發(fā)電廠(chǎng)外文翻譯

1、<p>　　國(guó)際電氣工程教育46/4 </p><p><b>　　虛擬發(fā)電廠(chǎng)</b></p><p>　　英國(guó)曼徹斯特大學(xué)電氣與電子工程學(xué)院：蓋.紐曼和馬泰爾.約瑟夫</p><p>　　電子郵件：g.newman-3@ student.manchester.ac.uk</p><p>　　摘要：依靠各種新能

2、源和可再生技術(shù)的小型發(fā)電機(jī)正在得到越來(lái)越普遍的應(yīng)用，這是由于這種發(fā)電機(jī)能夠減少溫室氣體的排放量，而這些溫室氣體正是導(dǎo)致氣候變化的首要原因。 隨著分布式要素的增加， 中央式結(jié)構(gòu)逐漸被取代，也就需要我們?cè)絹?lái)越多的了解分布式發(fā)電設(shè)備的復(fù)合運(yùn)行方式，這種設(shè)備也被稱(chēng)為虛擬電廠(chǎng)。本文介紹了一種在數(shù)學(xué)建?；A(chǔ)上進(jìn)行虛擬電廠(chǎng)開(kāi)發(fā)的一種用戶(hù)友好型工具，它可以用來(lái)作為一個(gè)電力系統(tǒng)工程教具，來(lái)輔助演示虛擬電廠(chǎng)的特點(diǎn)。</p><p>

3、　　關(guān)鍵詞:分布式發(fā)電；微型熱電聯(lián)產(chǎn)；太陽(yáng)能；虛擬電廠(chǎng)；風(fēng)力發(fā)電</p><p>　　社會(huì)環(huán)保意識(shí)的增強(qiáng)促使電力行業(yè)需要去發(fā)展新的業(yè)務(wù)，以減少二氧化碳的排放，而二氧化碳的釋放正是氣候變化的首要因素。這就導(dǎo)致需要通過(guò)新的途徑來(lái)發(fā)電，其中一些方法就是確立現(xiàn)有的綠色技術(shù)并且推廣它們，例如新型海上風(fēng)力場(chǎng)，還有一些方法就需要更新的技術(shù)，例如燃料電池。這種轉(zhuǎn)變由一系列因素引起，從由于化石燃料發(fā)電所引起的氣候變化而增強(qiáng)的環(huán)保意

4、識(shí)，到對(duì)長(zhǎng)期石油供應(yīng)安全的擔(dān)憂(yōu)。 </p><p>　　當(dāng)然，小規(guī)模的傳統(tǒng)化石燃料發(fā)電，也可以安裝備份。如光伏、風(fēng)能和微熱電聯(lián)產(chǎn)這樣的環(huán)保的綠色科技，就正在像高標(biāo)準(zhǔn)看齊，它們減少了來(lái)自電網(wǎng)的壓力，這樣就減少了傳統(tǒng)發(fā)電導(dǎo)致的二氧化碳的釋放。如果設(shè)備量夠大，那么通過(guò)這些技術(shù)，年度生產(chǎn)的電能就可以達(dá)到甚至超過(guò)每年建設(shè)所用的電量。如果一項(xiàng)建筑安裝了足夠數(shù)量的分布式發(fā)電廠(chǎng)，又能夠長(zhǎng)期的運(yùn)用這些技術(shù)，那么這一地區(qū)的電網(wǎng)就可以

5、被完全移除而自給自足。但是，光伏陣列在晚上處于休眠狀態(tài)，而風(fēng)渦輪機(jī)的運(yùn)行性能取決于風(fēng)速，微熱電聯(lián)產(chǎn)受現(xiàn)場(chǎng)供暖的要求的局限。要真正成為離網(wǎng)時(shí)需安裝的儲(chǔ)能設(shè)備，費(fèi)用是昂貴的，所以很多地方只有在產(chǎn)大于出的時(shí)候才將能量饋入電網(wǎng)，當(dāng)入不敷出時(shí)就只能從電網(wǎng)中取電。因此，依靠化石燃料燃燒發(fā)電還將持續(xù)一段時(shí)間。用這種分布式發(fā)電廠(chǎng)建模電網(wǎng)就假設(shè)了這些分布式發(fā)電廠(chǎng)可以作為負(fù)載的抵消，這種假設(shè)可以成為一個(gè)接受點(diǎn)，然而，如果小規(guī)模的發(fā)電不敷負(fù)載，或者減少了網(wǎng)路

