管光華,賈夢(mèng)浩
基于實(shí)測(cè)水位和流量數(shù)據(jù)的渠道控制模型參數(shù)辨識(shí)與驗(yàn)證
管光華,賈夢(mèng)浩
(武漢大學(xué)水資源與水電工程科學(xué)國(guó)家重點(diǎn)實(shí)驗(yàn)室,武漢 430072)
針對(duì)傳統(tǒng)渠道控制模型推導(dǎo)求解復(fù)雜,難以在實(shí)際工程中應(yīng)用的問題,該研究以渠道積分延遲(Integrator Delay,ID)模型和積分延遲零(Integrator Delay Zero,IDZ)模型為研究對(duì)象,利用遞推最小二乘法對(duì)2種渠道控制模型進(jìn)行了參數(shù)辨識(shí)和分析。結(jié)果表明,參數(shù)辨識(shí)方法可以簡(jiǎn)單、有效地獲取渠道系統(tǒng)的控制模型;在模型試驗(yàn)中,辨識(shí)誤差基本分布在(?1,1)cm之間,在南水北調(diào)原型工程中,辨識(shí)誤差基本分布在(?2,2)cm之間;2種辨識(shí)模型的均方誤差均小于7.675×10-5m2,且針對(duì)水流變化復(fù)雜的大型渠道工程,IDZ模型的辨識(shí)精度和穩(wěn)定性整體高于ID模型。該研究結(jié)論對(duì)于渠道控制建模理論及控制器設(shè)計(jì)方法具有一定的參考價(jià)值,可應(yīng)用于灌區(qū)或調(diào)水工程的輸配水渠道系統(tǒng)建模。
模型;渠道;水位;渠道控制模型;積分延遲(ID)模型;積分延遲零(IDZ)模型;參數(shù)辨識(shí);遞推最小二乘法(RLS)
中國(guó)作為人口大國(guó)與農(nóng)業(yè)大國(guó),水資源短缺和時(shí)空分布不均成為制約社會(huì)發(fā)展的重要因素,而水利自動(dòng)化的發(fā)展能夠有效提高水資源輸配效率并減少灌溉渠道輸水損失,從而緩解由于水資源短缺引起的社會(huì)矛盾。水利自動(dòng)化的核心是控制器的設(shè)計(jì),而控制器設(shè)計(jì)的基礎(chǔ)則是渠系控制模型的建立,因此如何快速、準(zhǔn)確地建立渠道控制模型成為輸配水工程運(yùn)行管理的難點(diǎn)。渠道控制模型是描述渠系建筑物-閘門-水流動(dòng)態(tài)關(guān)系的數(shù)學(xué)表達(dá)式,傳統(tǒng)的渠道控制模型根據(jù)圣維南方程組推導(dǎo)建立,該方程組是雙曲型偏微分方程,求解復(fù)雜,不利于控制器的設(shè)計(jì)。因此對(duì)圣維南方程組進(jìn)行線性化處理,并采用參數(shù)辨識(shí)方法對(duì)模型參數(shù)進(jìn)行辨識(shí)成為解決實(shí)際工程中建模難的有效方法[1]。
在渠道自動(dòng)化控制領(lǐng)域,有很多學(xué)者基于圣維南方程組提出了不同的控制模型,Malaterre等[2]和Reddy等[3]基于普萊士曼格式和泰勒展開式對(duì)圣維南方程組進(jìn)行線性化處理,推導(dǎo)出了渠道狀態(tài)空間模型,但該模型變量較多,難以應(yīng)用;Corriga等[4]和Ermolin[5]通過對(duì)圣維南方程組進(jìn)行線性變化和拉氏變換,建立了簡(jiǎn)化的渠道控制模型,但該模型僅適用于均勻流;Schuurmans等[6]通過考慮渠道中的均勻流段和回水段,提出了積分延遲(Integrator Delay,ID)模型,該模型較為完整地反映了渠道中的水流特性,但其只適用于低頻水流;Litrico等[7-8]和Clemmens等[9-10]在ID模型的基礎(chǔ)上提出了積分延遲零(Integrator Delay Zero,IDZ)模型,該模型同時(shí)考慮了渠道內(nèi)水流的低頻和高頻特性,能夠較好地捕捉到渠池內(nèi)水位對(duì)流量變化的響應(yīng)情況。
