康?艷,伊?麗,龔家國
基于分期設(shè)權(quán)理想點(diǎn)法的水文模型參數(shù)多目標(biāo)優(yōu)化
康?艷1, 2,伊?麗1, 2,龔家國3
(1. 西北農(nóng)林科技大學(xué)水利與建筑工程學(xué)院,楊凌 712100;2. 西北農(nóng)林科技大學(xué)旱區(qū)農(nóng)業(yè)水土工程教育部重點(diǎn)實(shí)驗(yàn)室,楊凌 712100;3. 中國水利水電科學(xué)研究院,北京 100038)
水文模型參數(shù)多目標(biāo)優(yōu)化;分期設(shè)權(quán)理想點(diǎn)法;ABCD水文模型;徑流模擬;涇河流域
提高徑流預(yù)報(bào)精度對掌握流域水文情勢,實(shí)現(xiàn)水庫安全調(diào)度、水資源科學(xué)管理與優(yōu)化配置具有重要的指導(dǎo)意義[1-3].水文模型參數(shù)優(yōu)化率定對提高徑流預(yù)報(bào)精度、提升水文模型整體預(yù)報(bào)性能有著極大的推動作用,如何確定適應(yīng)流域徑流預(yù)報(bào)需求的模型參數(shù)是水文領(lǐng)域亟待解決的難點(diǎn)問題[4].傳統(tǒng)水文模型參數(shù)率定主要采用單目標(biāo)函數(shù)優(yōu)化模型參數(shù),然而研究表明[5-6],采用單目標(biāo)函數(shù)優(yōu)化模型參數(shù)僅考慮了徑流過程的某一方面的特性,無法全面刻畫水文過程中所蘊(yùn)含的各種水文特征信息,相較而言,多目標(biāo)優(yōu)化方法能夠統(tǒng)籌考慮多個(gè)目標(biāo)函數(shù)、較全面地挖掘水文資料中蘊(yùn)含的多種水文信息特征.對此國內(nèi)外學(xué)者開展了大量的研究工作[7-9],為水文模型參數(shù)優(yōu)化提供了重要的理論基礎(chǔ)和技術(shù)方法.但是多目標(biāo)優(yōu)化問題仍然存在著全局最優(yōu)解不唯一、直接求解困難、各目標(biāo)函數(shù)相互影響很難同時(shí)實(shí)現(xiàn)最優(yōu)的問題,因此,多目標(biāo)優(yōu)化求解的關(guān)鍵是在決策空間中尋求一個(gè)最優(yōu)解集,并在各目標(biāo)函數(shù)的最優(yōu)解之間進(jìn)行協(xié)調(diào)和權(quán)衡,以使各目標(biāo)函數(shù)盡可能地達(dá)到近似最優(yōu).理想點(diǎn)法[10]可將多目標(biāo)問題轉(zhuǎn)化為易求解的單目標(biāo)問題,通過構(gòu)建評價(jià)函數(shù)使各目標(biāo)函數(shù)值盡可能逼近其最優(yōu)值,為各目標(biāo)函數(shù)之間提供一種合理的平衡,有效地解決多目標(biāo)問題求解尋優(yōu)困難的問題,該方法思路簡單,適用性強(qiáng).針對徑流模擬研究中參數(shù)多目標(biāo)優(yōu)化率定時(shí)表征高流量特征的目標(biāo)函數(shù)與表征低流量的目標(biāo)函數(shù)之間存在一定的沖突且很難同時(shí)達(dá)到最優(yōu)的問題[4,11],本文對豐水期與枯水期進(jìn)行分期模擬,構(gòu)建分期設(shè)權(quán)理想點(diǎn)法對多目標(biāo)問題進(jìn)行轉(zhuǎn)化,在不同時(shí)期,有所側(cè)重地考慮表征該時(shí)期水文特征的目標(biāo)函數(shù),分期設(shè)置評價(jià)函數(shù)中目標(biāo)函數(shù)權(quán)重,解決不同目標(biāo)函數(shù)之間相互矛盾的問題.