6、的可控性，這種假設(shè)就不能成立。在這一點(diǎn)上，通過(guò)電網(wǎng)的分配，電流就發(fā)生明顯的改變，電網(wǎng)內(nèi)母線(xiàn)的電壓也受到影響。最重要的是，安裝的網(wǎng)絡(luò)物理硬件能夠在一定的條件范圍內(nèi)控制功率，但是這些條件通常都是單向的，也</p><p>　　預(yù)測(cè)輸出的虛擬電廠(chǎng)所面臨的挑戰(zhàn)，</p><p>　　從光伏、風(fēng)能和微熱電聯(lián)產(chǎn)預(yù)測(cè)功率的輸出的首要問(wèn)題就是，這些設(shè)備是由不可預(yù)知源驅(qū)動(dòng)。</p><p

7、>　　準(zhǔn)確的預(yù)知風(fēng)速、太陽(yáng)輻射和溫度是一種高級(jí)主題的氣象研究，但是它們都是復(fù)雜的問(wèn)題。另外，精準(zhǔn)的預(yù)測(cè)還存在一些問(wèn)題，就是一些局部的差異已經(jīng)超出的預(yù)測(cè)的范圍?？傮w而言，這個(gè)問(wèn)題的難度還是很大的。在這項(xiàng)工作中，我們關(guān)注了三種分布式發(fā)電，分別稱(chēng)為光伏發(fā)電、風(fēng)能和微熱電聯(lián)產(chǎn)，為了獲得這些技術(shù)中的一個(gè)虛擬電廠(chǎng)的形式模型，首要任務(wù)就是要了解每個(gè)技術(shù)的特征。</p><p><b>　　風(fēng)力發(fā)電的基礎(chǔ)<

8、/b></p><p>　　如上所述，地面水平風(fēng)速在一定程度上的大規(guī)模預(yù)測(cè)具有不確定性，并且由于局部動(dòng)蕩也會(huì)導(dǎo)致預(yù)測(cè)不準(zhǔn)確。接下來(lái)的工作是在假定沒(méi)有預(yù)測(cè)誤差的情況下進(jìn)行的，盡管實(shí)際情況并非如此，這種假設(shè)是為了簡(jiǎn)化計(jì)算。測(cè)量風(fēng)速中，其余的不確定性就是由局部差異引起的，而這可以進(jìn)行模擬。在參考文獻(xiàn)2中，平均風(fēng)速為一個(gè)給定的海拔高度作為式1，由于地形帶來(lái)的標(biāo)準(zhǔn)偏差由式2所決定，從廣義上講，測(cè)量點(diǎn)越高，風(fēng)速將會(huì)越大

9、約平滑，方程1和2分別提供了平均風(fēng)速，和風(fēng)速的標(biāo)準(zhǔn)偏差。在短期測(cè)量時(shí)間內(nèi)，分配功能不能精確的呈現(xiàn)，由于假定分配沒(méi)有差異，有了這些數(shù)據(jù)，標(biāo)準(zhǔn)的分配就被選出。有了統(tǒng)一分配，標(biāo)準(zhǔn)偏差和平均風(fēng)速，就可以估計(jì)風(fēng)俗的可能性分布。為了產(chǎn)生有用的輸出，這種分布必須同渦輪機(jī)動(dòng)力曲線(xiàn)聯(lián)系起來(lái)。fIG1展示了理想的動(dòng)力曲線(xiàn)。用統(tǒng)計(jì)學(xué)理論，就有可能將兩種曲線(xiàn)聯(lián)系起來(lái)從而產(chǎn)生一種可能性功率輸出。盡管這可能作為獨(dú)立的單元，但以這種方式處理發(fā)電機(jī)輸出是不切實(shí)際的，因

10、為它們必須合并成單一的虛擬電廠(chǎng)輸出。雖然使用連續(xù)型數(shù)據(jù)時(shí)可能的，但是在過(guò)程中使用離散型數(shù)據(jù)是更方便的。因此，風(fēng)力渦輪機(jī)的輸出以離散形式的幾率更大。</p><p><b>　　太陽(yáng)能發(fā)電基礎(chǔ)知識(shí)</b></p><p>　　地面的太陽(yáng)能輻射也具有一定程度的不確定性，對(duì)于福照度，盡管可能涉及一些結(jié)構(gòu)性陰影取決于光伏陣列的安置，但最大的不確定性還是來(lái)自于云變的性質(zhì)，雖然較

11、大的云層可以從太空中看見(jiàn)也能被預(yù)測(cè)。但一些較小的云層則不容易被注意到，并且在預(yù)測(cè)輻照度上引入到一個(gè)誤區(qū)。參考文獻(xiàn)3給出了衛(wèi)星8在1平方公里在15分鐘內(nèi)的氣象分辨率，與r.m.s.在1分鐘內(nèi)存在20%的誤差。雖然在他們的工作中并沒(méi)有提到特別擬合的概率分布，但他們使用均勻分布同風(fēng)模型保持一致。均與分布的選擇呈現(xiàn)出了這些誤差。有了這個(gè)信息，光伏陣列的輸出功率就可以由研究中使用的模型確定。另外，這種輸出在本質(zhì)上是離散的，易于操作和組合。<