國(guó)內(nèi)外學(xué)者針對(duì)參數(shù)辨識(shí)也展開了一系列研究,Litrico[11]通過假設(shè)灌溉系數(shù)恒定,對(duì)明渠進(jìn)行辨識(shí)建模,初步驗(yàn)證了辨識(shí)方法的有效性;楊開林等[12-13]通過對(duì)南水北調(diào)部分渠段的糙率和閘門特性進(jìn)行參數(shù)辨識(shí)研究,進(jìn)一步驗(yàn)證了參數(shù)辨識(shí)方法的可靠性;Foo等[14-15]通過對(duì)澳大利亞布羅肯河進(jìn)行參數(shù)辨識(shí),驗(yàn)證了ID模型對(duì)于控制器的設(shè)計(jì)是有效的;Weyer等[16-19]通過對(duì)澳大利亞墨累河進(jìn)行試驗(yàn)研究,提出了完整的參數(shù)辨識(shí)流程,并驗(yàn)證了不同辨識(shí)方法下控制模型的精度,但該研究只針對(duì)低頻下的水流數(shù)據(jù)進(jìn)行;Liao等[20]通過對(duì)漳河灌區(qū)進(jìn)行仿真研究,驗(yàn)證了IDZ模型對(duì)渠道內(nèi)高頻水流的適用性。
目前,關(guān)于渠道控制模型及參數(shù)辨識(shí)方法的研究多是基于天然河流或仿真數(shù)據(jù),對(duì)于實(shí)際渠道運(yùn)行調(diào)度過程的研究較為匱乏。因此,針對(duì)中國(guó)當(dāng)前灌區(qū)和調(diào)水工程中渠道系統(tǒng)運(yùn)行管理難的問題,本研究基于模型試驗(yàn)數(shù)據(jù)和原型工程觀測(cè)數(shù)據(jù),利用參數(shù)辨識(shí)方法對(duì)渠道控制模型進(jìn)行參數(shù)辨識(shí),從而獲取更適用于實(shí)際工程的簡(jiǎn)單模型。
1.1.1 試驗(yàn)場(chǎng)地
本研究進(jìn)行模型試驗(yàn)的渠道系統(tǒng)位于武漢大學(xué)水資源與水電工程科學(xué)國(guó)家重點(diǎn)實(shí)驗(yàn)室,主要利用的是灌排渠系水量實(shí)時(shí)測(cè)控系統(tǒng)。該測(cè)控系統(tǒng)由眾多模塊組成,可以實(shí)現(xiàn)對(duì)輸水渠道系統(tǒng)的實(shí)時(shí)信息采集、閘門自動(dòng)調(diào)節(jié)和中央集中監(jiān)控等多種功能。試驗(yàn)系統(tǒng)主要包括5個(gè)部分:主測(cè)控系統(tǒng)、取水泵、節(jié)制閘、試驗(yàn)渠系和地下水庫(kù)。試驗(yàn)渠道全長(zhǎng)(,m)約134 m,沿程設(shè)有20個(gè)水位計(jì)(相鄰間隔約6 m),2個(gè)流量計(jì),2個(gè)節(jié)制閘,2個(gè)水力自動(dòng)閘門。試驗(yàn)渠道的上下游渠底高程分別為22.65和22.64 m,渠底縱向坡度為0.000 1,糙率為0.015,設(shè)計(jì)流量為0.12 m3/s,其平面布置及典型斷面如圖1所示。
本研究選取的原型工程為南水北調(diào)中線京石段第五渠池。與模型試驗(yàn)渠道相比,二者的長(zhǎng)度之比約為60:1,寬度之比約為15:1,流量之比約為1 100:1。
注:尺寸單位為cm;渠道面層填充物為混凝土砂漿;A-A為渠道上游橫剖面;B-B為渠道下游橫剖面;R表示彎道半徑;1:1.0和1:0.5表示渠道坡度(垂直:水平)。
1.1.2 試驗(yàn)工況設(shè)計(jì)
為了充分觀察渠池內(nèi)水位對(duì)流量變化的響應(yīng)情況并獲得模型的輸入、輸出數(shù)據(jù),模型試驗(yàn)設(shè)計(jì)了2種典型的流量工況,分別為正向階躍工況和反向階躍工況,如圖2所示。
圖2 試驗(yàn)流量工況
1.1.3 試驗(yàn)方法
首先,通過安裝在渠首的取水泵進(jìn)行水流的抽取,同時(shí)主測(cè)控系統(tǒng)可以調(diào)節(jié)取水泵的轉(zhuǎn)速和節(jié)制閘的開度,以此實(shí)現(xiàn)不同流量大小的變化。其次,當(dāng)水流進(jìn)入試驗(yàn)渠道時(shí),安裝在渠道內(nèi)的水位計(jì)和流量計(jì)會(huì)將實(shí)時(shí)監(jiān)測(cè)數(shù)據(jù)反饋至主測(cè)控系統(tǒng)進(jìn)行存儲(chǔ)。最后,渠道內(nèi)的尾水會(huì)流入地下水庫(kù)循環(huán)使用,整個(gè)試驗(yàn)的流程如圖3所示。