概念性水文模型物理過程清晰,結(jié)構(gòu)簡單,對驅(qū)動數(shù)據(jù)要求低,能完整、科學(xué)地描述水文循環(huán)機(jī)??理[12-13],與輸入資料復(fù)雜、參數(shù)繁多的分布式水文模型相比,概念性水文模型參數(shù)少、適用性強(qiáng),在實(shí)踐中被廣泛應(yīng)用,代表模型有水箱模型、新安江模型、HBV模型、ABCD模型等.其中,ABCD模型因參數(shù)少、資料易收集、模擬精度高、適用性強(qiáng)等特點(diǎn),在國外已被廣泛應(yīng)用于年、月尺度的徑流預(yù)測[14-17].目前,ABCD水文模型在國內(nèi)流域的研究運(yùn)用還相對較少[18-20].基于此,本文以ABCD模型為預(yù)報(bào)模型,提出基于分期設(shè)權(quán)理想點(diǎn)法的水文模型參數(shù)多目標(biāo)優(yōu)化率定方法,將其應(yīng)用于涇河流域的月徑流預(yù)報(bào),分析不同組合方案的模型參數(shù)優(yōu)化效果,尋找模擬效果最佳的組合方案,以期為實(shí)際工程應(yīng)用提供更為可靠的決策依據(jù),為水文模型參數(shù)優(yōu)化率定提供一種高效實(shí)用的方法.
ABCD模型是Thomas[14]于1981年提出的非線性水量平衡模型,該模型只有4個(gè)參數(shù),以降水和潛在蒸發(fā)作為輸入,將流域氣象水文要素之間的關(guān)系概化為經(jīng)驗(yàn)公式,模擬流域?qū)嶋H蒸發(fā)、土壤水及蓄水量、地下水儲量以及地表、地下徑流.與新安江模型等國內(nèi)常用概念性模型相比,該模型具有結(jié)構(gòu)簡單,過程機(jī)理清晰,數(shù)據(jù)易收集,參數(shù)少易優(yōu)化,在月、年尺度上適用性強(qiáng)等優(yōu)點(diǎn)[21].
模型將流域儲水空間概化為土壤水層和地下水層,以水量平衡為基本原理構(gòu)建土壤水層、地下水層水量均衡方程.
圖1?ABCD模型結(jié)構(gòu)
式中a和b為參數(shù),分別代表土壤完全飽和前徑流的傾向性、土壤含水量與蒸發(fā)量的和.Yi和Wi的關(guān)系如圖2[15]所示.
假設(shè)土壤蓄水量與蒸發(fā)量成正比,則有
式中為土壤水補(bǔ)給地下水的比例.
將地下水層概化為線性水庫,地下徑流量可表?示為
式中表示地下水出流速度.
ABCD模型的產(chǎn)流量為
表1?模型參數(shù)(狀態(tài)變量)的物理意義及取值范圍
Tab.1 Meaning and range of the model parameters (ini-tial values of the state variables)
水文模型多目標(biāo)參數(shù)優(yōu)化的目的是通過對模型參數(shù)進(jìn)行全局優(yōu)化搜索,得到多個(gè)目標(biāo)函數(shù)條件下的最優(yōu)非劣解集,以提高水文模型的模擬精度.假設(shè)目標(biāo)函數(shù)均為最小化,水文模型參數(shù)多目標(biāo)優(yōu)化問題可以表示為
水文模型參數(shù)多目標(biāo)優(yōu)化問題無約束條件限制,其搜索的可行域由模型各參數(shù)的取值范圍限定.
直接求解多目標(biāo)優(yōu)化問題具有一定的難度和局限性,在實(shí)際應(yīng)用中,將復(fù)雜的多目標(biāo)問題轉(zhuǎn)化為較易求解的單目標(biāo)問題更為高效.理想點(diǎn)法是由Hwang和Yoor于1981年首次提出,其核心思想是使各目標(biāo)函數(shù)值盡量逼近其理想值,通過調(diào)節(jié)各目標(biāo)的權(quán)重,使得各目標(biāo)函數(shù)之間達(dá)到一種合理的平衡,以滿足不同模擬情景下參數(shù)優(yōu)化需要.具體過程如下.
