threshold value P 3,low, , determination of a long window rainfall high threshold value P L,high and a long window rainfall low threshold value P L,low , sand supply state representation, event type judgment rules, event type judgment, continuous event sand reduction and recovery, cold start treatment, parameter calibration, result output, and introduction of a multi-scale early-stage influence rainfall index, sand supply state representation and continuous event sand reduction factor.

Inventors

• CHEN ZHONGXIAN
• CHEN FANG
• ZHANG BO
• ZENG BIN
• LIU FANGZHI
• REN YI
• LI SHUYAN
• WANG ZHILI
• ZHAO MINGLIANG
• ZHU LINGLING
• ZHENG CHUANDONG
• JIN YE
• GAO CE
• XIANG HENG
• Feng Shenghang

Assignees

• 中国长江三峡集团有限公司
• 三峡金沙江川云水电开发有限公司
• 长江水利委员会水文局

Dates

Publication Date

20260512

Application Date

20260120

Claims (10)

1. 1. A method for forecasting the sand production in a flow domain by considering the rainfall accumulation effect is characterized by comprising the following steps: Step 1, data acquisition, namely acquiring a precipitation sequence P, a flow sequence Q and an observed sand conveying quantity sequence S of a target river basin, wherein the duration time and the time resolution of the sequences { P, Q and S } are required to be the same, and meanwhile, missing measurement value rejection and time alignment of the sequences { P, Q and S } are completed; step 2, calculating early-stage influence rainfall, namely calculating short window accumulated rainfall P 3,t and long window accumulated rainfall P L,t according to a rainfall sequence P, wherein the window L epsilon [15,30] days and parameter calibration are needed, and calculating an exponential decay type early-stage rainfall index API t ; Step 3, constructing a threshold set, namely calculating specified probability quantiles of a flow sequence Q, wherein the specified probability quantiles comprise 5%, 25%, 75% and 95%, corresponding flow values are respectively determined to be a flow threshold Q 5 , Q 25 , Q 75 , Q 95 , sorting a whole sample data set of short window accumulated rainfall P 3,t , determining a short window rainfall high threshold P 3,high and a short window rainfall low threshold P 3,low according to a group of preset quantile values, sorting a whole sample data set of long window accumulated rainfall P L,t , determining a long window rainfall high threshold P L,high and a long window rainfall low threshold P L,low according to the group of preset quantile values, and optimally adjusting the group of values in a model rating stage; step 4, sand supply state representation, namely calculating a sand supply state X t , and introducing a first-order memory coefficient gamma epsilon (0, 1); Step 5, event type judging rules, namely setting a high intensity threshold r hi by taking a baseline function S m (Q t ) as a reference, defining a high sand supply condition as meeting any condition of P 3,t ≥ P 3,high 、P L,t ≥ P L,high or X t ≥ X high , wherein the value range of the high sand supply threshold X high is [0.6,0.9] and the parameters are required to be set, defining a low sand supply condition as meeting any condition of P 3,t ≤ P 3,low 、P L,t ≤ P L,low or X t ≤ X low , wherein the value range of the low sand supply threshold X low is [0.1,0.4], and the parameters are also required to be set; And 6, judging the event type, namely judging a high-strength sand conveying event if the observed sand conveying amount meets S t / S m (Q t ) ≥ r hi and is simultaneously in a high sand supply condition, judging a large water xiao Sha event if the observed sand conveying amount does not meet S t / S m (Q t ) ≥ r hi , and the flow rate meets Q t ≥ Q 75 and is simultaneously in a low sand supply condition, and judging a small water large sand event if the observed sand conveying amount does not meet S t / S m (Q t ) ≥ r hi , and the flow rate meets Q t ≤ Q 25 and is simultaneously in a high sand supply condition. Defining phi h,t as an indication function of a 'small water and big sand' event as a 'high-strength sand conveying' event, phi A,t as an indication function of a 'small water and big sand' event, and phi B,t as an indication function of a 'big water xiao Sha' event; Step 7, continuous event sand reduction and recovery, namely counting the times N t of the accumulated high-intensity sand conveying event in the period of the near window H, calculating a sand reduction factor D t , and correcting the sand production of the next heavy rainfall by combining a recovery function g (delta t) to obtain a predicted value ; Step 8. Cold Start treatment, predictive value when t < L-1 or sufficient sample is absent The prediction result in the COLD start state does not participate in error statistics and is only used for service prompt; Step 9, parameter calibration, namely carrying out joint calibration on a parameter set { short-term precipitation state coefficient alpha s , long-term precipitation state coefficient alpha l , flow state coefficient beta, first-order memory coefficient gamma, high-strength sand conveying event amplification coefficient theta h , small water and large sand event amplification coefficient theta A , large water xiao Sha event shrinkage coefficient theta B , high-strength