1.1.4 實(shí)測(cè)數(shù)據(jù)
試驗(yàn)獲取的數(shù)據(jù)信息主要包括上下游的水位和流量。水位、流量信息分別通過水位計(jì)、流量計(jì)進(jìn)行實(shí)時(shí)監(jiān)測(cè)獲取;其中,水位計(jì)的最小精度為0.01 m,流量計(jì)的最小精度為0.001 m3/s,數(shù)據(jù)的采樣周期為20 s。
注:+代表主控系統(tǒng)起正調(diào)節(jié)作用,加大入流量;?代表主控系統(tǒng)起負(fù)調(diào)節(jié)作用,減小入流量;U為比較點(diǎn)。
在渠道控制系統(tǒng)中,基于圣維南方程組建立的控制模型能較好地描述水流的運(yùn)動(dòng)規(guī)律,但是圣維南方程組是雙曲型偏微分方程,其推導(dǎo)、求解過程復(fù)雜,難以用于控制器的設(shè)計(jì)。因此,Schuurmans等[6]和Litrico等[7-8]對(duì)圣維南方程組進(jìn)行了線性、拉普拉斯變換,從而推導(dǎo)出了適用于明渠渠系控制的簡(jiǎn)單模型——積分延遲(ID)模型和積分延遲零(IDZ)模型。
1.2.1 圣維南方程組
圣維南方程組[6-8]是描述明渠水流運(yùn)動(dòng)的經(jīng)典方程,其時(shí)域表達(dá)式如下:
式中為流量,m3/s;為時(shí)間,s;為渠池橫斷面面積,m2;為渠池水深,m;為水面寬度,m;為水流沿程距離,m;為糙率;0為水力半徑,m;為渠底縱向坡度;為重力加速度,m/s2。
1.2.2 積分延遲(ID)模型
1.2.3 積分延遲零(IDZ)模型
渠道積分延遲零(IDZ)模型是Litrico等[7-8]于2004年提出的,IDZ模型是在經(jīng)典的積分延遲(ID)模型基礎(chǔ)上擴(kuò)展出的。相比于ID模型,IDZ模型能更好地描述渠道內(nèi)水流在低頻和高頻下的運(yùn)動(dòng)特性。IDZ模型中,積分延遲用來描述低頻,而零項(xiàng)則反映高頻下水位對(duì)流量的響應(yīng)情況。IDZ模型主要有三點(diǎn)進(jìn)步:1)為經(jīng)典的ID模型提供了準(zhǔn)確的時(shí)間延遲和積分增益的近似值;2)在高頻中增加了積分延遲零點(diǎn)和積分零點(diǎn)2項(xiàng),從而彌補(bǔ)了ID模型在高頻下的偏差;3)模型中的所有參數(shù)均可進(jìn)行解析計(jì)算。根據(jù)Litrico等[7]的推導(dǎo)結(jié)果,IDZ模型的時(shí)域表達(dá)式如式(4)和式(5)所示:
(5)
水利自動(dòng)化控制系統(tǒng)的設(shè)計(jì)、操作、實(shí)現(xiàn)等都需要有精確的渠系控制模型,用來描述渠系內(nèi)水流的運(yùn)動(dòng)過程、變化規(guī)律等。常用的渠系控制模型,其參數(shù)受外界條件及水力因素等影響,難以求解,因此常常采用試驗(yàn)建模的方式來觀測(cè)系統(tǒng)的輸入、輸出信息,從而確定系統(tǒng)的模型,這個(gè)過程稱作參數(shù)辨識(shí)[21]。相比于傳統(tǒng)的仿真建模,本研究利用1.1節(jié)介紹的小型渠道進(jìn)行試驗(yàn)獲取實(shí)測(cè)數(shù)據(jù),進(jìn)而采用最小二乘法對(duì)2種渠系控制模型(ID、IDZ)進(jìn)行參數(shù)辨識(shí),最終獲得滿足控制器設(shè)計(jì)要求的渠系控制模型。