12、/p><p><b>　　微型熱電聯(lián)產(chǎn)的基礎(chǔ)</b></p><p>　　微型熱電聯(lián)產(chǎn)更難預(yù)測(cè)，因?yàn)樗枰毡榈臄?shù)據(jù)，以便產(chǎn)生輸出。微型熱電聯(lián)產(chǎn)是以鍋爐裝置為單元，通過(guò)它們加熱，把廢熱轉(zhuǎn)換成一些電能效率。這些設(shè)置然后就一直加熱直到溫控器超過(guò)上限閥值，就關(guān)閉鍋爐加熱用途。一旦設(shè)置失去足夠的能量而降到下面較低的閥值時(shí)，溫控器的鍋爐就再次打開(kāi)。鑒于預(yù)測(cè)數(shù)據(jù)包含周?chē)h(huán)境溫度，就

13、有可能決定提供鍋爐可能的占空比，也可預(yù)測(cè)設(shè)備的熱工參數(shù)。例如從內(nèi)部到外觀(guān)的的整體熱阻和熱容量。通過(guò)對(duì)設(shè)備的熱容量和熱電阻選擇適當(dāng)?shù)闹担⑶抑榔渌鐭犷~定功率和轉(zhuǎn)換效率的鍋爐參數(shù)，就可能確定出鍋爐的占空比，從而確定微熱電聯(lián)產(chǎn)的瞬時(shí)輸出功率。</p><p><b>　　瞬時(shí)功率和平均功率</b></p><p>　　如上面所述，確定瞬時(shí)功率是非常重要的，但也不能忽略另

14、一個(gè)重要的數(shù)據(jù)也就是平均功率的消耗。在預(yù)測(cè)的時(shí)間范圍內(nèi)，瞬時(shí)功率描述了在任何一點(diǎn)時(shí)的虛擬電廠(chǎng)發(fā)電，這在整個(gè)預(yù)測(cè)的時(shí)間范圍內(nèi)是不同于平均功率的，平均功率是長(zhǎng)期時(shí)間的功率。雖然瞬時(shí)功率是用于在一定時(shí)間范圍內(nèi)確立網(wǎng)絡(luò)流量的，而靜態(tài)的平均功率則對(duì)長(zhǎng)期時(shí)間內(nèi)可用功率更為有益，并且會(huì)顯示出整體預(yù)測(cè)流量。風(fēng)力發(fā)電機(jī)組長(zhǎng)期的功率輸出可以用一種簡(jiǎn)單的途徑被聚集起來(lái)。雖然這是一種簡(jiǎn)化的途徑，但在確定這種方法的準(zhǔn)確度時(shí)，時(shí)間長(zhǎng)短的選擇還是起著關(guān)鍵的作用。通過(guò)

15、選擇下降到10到60分之間的長(zhǎng)度，就可以得出風(fēng)速方差的最小化。雖然這并沒(méi)有保證速度會(huì)完全的適合分布，但是它已經(jīng)最大限度的提高了它所能提高的機(jī)遇。光伏的平均輸出功率也可以用一個(gè)簡(jiǎn)單的函數(shù)算出，因?yàn)樵频男纬?，造成了初始誤差。雖然構(gòu)成基準(zhǔn)預(yù)測(cè)輻射度的較大的云層比較容易預(yù)測(cè)，但是那些規(guī)模較小的浮云會(huì)引起誤差也可以被平均掉。由于其規(guī)模和速度，它們產(chǎn)生的波動(dòng)類(lèi)似占空比。雖然這在瞬時(shí)功率是會(huì)產(chǎn)生誤差，但這種行為較為平滑，因此在較長(zhǎng)時(shí)間內(nèi)能夠得到一個(gè)相

16、對(duì)準(zhǔn)確的平均結(jié)果。</p><p>　　微型熱電聯(lián)產(chǎn)更加難以描述，由于發(fā)電機(jī)偏振和關(guān)閉行為，輸出就作為一個(gè)脈沖周期出現(xiàn)。由此獲得的輸出功率完全依賴(lài)于鍋爐循環(huán)頻率和觀(guān)察期。雖然方程冗長(zhǎng)派生，但并不復(fù)雜，可以分為四種不同情況分析。</p><p><b>　　結(jié)合總設(shè)備</b></p><p>　　如上所述，隨機(jī)輸出功率圖的組合形成了一個(gè)整體的功率