threshold r hi , attenuation coefficient k, window L, near window period H, sand reduction factor coefficient lambda, recovery function coefficient mu, long-window accumulated precipitation and short-window accumulated precipitation threshold quantile set } by adopting a global optimization algorithm to obtain global optimal parameters; And step 10, outputting a result, namely outputting an optimal model parameter set { short-term precipitation state coefficient alpha s , long-term precipitation state coefficient alpha l , flow state coefficient beta, first-order memory coefficient gamma, high-strength sand transmission event amplification coefficient theta h , small water and large sand transmission event amplification coefficient theta A , large water xiao Sha event shrinkage coefficient theta B , high-strength threshold r hi , attenuation coefficient k, window L, near window period H, sand reduction factor coefficient lambda, recovery function coefficient mu, long window accumulated precipitation and short window accumulated precipitation threshold quantile set }, changing step 1 into data acquisition when the model is in use period, acquiring a precipitation sequence P and a flow sequence Q of a target river basin, requiring the same duration time and time resolution of the sequence { P, Q } and simultaneously completing missing measurement value rejection and time alignment of the sequence { P, Q }, directly carrying out calculation by the parameter set after the step 1, and outputting a sand transmission sequence predicted value S.
2. 2. A method for domain sand production prediction considering rainfall accumulation effect as in claim 1 wherein the short window accumulation P 3,t and the long window accumulation P L,t are respectively calculated as follows: (1) (2)。
3. 3. A domain sand production forecasting method taking into account the cumulative effects of rainfall as defined in claim 1, wherein said API t is calculated as follows: (3) Wherein k epsilon (0, 1) is the attenuation coefficient, M is more than or equal to L.
4. 4. A method for domain sand production prediction considering rainfall accumulation effect as in claim 1 wherein the sand supply state X t is defined as: (4) wherein R (·) is a normalized mapping, alpha s is a short-term precipitation state coefficient, alpha l is a long-term precipitation state coefficient, beta is a flow state coefficient, and the three coefficients meet constraint conditions that alpha s , α l , beta is more than or equal to 0 and alpha s + α l +beta=1; The sand state memory updating mode is that ;; The calculation mode of the normalization mapping is as follows: (5) Where x is any amount in the sequence, x min is the minimum in the sequence, and x max is the maximum in the sequence.
5. 5. The method for domain sand production prediction considering rainfall accumulation effect according to claim 1, wherein the baseline function S m (Q) is quadratic or power law, and is obtained by two-step robust screening, wherein the first step is to perform primary fitting by using quadratic or power law: ,, , And the second step is to remove samples with abnormal ratio S/S m ∉ [0.5,2] and then fit the samples again by using a quadratic form or a power law form.
6. 6. A domain sand production forecasting method considering rainfall accumulation effect as in claim 1 wherein the continuous event sand reduction factor D t and the recovery function g (Δt) are defined as: (6a) (6b) Wherein, deltat is the time interval from the last high-intensity sand conveyance event, lambda is the factor coefficient of sand reduction, mu is the coefficient of recovery function, lambda, mu > 0, and the parameter calibration is needed to participate.
7. 7. The method for predicting sand yield in a flow domain taking into account the cumulative effect of rainfall as defined in claim 1, wherein the sand yield prediction formula is: (7) Wherein, θ h is the "high-intensity sand-transporting" event amplification factor, θ h ≥ 1.5;θ A is the "small water and large sand" event amplification factor, θ A > 1;θ B is the "large water xiao Sha" event shrinkage factor, θ B e (0, 1), and all three parameters are waiting parameters.
8. 8. A domain sand production forecasting method taking into account the cumulative effects of rainfall as defined in claim 1, wherein said parameter scaling objective function is: (8) Wherein σ s is the standard deviation of the observed sand transport sequence S t .
9. 9. A system for predicting the sand yield in a flow domain taking into account the effect of rainfall accumulation, comprising a memory and a processor, wherein the memory comprises a program of a method for predicting the sand yield in the flow domain taking into account the effect of rainfall accumulation, and the program of the method for predicting the sand yield in the flow domain taking into account the effect of rainfall accumulation realizes the steps of any one of claims 1 to 8 when being executed by the processor.
10. 10. A computer readable storage medium storing program code which, when executed by a processor, performs the steps of a domain sand production forecasting method taking into account the effects of rainfall accumulation as claimed in any one of claims 1 to 8.