常用的最小二乘法包括一般最小二乘法和遞推最小二乘法,由于一般最小二乘法的計(jì)算數(shù)據(jù)量會(huì)隨著時(shí)間的延長(zhǎng)不斷增加,因此難以用于實(shí)際渠道模型參數(shù)的在線辨識(shí)。為了減少計(jì)算量,實(shí)現(xiàn)參數(shù)的在線辨識(shí),本研究采用遞推最小二乘法(Recursive Least Square,RLS)進(jìn)行計(jì)算求解。遞推最小二乘法的具體推導(dǎo)及求解過程可參考文獻(xiàn)[22],其基本思想是利用新觀測(cè)的數(shù)據(jù),在上一次辨識(shí)結(jié)果的基礎(chǔ)上,根據(jù)遞推算法進(jìn)行修正,從而得到新的參數(shù)估計(jì)值。
在本研究中,為了實(shí)現(xiàn)參數(shù)辨識(shí),需要對(duì)ID、IDZ的時(shí)域模型進(jìn)行轉(zhuǎn)化,將其變?yōu)殡x散時(shí)間的最小二乘形式,離散后的模型表達(dá)式如式(6)和式(8)所示。
1)ID模型
用矩陣表示如式(7)所示:
2)IDZ模型
用矩陣表示如式(9)所示:
為了實(shí)現(xiàn)2種渠系控制模型(ID、IDZ)的參數(shù)辨識(shí),需要掌握邊界條件、輸入數(shù)據(jù)和滯后時(shí)間的相關(guān)信息。
2.1.1 邊界條件
根據(jù)試驗(yàn)工況設(shè)計(jì),在小型渠道試驗(yàn)過程中,為了獲得滿足研究目的的計(jì)算數(shù)據(jù),試驗(yàn)渠道上下游均采用流量邊界進(jìn)行控制[23],即將渠首的入流量in(0,)和渠尾的出流量out()作為邊界條件。
2.1.2 輸入數(shù)據(jù)
本研究設(shè)計(jì)了2種典型的試驗(yàn)工況,分別為正向階躍工況和反向階躍工況,試驗(yàn)時(shí)間均為40 min。輸入數(shù)據(jù)為渠池的入流量in()和出流量out(),輸出數(shù)據(jù)為渠池的下游水位()。由于小型試驗(yàn)渠道內(nèi)水位計(jì)、流量計(jì)的最小采樣周期為20 s,故在整個(gè)試驗(yàn)時(shí)間內(nèi),可獲得120組關(guān)于in()out()和()的實(shí)測(cè)數(shù)據(jù)。
2.1.3 滯后時(shí)間
由式(7)和式(9)可知,獲得滯后時(shí)間是進(jìn)行模型參數(shù)辨識(shí)的前提。根據(jù)文獻(xiàn)[24-25]中的推導(dǎo),可得滯后時(shí)間的計(jì)算表達(dá)式如式(10)和式(11)所示:
式中為渠池長(zhǎng)度(),m;0為波速,m/s;0為平均流速,m/s;0為渠池橫斷面面積(穩(wěn)定狀態(tài)下),m2;0為水面寬度(穩(wěn)定狀態(tài)下),m。
將試驗(yàn)渠道的幾何參數(shù)代入式(10)和式(11)進(jìn)行計(jì)算,可得正向階躍工況下的滯后時(shí)間約為114 s,反向階躍工況下的滯后時(shí)間約為124 s。由于輸入數(shù)據(jù)的采樣周期為20 s,且在離散模型中,滯后時(shí)間必須為采樣周期的整數(shù)倍,故2種工況下的滯后時(shí)間最終選為120 s,即6個(gè)采樣周期。
根據(jù)遞推最小二乘法的推導(dǎo)過程[22],利用MATLAB對(duì)其進(jìn)行編程求解。將模型試驗(yàn)獲得的實(shí)測(cè)數(shù)據(jù)in()out()和()輸入,對(duì)ID、IDZ模型中的待定參數(shù)進(jìn)行辨識(shí),辨識(shí)結(jié)果如表1所示,辨識(shí)過程如圖4所示。由該結(jié)果可知,隨著遞推的進(jìn)行,2種控制模型的待定參數(shù)逐漸收斂,直至達(dá)到穩(wěn)定。
表1 試驗(yàn)渠道控制模型參數(shù)辨識(shí)結(jié)果
注:1、1為積分延遲模型中的系數(shù),(s·m-2);1、1、2、3為積分延遲零模型中的系數(shù),(s·m-2)。
Note:1and1are coefficients in the Integrator Delay (ID) model, (s·m-2);1,1,2and3are coefficients in the Integrator Delay Zero (IDZ) model, (s·m-2).