(2) 得到各個(gè)目標(biāo)函數(shù)的最優(yōu)值構(gòu)成理想點(diǎn)
由于模型參數(shù)多目標(biāo)優(yōu)化率定時(shí),表征高水流量特征的目標(biāo)函數(shù)與表征低水流量的目標(biāo)函數(shù)有時(shí)會存在一定的沖突[11,22],且很難同時(shí)達(dá)到最優(yōu),為解決這一矛盾,在豐水期與枯水期分別有所側(cè)重地考慮表征高水流量與低水流量的目標(biāo)函數(shù),分期設(shè)置評價(jià)函數(shù)式(17)中目標(biāo)函數(shù)的權(quán)重,構(gòu)建分期設(shè)權(quán)理想點(diǎn)法,公式可改寫為
2.6 飲食指導(dǎo) 腹腔積液患者應(yīng)進(jìn)食易消化、富含維生素和蛋白質(zhì)、低脂肪、低鹽的飲食,禁食刺激性食物及飲酒,防止誘發(fā)消化道出血。
采用式(18)將多目標(biāo)函數(shù)在不同時(shí)期分別進(jìn)行單目標(biāo)轉(zhuǎn)化,每個(gè)分期內(nèi)設(shè)置目標(biāo)函數(shù)的權(quán)重,具體見第4.2節(jié)模擬方案設(shè)置.
分期設(shè)權(quán)理想點(diǎn)法將多目標(biāo)問題轉(zhuǎn)化為單目標(biāo)問題,可采用粒子群優(yōu)化算法求解.粒子群優(yōu)化(particle swarm optimization,PSO)算法是基于動物群體的智能優(yōu)化算法[23],該算法全局尋優(yōu)能力強(qiáng),迭代次數(shù)少,對求解水文模型參數(shù)優(yōu)化這種大規(guī)模、多峰值、高離散的非線性優(yōu)化的問題具有很強(qiáng)大的處理能力[24],其基本思路及相關(guān)計(jì)算過程參照文獻(xiàn)[25].
涇河是渭河最大的支流,全長455.1km,控制流域面積45421km2.張家山水文站是涇河干流控制站,控制涇河流域95%的面積,近60年,多年平均年徑流量為15.5×108m3,其中,豐水期徑流量為10.9×108m3,枯水期徑流量為2.1×108m3,豐、枯水期徑流量相差較大.張家山斷面流量的準(zhǔn)確模擬預(yù)報(bào)可為涇河下游水資源開發(fā)利用提供科學(xué)依據(jù),對渭河、黃河以及三門峽庫區(qū)防汛抗旱的管理決策有重要的指導(dǎo)意義.
本文共收集了涇河流域7個(gè)氣象站點(diǎn)1960—2016年的氣象資料及張家山水文站1960—2016年的月徑流資料.氣象數(shù)據(jù)資料來源于中國氣象科學(xué)數(shù)據(jù)共享服務(wù)網(wǎng),張家山站徑流資料來源于黃河流域水文年鑒.流域地理位置及氣象站、水文站點(diǎn)分布情況見圖3.
圖3?涇河流域地理位置及水文站點(diǎn)分布
1990年后,流域修建水利工程、灌區(qū)引水以及退耕還林、還草等措施的實(shí)施,對張家山斷面徑流序列一致性影響較大,20世紀(jì)90年代至今,多年平均年徑流量明顯小于長系列多年平均值.徑流模擬前需要對徑流的一致性進(jìn)行分析.本文采用Mann-Kendall檢驗(yàn)法和降雨徑流雙累積曲線法對徑流序列變異性進(jìn)行分析,結(jié)果表明,徑流序列在1993年左右發(fā)生變異.以變異點(diǎn)為分界點(diǎn)將徑流序列劃分為天然期(1960—1992年)和影響期(1993—2016年)兩部分,為避免非一致性序列對模型精度的影響,模擬采用天然期徑流序列,其中1960—1985年為ABCD模型率定期,1986—1992年為模型檢驗(yàn)期.
方案D為方案A、方案B和方案C的組合方案,即選取方案B和方案C中模擬效果最佳的子方案作為豐水期和枯水期徑流的模擬方案,選取方案A中模擬效果最佳的子方案作為平水期徑流的模擬方案.具體方案及各目標(biāo)函數(shù)對應(yīng)的權(quán)重見表2.