Description

Method, system and medium for forecasting flow-domain sand production in consideration of rainfall accumulation effect Technical Field The invention relates to the field of hydrologic water resources and river sediment, in particular to a method, a system and a medium for forecasting flow-domain sand production in consideration of rainfall accumulation effect. Background The traditional water-sand relationship is adopted mostlyOr variants thereof, the assumption of "sufficient supply of movable silt" is implied, and it is difficult to characterize water-sand mismatch in the case of "limited supply of sand" or "significant accumulation/depletion of precursor sources", on the one hand, that "large water xiao Sha" is often present during drought-heavy rain transitions, and on the other hand, that small floods may trigger "small water-large sand" after successive rainfall times. Existing methods generally ignore the time-varying effects of early rain accumulation across events and material source depletion on sand delivery, resulting in systematic deviations in annual differences significantly or in multiple serial years. Disclosure of Invention The invention aims to provide a method, a system and a medium for predicting the sand production in a flow domain, which take the rainfall accumulation effect into consideration, introduce multiscale early-stage influence rainfall indexes, sand supply state representation and continuous event sand reduction factors, and the system characterizes a mismatch mechanism of 'sand supply-sand conveying capacity', so that the prediction stability and the interpretation under the conditions of different hydrologic years and multiple event sequences are improved. In order to achieve the above purpose, the present invention provides the following technical solutions: in a first aspect, an embodiment of the present application provides a method for predicting sand production in a flow domain in consideration of a rainfall accumulation effect, including the following steps: Step 1, data acquisition, namely acquiring a precipitation sequence P, a flow sequence Q and an observed sand conveying quantity sequence S of a target river basin, wherein the duration time and the time resolution of the sequences { P, Q and S } are required to be the same, and meanwhile, missing measurement value rejection and time alignment of the sequences { P, Q and S } are completed; step 2, calculating early-stage influence rainfall, namely calculating short window accumulated rainfall P 3,t and long window accumulated rainfall P L,t according to a rainfall sequence P, wherein the window L epsilon [15,30] days and parameter calibration are needed, and calculating an exponential decay type early-stage rainfall index API t; Step 3, constructing a threshold set, namely calculating specified probability quantiles of a flow sequence Q, wherein the specified probability quantiles comprise 5%, 25%, 75% and 95%, corresponding flow values are respectively determined to be a flow threshold Q 5, Q25, Q75, Q95, sorting a whole sample data set of short window accumulated rainfall P 3,t, determining a short window rainfall high threshold P 3,high and a short window rainfall low threshold P 3,low according to a set of preset quantile values, sorting a whole sample data set of long window accumulated rainfall P L,t, and determining a long window rainfall high threshold P L,high and a long window rainfall low threshold P L,low according to the set of preset quantile values, wherein the set of preset quantile values comprise 0.75 and 0.25 by default, and the set of values can be optimally adjusted in a model calibration stage; step 4, sand supply state representation, namely calculating a sand supply state X t, and introducing a first-order memory coefficient gamma epsilon (0, 1); Step 5, event type judging rules, namely setting a high intensity threshold value r hi (more than or equal to 1.5 and needing parameter calibration) by taking a baseline function S m(Qt) as a reference, defining a high sand supply condition as meeting any condition in P 3,t ≥ P3,high、PL,t ≥ PL,high or X t≥ Xhigh, wherein the value range of the high sand supply threshold value X high is [0.6,0.9] and needing parameter calibration, defining a low sand supply condition as meeting any condition in P 3,t ≤ P3,low、PL,t ≤ PL,low or X t ≤ Xlow, wherein the value range of the low sand supply threshold value X low is [0.1,0.4], and also needing parameter calibration; And 6, judging the event type, namely judging a high-strength sand conveying event if the observed sand conveying amount meets S t / Sm(Qt) ≥ rhi and is simultaneously in a high sand supply condition, judging a large water xiao Sha event if the observed sand conveying amount does not meet S t / Sm(Qt) ≥ rhi, and the flow rate meets Q t≥ Q75 and is simultaneously in a low sand supply condition, and judging a small water large sand event if the observed sand conveying amount does not meet S t / S