根據(jù)表2中計(jì)算出的模型參數(shù),可對(duì)ID、IDZ模型的精度進(jìn)行驗(yàn)證?;舅悸窞?)將參數(shù)辨識(shí)結(jié)果代入渠道控制模型;2)利用辨識(shí)模型進(jìn)行計(jì)算預(yù)測(cè);3)將預(yù)測(cè)結(jié)果與試驗(yàn)觀測(cè)結(jié)果進(jìn)行對(duì)比;4)采用相關(guān)性能指標(biāo)對(duì)模型精度進(jìn)行判斷。利用表1的辨識(shí)結(jié)果及模型試驗(yàn)數(shù)據(jù)進(jìn)行計(jì)算,可繪制辨識(shí)結(jié)果與實(shí)測(cè)結(jié)果的下游水位變化曲線圖及誤差分布圖,如圖5所示。
圖4 待定參數(shù)辨識(shí)過程
選用平均絕對(duì)誤差(Mean Absolute Error,MAE)和均方誤差(Mean Square Error,MSE)2個(gè)性能指標(biāo)[26]對(duì)小型試驗(yàn)渠道控制模型參數(shù)的辨識(shí)結(jié)果進(jìn)行計(jì)算分析,結(jié)果如表2所示。
表2 試驗(yàn)渠道參數(shù)辨識(shí)性能
由圖5和表2可知,辨識(shí)誤差基本分布在(?1, 1)cm之間,辨識(shí)結(jié)果與實(shí)測(cè)值吻合度較高,表明利用遞推最小二乘法(RLS)對(duì)渠系控制模型(ID、IDZ)進(jìn)行參數(shù)辨識(shí)是可行的。且由表2可知,針對(duì)試驗(yàn)渠道,ID模型與IDZ模型的辨識(shí)結(jié)果相近。其原因在于試驗(yàn)渠道的流量較小,水流變化較為單一,渠道內(nèi)的水流為低頻流動(dòng),無明顯的高頻變化。
圖5 各工況下辨識(shí)水位及誤差
為了進(jìn)一步驗(yàn)證渠系控制模型參數(shù)辨識(shí)方法在復(fù)雜實(shí)際工程中應(yīng)用的可行性,本研究選取了南水北調(diào)中線京石段第五渠池的實(shí)測(cè)水位、流量數(shù)據(jù)進(jìn)行ID、IDZ模型的參數(shù)辨識(shí),該渠池的幾何參數(shù)如表3所示。本研究選取了該渠池在2018年4月3日—9日(共144 h)的水位、流量數(shù)據(jù)進(jìn)行分析計(jì)算,數(shù)據(jù)采樣周期為2 h。在此時(shí)間段內(nèi),水位的變化分為2個(gè)明顯過程,一是水位下降段,二是水位上升段。
表3 原型工程幾何參數(shù)
根據(jù)式(10),將原型工程渠道的幾何參數(shù)代入進(jìn)行計(jì)算,可得滯后時(shí)間約為3 210 s。由于輸入數(shù)據(jù)的采樣周期為2 h,且在離散模型中,滯后時(shí)間必須為采樣周期的整數(shù)倍,故滯后時(shí)間最終選為2 h,即1個(gè)采樣周期。
對(duì)原型觀測(cè)數(shù)據(jù)進(jìn)行參數(shù)辨識(shí),計(jì)算結(jié)果如表4所示。
表4 原型工程參數(shù)辨識(shí)結(jié)果
根據(jù)表4的辨識(shí)結(jié)果,將辨識(shí)參數(shù)分別代入ID、IDZ模型的表達(dá)式(式(6)和式(8))中,利用原型觀測(cè)數(shù)據(jù)進(jìn)行計(jì)算,繪制辨識(shí)結(jié)果與實(shí)測(cè)結(jié)果的下游水位變化曲線圖以及誤差分布圖,如圖6所示。
圖6 原型工程辨識(shí)水位及誤差
同樣選用平均絕對(duì)誤差(MAE)和均方誤差(MSE)2個(gè)性能指標(biāo)對(duì)原型工程渠道控制模型參數(shù)的辨識(shí)結(jié)果進(jìn)行計(jì)算分析,結(jié)果如表5所示。
表5 原型工程參數(shù)辨識(shí)性能