表2?模擬方案及多目標(biāo)函數(shù)權(quán)重設(shè)置
Tab.2 Simulation scheme and weight setting of the multi-objective function
4.3.1?多目標(biāo)優(yōu)化模擬效果分析
方案A1~A4分別以式(11)~(14)作為單目標(biāo)函數(shù),采用PSO算法優(yōu)化求解模型參數(shù),得出單目標(biāo)最優(yōu)目標(biāo)函數(shù)值;方案A5為多目標(biāo)優(yōu)化方案,采用等權(quán)理想點(diǎn)法將其轉(zhuǎn)化為單目標(biāo)函數(shù)后采用PSO算法求解.方案A的目標(biāo)函數(shù)值、率定期與驗(yàn)證期模型評價(jià)指標(biāo)值見表3,方案A5模擬徑流過程與實(shí)測徑流過程對比分析見圖4.
表3?模型模擬效果評價(jià)(方案A)
Tab.3?Evaluation of the model simulation effect(scheme A)
注:*為單目標(biāo)優(yōu)化時(shí)的最優(yōu)目標(biāo)函數(shù)值.
圖4?模擬徑流與實(shí)測徑流對比分析(方案A5)
從圖4可以看出,不論率定期還是檢驗(yàn)期,模擬徑流與實(shí)測徑流在變化趨勢上基本保持一致,時(shí)程同步性較高,模擬徑流和實(shí)測徑流基本呈現(xiàn)對應(yīng)關(guān)系,整體而言模擬結(jié)果較好,說明ABCD模型在涇河張家山站月徑流模擬研究中表現(xiàn)出較好的適用性.但是在豐水期和枯水期模擬偏差相對較大的問題還有待進(jìn)一步解決.
4.3.2?豐水期和枯水期分期優(yōu)化模擬
表4?豐水期徑流模擬效果評價(jià)(方案B)
Tab.4?Evaluation of the runoff simulation effect in the wet period(scheme B)
表5?枯水期徑流模擬效果評價(jià)(方案C)
Tab.5?Evaluation of the runoff simulation effect in the dry period(scheme C)
4.3.3?組合徑流模擬效果分析
為提高徑流整體模擬效果,以方案B3作為豐水期徑流的模擬方案,C3作為枯水徑流的模擬方案,A5作為平水期徑流的模擬方案,提出分期模擬組合方案D.將方案D與全時(shí)期方案A5進(jìn)行對比分析,2種方案模型模擬效果見表6,模擬徑流過程與實(shí)測徑流過程對比分析見圖5,模擬徑流與實(shí)測徑流相關(guān)分析散點(diǎn)圖見圖6.
由圖5可以看出,無論在率定期還是檢驗(yàn)期,方案D模擬的高水流量及低水流量更接近實(shí)測徑流,模擬徑流與實(shí)測徑流在變化趨勢上更加一致,具有更好的時(shí)程同步性;由圖6可以直觀地看出,兩種方案模擬徑流值與實(shí)測徑流值圍繞在1∶1線周圍,點(diǎn)群表現(xiàn)出來的整體的模擬效果較好,但是相較而言,在高流量和低流量部分,方案D的點(diǎn)群更集中于1∶1線周圍.以上的結(jié)果均表明方案D的模擬效果優(yōu)于方案A5,也驗(yàn)證了對豐水期、枯水期進(jìn)行分期模擬能夠有效地提高高水流量與低水流量的模擬效果,同時(shí)進(jìn)一步說明分期模擬組合方案能更有效地提高徑流模擬精度,以期為徑流模擬提供一種新的思路.
表6?方案A5與方案D模擬效果對比分析
Tab.6?Comparative analysis of the simulation effects of schemes A5 and D
圖5?方案A5和方案D模擬徑流與實(shí)測徑流對比分析
圖6?方案A5和方案D模擬徑流與實(shí)測徑流離散分析
(1) 構(gòu)建了基于水量平衡原理的ABCD水文模型,并將其應(yīng)用于涇河流域月徑流模擬,通過采用單目標(biāo)、多目標(biāo)等多種優(yōu)化方案對模型參數(shù)進(jìn)行率定,多目標(biāo)優(yōu)化結(jié)果在率定期、驗(yàn)證期均表現(xiàn)出了較好的模擬精度和效果,研究表明ABCD模型在涇河流域具有較好的適用性,可用于該流域的月徑流預(yù)報(bào).