由圖6和表5可知,在南水北調(diào)原型工程的參數(shù)辨識(shí)中,辨識(shí)誤差基本分布在(?2, 2)cm之間,ID模型與IDZ模型的辨識(shí)水位與實(shí)測(cè)水位吻合度依然較高。且由表5可知,IDZ模型的MAE最大值為4.740×10-3m,MSE最大值為3.604×10-5m2;ID模型的MAE最大值為6.433×10-3m,MSE最大值為7.675×10-5m2。因此,在水流變化復(fù)雜的實(shí)際工程中,IDZ模型的辨識(shí)效果優(yōu)于ID模型,其響應(yīng)精度更高,能更好地反映水位的變化趨勢(shì),這也證明了其對(duì)于高頻水流的適用性[27]。所以,IDZ模型更適用于大型輸配水工程的渠系控制建模。
針對(duì)中國(guó)當(dāng)前灌區(qū)和調(diào)水工程中渠道系統(tǒng)運(yùn)行管理難的問題,本研究根據(jù)參數(shù)辨識(shí)原理,基于模型試驗(yàn)數(shù)據(jù)和原型觀測(cè)數(shù)據(jù),采用遞推最小二乘法(Recursive Least Square,RLS)對(duì)積分延遲(Integrator Delay,ID)模型和積分延遲零(Integrator Delay Zero,IDZ)模型進(jìn)行了參數(shù)辨識(shí)研究。根據(jù)辨識(shí)結(jié)果及驗(yàn)證分析,得出的主要結(jié)論如下:
1)針對(duì)模型試驗(yàn)數(shù)據(jù)和原型觀測(cè)數(shù)據(jù),利用遞推最小二乘法(RLS)對(duì)渠道控制模型(ID、IDZ)進(jìn)行參數(shù)辨識(shí)是可行且有效的。
2)模型試驗(yàn)數(shù)據(jù)分析表明,ID模型與IDZ模型的辨識(shí)結(jié)果相近,二者的水位辨識(shí)誤差基本分布在(?1, 1)cm之間。
3)原型觀測(cè)數(shù)據(jù)分析表明,在南水北調(diào)實(shí)際工程中,IDZ模型的平均絕對(duì)誤差最大值為4.740×10-3m,均方誤差最大值為3.604×10-5m2;ID模型的平均絕對(duì)誤差最大值為6.433×10-3m,均方誤差最大值為7.675×10-5m2。IDZ模型的辨識(shí)精度和穩(wěn)定性高于ID模型。
[1]楊開林. 長(zhǎng)距離輸水水力控制的研究進(jìn)展與前沿科學(xué)問題[J]. 水利學(xué)報(bào),2016,47(3):424-435. Yang Kailin. Review and frontier scientific issues of hydraulic control for long distance water diversion[J]. Journal of Hydraulic Engineering, 2016, 47(3):424-435. (in Chinese with English abstract)
[2]Malaterre P O, Rodellar J. Multivariable predictive control of irrigation canals: Design and evaluation on a 2-pool model[J]. Reliable Computing, 1997, 3(3): 239-248.
[3]Reddy J M, Jacquot R G. Stochastic optimal and suboptimal control of irrigation canals[J]. Journal of Water Resources Planning and Management, 1999, 125(6): 369-378.
[4]Corriga G, Sanna S, Usai G, et al. Estimation of uncertainty in an open-channel network mathematical model[J]. Applied Mathematical Modelling, 1989, 13(11): 651-657.
[5]Ermolin Y A. Study of open-channel dynamics as controlled process[J]. Journal of Hydraulic Engineering, 1992, 118(1): 59-72.
[6]Schuurmans J, Bosgra O H, Brouwer R. Open-channel flow model approximation for controller design[J]. Applied Mathematical Modelling, 1995, 19(9): 525-530.
[7]Litrico X, Fromion V. Simplified modeling of irrigation canals for controller design[J]. Journal of Irrigation and Drainage Engineering, 2004, 130(5): 373-383.
[8]Litrico X, Fromion V. Analytical approximation of open-channel flow for controller design[J]. Applied Mathematical Modelling, 2004, 28(7): 677-695.
[9]Clemmens A J, Wahlin B T. Simple optimal downstream feedback canal controllers: ASCE test case results[J]. Journal of Irrigation and Drainage Engineering, 2004, 130(1): 35-46.
[10]Clemmens A J, Schuurmans J. Simple optimal downstream feedback canal controllers: Theory[J]. Journal of Irrigation and Drainage Engineering, 2004, 130(1): 26-34.
[11]Litrico X. Nonlinear diffusive wave modeling and identification of open channels[J]. Journal of Hydraulic Engineering, 2001, 127(4): 313-320.
[12]楊開林,汪易森. 南水北調(diào)中線工程渠道糙率的系統(tǒng)辨識(shí)[J]. 中國(guó)工程科學(xué),2012,14(11):17-23. Yang Kailin, Wang Yisen. System identification of channel roughness for middle route project of south-to-north water diversion[J]. Strategic Study of CAE, 2012, 14(11): 17-23. (in Chinese with English abstract)
[13]楊開林,汪易森. 調(diào)水工程閘門特性的動(dòng)態(tài)系統(tǒng)辨識(shí)[J]. 水利學(xué)報(bào),2011,42(11):1289-1294. Yang Kailin, Wang Yisen. Dynamic system identification of gate characteristics for water diversion projects[J]. Journal of Hydraulic Engineering, 2011, 42(11): 1289-1294. (in Chinese with English abstract)
[14]Foo M, Ooi S K, Weyer E. Modelling of rivers for control design[C]// System Identification, Environmental Modelling, and Control System Design. 2012: 423-447.
[15]Foo M, Ooi S K, Weyer E. System identification and control of the broken river[J]. IEEE Transactions on Control Systems Technology, 2014, 22(2): 618-634.
[16]Weyer E. System identification of an open water channel[J]. Control Engineering Practice, 2001, 9(12): 1289-1299.
[17]Eurén K, Weyer E. System identification of open water channels with undershot and overshot gates[J]. Control Engineering Practice, 2007, 15(7): 813-824.
[18]Nasir H A, Weyer E. System identification of the upper part of Murray River[J]. Control Engineering Practice, 2016, 52(7): 70-92.
[19]Nasir H A, Weyer E. System identification of the upper part of Murray River[C]// European Control Conference. America: New York, 2014.
[20]Liao Wenjun, Guan Guanghua, Zhong Le, et al. Online model identification of open-channel system with high order IDZ model[C]// MATEC Web of Conferences. China: Beijing, 2018.
[21]Isermann R,Munchhof M. 動(dòng)態(tài)系統(tǒng)辨識(shí):導(dǎo)論與應(yīng)用[M]. 北京:機(jī)械工業(yè)出版社,2016.
[22]侯媛彬,汪梅,王立琦. 系統(tǒng)辨識(shí)及其MATLAB仿真[M]. 北京:科學(xué)出版社,2004.
[23]Litrico X, Fromion V. Modeling and Control of Hydrosystems[M]. London: Springer, 2009.
[24]Schuurmans J, Hof A, Dijkstra S, et al. Simple water level controller for irrigation and drainage canals[J]. Journal of Irrigation and Drainage Engineering, 1999, 125(4): 189-195.
[25]Schuurmans J, Clemmens A J, Dijkstra S, et al. Modeling of irrigation and drainage canals for controller design[J]. Journal of Irrigation and Drainage Engineering, 1999, 25(6): 338-344.
[26]管光華,鐘錁,廖文俊,等. 基于無量綱性能指標(biāo)的渠系控制器參數(shù)優(yōu)化[J]. 農(nóng)業(yè)工程學(xué)報(bào),2018,34(7):90-99. Guan Guanghua, Zhong Ke, Liao Wenjun, et al. Optimization of controller parameters based on nondimensional performance indicators for canal systems[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2018, 34(7): 90-99. (in Chinese with English abstract)
[27]Clemmens A J, Tian X, Overloop P J V, et al. Integrator delay zero model for design of upstream water-level controllers[J]. Journal of Irrigation and Drainage Engineering, 2017, 143(3): 2-9.