(2) 基于理想點(diǎn)法空間距離最小的基本原理,提出了分期設(shè)權(quán)理想點(diǎn)法,該方法將多目標(biāo)優(yōu)化問題轉(zhuǎn)化為單目標(biāo)優(yōu)化問題,有效地解決了直接求解多目標(biāo)優(yōu)化計(jì)算復(fù)雜的問題,且通過調(diào)節(jié)目標(biāo)函數(shù)權(quán)重來協(xié)調(diào)目標(biāo)函數(shù)之間的相互矛盾,該方法相對簡單,實(shí)用性強(qiáng).
[1] 劉?寧. 大江大河防洪關(guān)鍵技術(shù)問題與挑戰(zhàn)[J]. 水利學(xué)報(bào),2018,49(1):19-25.
Liu Ning. Problems and challenges on flood control for large rivers[J]. Journal of Hydraulic Engineering,2018,49(1):19-25(in Chinese).
[2] 金君良,舒章康,陳?敏,等. 基于數(shù)值天氣預(yù)報(bào)產(chǎn)品的氣象水文耦合徑流預(yù)報(bào)[J]. 水科學(xué)進(jìn)展,2019,30(3):316-325.
Jin Junliang,Shu Zhangkang,Chen Min,et al. Meteo-hydrological coupled runoff forecasting based on numerical weather prediction products[J]. Advances in Water Science,2019,30(3):316-325(in Chinese).
[3] 袁?飛,趙晶晶,任立良,等. TRMM多衛(wèi)星測雨數(shù)據(jù)在贛江上游徑流模擬中的應(yīng)用[J]. 天津大學(xué)學(xué)報(bào):自然科學(xué)與工程技術(shù)版,2013,46(7):611-616.
Yuan Fei,Zhao Jingjing,Ren Liliang,et al. Stream flow simulation in the upper Ganjiang River Basin using the TRMM multi-satellite precipitation data[J]. Journal of Tianjin University:Science and Technology,2013,46(7):611-616(in Chinese).
[4] 歐陽碩,徐高洪,戴明龍,等. 概念性水文模型參數(shù)多目標(biāo)率定及參數(shù)組合預(yù)報(bào)[J]. 四川大學(xué)學(xué)報(bào):工程科學(xué)版,2014,46(6):63-70.
Ouyang Shuo,Xu Gaohong,Dai Minglong,et al. Study on parameter multi-objective calibration and parameter combination forecast of conceptual hydrological model[J]. Journal of Sichuan University:Engineering Science Edition,2014,46(6):63-70(in Chinese).
[5] Vrugt J,Gupta H V,Bastidas L A,et al. Effective and efficient algorithm for multi-objective optimization of hydrologic models[J]. Water Resources Research,2003,39(8):12-14.
[6] 周建中,盧韋偉,孫?娜,等. 水文模型參數(shù)多目標(biāo)率定及最優(yōu)非劣解優(yōu)選[J]. 水文,2017,37(2):1-7.
Zhou Jianzhong,Lu Weiwei,Sun Na,et al. Study on multi-objective calibration of hydrological model and optimization method of optimal pareto solutions[J]. Journal of China Hydrology,2017,37(2):1-7(in Chinese).
[7] Yapo P O,Gupta H V,Sorooshian S. Multi-objective global optimization for hydrologic models[J]. Journal of Hydrology,1998,204(1):83-97.
[8] 張洪剛,郭生練,王才君,等. 概念性流域水文模型參數(shù)優(yōu)選技術(shù)研究[J]. 武漢大學(xué)學(xué)報(bào):工學(xué)版,2004,37(3):18-22,26.
Zhang Honggang,Guo Shenglian,Wang Caijun,et al. Review and research of calibration techniques for conceptual hydrological model[J]. Engineering Journal of Wuhan University,2004,37(3):18-22,26(in Chinese).
[9] 宋曉猛,張建云,孔凡哲,等. 基于代理模型的水文模型參數(shù)多目標(biāo)優(yōu)化[J]. 四川大學(xué)學(xué)報(bào):工程科學(xué)版,2014,46(2):36-45.
Song Xiaomeng,Zhang Jianyun,Kong Fanzhe,et al. Multi-objective optimization for hydrological models using surrogate modeling[J]. Journal of Sichuan University:Engineering Science Edition,2014,46(2):36-45(in Chinese).
[10] Hwang C L,Yoon K L. Multiple Attribute Decision Making Methods and Applications[M]. New York:Spring-Verlag,1981.