Identification and verification of the parameters for canal control model based on measured water level and flow
Guan Guanghua, Jia Menghao
(,,430072,)
A canal control model can be considered as a mathematical expression to represent the dynamic relationship between the canal building-gate-flow. The accuracy of the model determines the design effect of the controller. In view of the traditional canal control model, it is difficult to solve the problems which are complicated and difficult to apply in the actual engineering. In this study, a small canal monitoring system was first developed to conduct model tests. A parameter identification method was used to identify the canal control model, and then to verify the accuracy of the identification model. The parameter identification method was extended to the prototype project of South-to-North Water Diversion Project to further verify the reliability of parameter identification. Two typical flow conditions were designed in the model test, including the forward step condition and the reverse step condition, in order to fully observe the response of water level in the canal pond to the flow change, and thereby obtain the input and output information of the model. During the test, water level gauges and flow meters were used to monitor the water level and flow information in the test canal in real time, and then the monitored information was transmitted to the central monitor for storage. In the prototype project, the observation data were provided by the Administration of the Middle Route of South-to-North Water Diversion Project. The Integrator Delay (ID) model and the Integrator Delay Zero (IDZ) model were used in the control model, while, the Recursive Least Square (RLS) method was used in the identification method to identify and analyze the measured data. The results showed that recursive least squares were feasible to identify the parameters of canal control models (ID, IDZ), according to the principle of system identification in a model test and prototype observation. In the model test, the identification error was basically distributed between (-1, 1) cm, and the mean square error was less than 1.492×10-5m2. The identification results were highly consistent with the measured values, and the identification data of ID model and IDZ model were similar. This was mainly because the flow rate in the test canal was small, together with the single flow change and the low frequency flow without obviously high frequency. In the prototype project of South-to-North Water Diversion Project, the identification error was basically distributed between (-2, 2) cm, and the mean square error was less than 7.675×10-5m2. The identification data was still in high agreement with the measured value. However, when the water flow in the canal changed dramatically, the IDZ model can capture this change trend, while the ID model cannot reflect it. This was mainly because, when the water level changed drastically, the vibration frequency of water wave was faster, and the water flow was in high-frequency motion, so the response accuracy of the IDZ model was higher, which also proved its applicability to high-frequency water flow. Therefore, the IDZ model has better identification accuracy and stability than the ID model, particularly for large-scale canal projects with drastic changes in water flow. The findings can provide a certain reference for the modeling theory in the canal control and controller design, thereby to the modeling of the water transmission and distribution canal system in irrigation areas or water diversion projects.
models; canals; water levels; canal control model; Integrator Delay (ID) model; Integrator Delay Zero (IDZ) model; parameter identification; Recursive Least Square (RLS)
10.11975/j.issn.1002-6819.2020.23.011
S274.2; TV91
A
1002-6819(2020)-23-0092-07
管光華,賈夢(mèng)浩. 基于實(shí)測(cè)水位和流量數(shù)據(jù)的渠道控制模型參數(shù)辨識(shí)與驗(yàn)證[J]. 農(nóng)業(yè)工程學(xué)報(bào),2020,36(23):92-98.doi:10.11975/j.issn.1002-6819.2020.23.011 http://www.tcsae.org
Guan Guanghua, Jia Menghao. Identification and verification of the parameters for canal control model based on measured water level and flow[J]. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 36(23): 92-98. (in Chinese with English abstract) doi:10.11975/j.issn.1002-6819.2020.23.011 http://www.tcsae.org
2020-06-08
2020-11-20
國(guó)家自然科學(xué)基金項(xiàng)目(51979202,51009108)
管光華,副教授,博士,主要從事渠道系統(tǒng)自動(dòng)化運(yùn)行調(diào)度理論與技術(shù)、灌區(qū)量水理論與方法、灌排工程新結(jié)構(gòu)研究。Email:GGH@whu.edu.cn
中國(guó)農(nóng)業(yè)工程學(xué)會(huì)會(huì)員:管光華(E041700033M)
農(nóng)業(yè)工程學(xué)報(bào)2020年23期