[11] 郭?俊,周建中,周?超,等. 概念性流域水文模型參數(shù)多目標(biāo)優(yōu)化率定[J]. 水科學(xué)進(jìn)展,2012,23(4):447-456.
Guo Jun,Zhou Jianzhong,Zhou Chao,et al. Multi-objective optimization for conceptual hydrological models[J]. Advances in Water Science,2012,23(4):447-456(in Chinese).
[12] 徐宗學(xué). 水文模型:回顧與展望[J]. 北京師范大學(xué)學(xué)報(bào):自然科學(xué)版,2010,46(3):278-289.
Xu Zongxue. Hydrological models:Past,present and future[J]. Journal of Beijing Normal University:Natural Science,2010,46(3):278-289(in Chinese).
[13] 董艷萍,袁晶瑄. 流域水文模型的回顧與展望[J]. 水力發(fā)電,2008(3):20-23.
Dong Yanping,Yuan Jingxuan. The retrospect and looking into the future of watershed hydrological model[J]. Water Power,2008(3):20-23(in Chinese).
[14] Thomas H A. Improved Methods for National Water Assessment Water Resources Contract:WR15249270 [R]. Cambridge:Harvard University,1981.
[15] Alley W M. On the treatment of evapotranspiration,soil moisture accounting,and aquifer recharge in monthly water balance models[J]. Water Resources Research,1984,20(8):1137-1149.
[16] Alley W M. Water balance models in one-month-ahead stream flow forecasting[J]. Water Resources Research,1985,21(4):597-606.
[17] Martinez G F,Gupta H V. Toward improved identification of hydrological models:A diagnostic evaluation of the “abcd”monthly water balance model for the conterminous United States[J]. Water Resources Research,2010,46(8):W08507.1-W08507.21.
[18] 李?帥,熊立華,萬?民. 月水量平衡模型的比較研究[J]. 水文,2011,31(5):35-41.
Li Shuai,Xiong Lihua,Wan Min. Comparison of monthly water balance models[J]. Journal of China Hydrology,2011,31(5):35-41(in Chinese).
[19] 王?喆,李?昱,丁?偉,等. 基于擴(kuò)展Budyko假設(shè)的季節(jié)徑流變化歸因分析[J]. 水力發(fā)電學(xué)報(bào),2017,36(8):22-33.
Wang Zhe,Li Yu,Ding Wei,et al. Attribution analysis on seasonal river runoff based on extended Budyko hypothesis[J]. Journal of Hydroelectric Engineering,2017,36(8):22-33(in Chinese).
[20] 吳光東,許繼軍,Hoshin Gupta,等. 新安江流域abcd水量平衡模型及參數(shù)敏感性分析[J]. 長江科學(xué)院院報(bào),2019,36(7):23-27,40.
Wu Guangdong,Xu Jijun,Hoshin Gupta,et al. The “abcd” water balance model:Application to Xin’an River Basin and sensitivity analysis[J]. Journal of Yangtze River Scientific Research Institute,2019,36(7):23-27,40(in Chinese).
[21] 韓鵬飛,王旭升. 利用ABCD模型預(yù)測流域水文對極端氣候的響應(yīng)[J]. 人民黃河,2016,38(11):16-22.
Han Pengfei,Wang Xusheng. Forecasting the response of a catchment on extreme climate change with ABCD model[J]. Yellow River,2016,38(11):16-22(in Chinese).
[22] 郭?俊,周建中,鄒?強(qiáng),等. 水文模型參數(shù)多目標(biāo)優(yōu)化率定及目標(biāo)函數(shù)組合對優(yōu)化結(jié)果的影響[J]. 四川大學(xué)學(xué)報(bào):工程科學(xué)版,2011,43(6):58-63.
Guo Jun,Zhou Jianzhong,Zou Qiang,et al. Study on multi-objective calibration of hydrological model and effect of objective functions combination on optimization results[J]. Journal of Sichuan University:Engineering Science Edition,2011,43(6):58-63(in Chinese).
[23] 劉洪波,鄭博一,蔣博齡. 基于人工魚群神經(jīng)網(wǎng)絡(luò)的城市時(shí)用水量預(yù)測方法[J]. 天津大學(xué)學(xué)報(bào):自然科學(xué)與工程技術(shù)版,2015,48(4):373-378.
Liu Hongbo,Zheng Boyi,Jiang Boling. Forecast method of city water consumption per hour based on artificial fish-swarm neural network[J]. Journal of Tianjin University:Science and Technology,2015,48(4):373-378(in Chinese).
[24] 張文明,董增川,朱成濤,等. 基于粒子群算法的水文模型參數(shù)多目標(biāo)優(yōu)化研究[J]. 水利學(xué)報(bào),2008,39(5):528-534.
Zhang Wenming,Dong Zengchuan,Zhu Chengtao,et al. Automatic calibration of hydrologic model based on multi-objective particle swarm optimization method[J]. Journal of Hydraulic Engineering,2008,39(5):528-534(in Chinese).
[25] 路志英,李艷英,陸?潔,等. 粒子群算法優(yōu)化RBF-SVM沙塵暴預(yù)報(bào)模型參數(shù)[J]. 天津大學(xué)學(xué)報(bào),2008,41(4):413-418.
Lu Zhiying,Li Yanying,Lu Jie,et al. Parameters optimization of RBF-SVM sand-dust storm forecasting model based on PSO[J]. Journal of Tianjin University,2008,41(4):413-418(in Chinese).
Multi-Objective Optimization of Hydrological Model Parameters Based on the Stage-Weighted Ideal Point Method
Kang Yan1, 2,Yi Li1, 2,Gong Jiaguo3
(1. School of Water Resource and Architectural Engineering,Northwest Agriculture and Forest University,Yangling 722100,China;2. Key Laboratory of Agricultural Soil and Water Engineering in Arid and Semiarid Areas,Ministry of Education,Northwest Agriculture and Forest University,Yangling 722100,China;3. China Institute of Water Resources and Hydropower Research,Beijing 100038,China)
Because of the complexity of calculation and the contradiction between multiple objectives in the multi-objective optimization of hydrological model parameters,a multi-objective optimization method based on the stage-weighted ideal point method was proposed. Taking the ABCD hydrological model as a monthly runoff prediction model,four kinds of objective functions,namely,overall water balance error,average relative error,Nash-Sutcliffe coefficient of high flow,and Nash-Sutcliffe coefficient of low flow,were constructed. The multi-objective functions were converted into a single objective by the stage-weighted ideal point method and solved by particle swarm optimization. The models were applied to the simulation of monthly runoff in the Jinghe River Basin. Four simulation schemes were set up,namely,full period(A),wet period(B),dry period(C),and staged simulation combination(D). The simulation results under different schemes were analyzed. The results show that in the full period(A)simulation scheme,the multi-objective optimization scheme(A5)can effectively represent multiple hydrological process characteristics and coordinate the mutual exclusion relationship between objective functions. Moreover,the simulation effects of multi-objective optimization are better than those of single-objective optimization. In the staged simulation combination scheme,the sub-scheme can not only reflect the hydrological characteristics of the objective function reasonably,but also take into account other objective functions when the weights of the rainy-period and low-water objectives are 0.75;thus,this scheme is better than the other schemes. The staged simulation combination scheme had a better simulation effect than the multi-objective optimization scheme in the high-flow and low-flow periods. The correlation and efficiency coefficients of the evaluation indices are all greater than 0.8,and the average absolute percentage error is less than 1%. This finding shows that the scheme of the staged simulation combination scheme can effectively improve the simulation accuracy of the model.
multi-objective optimization of hydrological model parameters;stage-weighted ideal point method;ABCD hydrological model;runoff simulation;Jinghe River Basin
TV11
A
0493-2137(2021)05-0458-10
10.11784/tdxbz202004033
2020-04-14;
2020-06-05.
康?艷(1977—??),女,博士,副教授.
康?艷,kangyan@nwsuaf.edu.cn.
陜西省水利科技計(jì)劃資助項(xiàng)目(2019slkj-14);國家自然科學(xué)基金資助項(xiàng)目(51409222);國家重點(diǎn)研發(fā)計(jì)劃資助項(xiàng)目(2016YFC0401306).
Supported by the Water Conservancy Science-Technology Plan Program of Shaanxi Province(No. 2019slkj-14),the National Natural Science Foundation of China(No. 51409222),the National Key Research and Development Program of China(No. 2016YFC0401306).
(責(zé)任編輯:金順愛)