17、輸出圖，大大加快利用離散數(shù)據(jù)的機(jī)會(huì)。要添加兩個(gè)分布式發(fā)電廠(chǎng)單元的功率輸出就要使用一系列連續(xù)的數(shù)據(jù)，包括使用大量的集成，而這一過(guò)程中也要使用離散型數(shù)據(jù)，因?yàn)檫@些離散型數(shù)據(jù)能能夠?yàn)槊總€(gè)據(jù)點(diǎn)解決簡(jiǎn)單的乘除。這樣做是為了每個(gè)數(shù)據(jù)點(diǎn)的輸出數(shù)據(jù)，知道每個(gè)分布式發(fā)電廠(chǎng)單元都包含兩種功率即輸出功率和平均功率。考慮到隨機(jī)過(guò)程，在哪里要求分布式發(fā)電廠(chǎng)被認(rèn)為是不重要的。</p><p><b>　　虛擬電廠(chǎng)的實(shí)施</b

18、></p><p>　　為了生成具有自動(dòng)計(jì)算和組合的能力的程序，就需要精心的組織程序結(jié)構(gòu)，還要知道怎樣交互程序以及同用戶(hù)的的數(shù)據(jù)交流。該程序的用戶(hù)終端是分布式發(fā)電廠(chǎng)的管理者，他們應(yīng)該能夠盡可能的聚集輸出功率，這也是最頻繁重復(fù)的工作。此外，管理者應(yīng)該能夠毫不費(fèi)力的從系統(tǒng)中添加或移走發(fā)電機(jī)。并且，管理者應(yīng)該能夠一次處理多種情況，無(wú)論是手動(dòng)的重復(fù)聚集工作，還是一起進(jìn)行的裝載預(yù)測(cè)程序列表的任務(wù)。從系統(tǒng)運(yùn)營(yíng)商進(jìn)來(lái)定的

19、數(shù)據(jù)分為兩個(gè)階段：分布式發(fā)電機(jī)的單位參數(shù)輸入，和預(yù)測(cè)參數(shù)項(xiàng)的輸入。在使用預(yù)測(cè)參數(shù)前需要給定單位參數(shù)，預(yù)測(cè)參數(shù)需要執(zhí)行程序模式。然而，無(wú)論程序模式運(yùn)行到何時(shí)都需要進(jìn)行第二次輸入， 除非虛擬電場(chǎng)的配置被修改，否則只是在虛擬電廠(chǎng)設(shè)置期間才需要輸入單位數(shù)據(jù)。這一代的單位安裝是費(fèi)時(shí)的，并且需要由系統(tǒng)內(nèi)部的發(fā)電機(jī)支配。因此，虛擬電廠(chǎng)發(fā)電機(jī)安裝應(yīng)該能夠被保存和加載，以確保能夠迅速的轉(zhuǎn)換到程序的主要部分。程序的輸出數(shù)據(jù)以數(shù)據(jù)集的形式來(lái)針對(duì)輸出功率。每個(gè)

20、預(yù)測(cè)數(shù)據(jù)集會(huì)產(chǎn)生兩套數(shù)據(jù)，瞬時(shí)輸出功率和長(zhǎng)期的平均輸出功率。</p><p>　　這是程序內(nèi)信息流動(dòng)的終端，而不是也不嘗試成為一個(gè)圖形顯示屏或電子表格類(lèi)型的應(yīng)用程序。在創(chuàng)造這些程序方面，另外一些人已經(jīng)做得很好了，所以數(shù)據(jù)應(yīng)該以這樣一種方式傳達(dá)給用戶(hù)，即它可以很容易的從程序轉(zhuǎn)移成圖形顯示包。</p><p><b>　　圖形用戶(hù)界面</b></p><

21、;p>　　根據(jù)程序用戶(hù)流量保持到最低限度的規(guī)定，最基本的界面要建立整合上述的特定的因素。產(chǎn)生和預(yù)測(cè)的輸入數(shù)據(jù)，都能夠由端口完成，同時(shí)上述的下載及虛擬電廠(chǎng)設(shè)備的保存也可以完成。此外，它還可以使用窗口上最左面的兩個(gè)大框編輯和加載發(fā)電機(jī)參數(shù)，最左面的框能夠選擇虛擬電廠(chǎng)的發(fā)電機(jī)，而右面的這個(gè)可以顯示出所選發(fā)電機(jī)的參數(shù)。通過(guò)簡(jiǎn)單的編輯相關(guān)值就可以改變?nèi)魏伟l(fā)生器的參數(shù)，當(dāng)然，這些新值不會(huì)自動(dòng)保存到文件內(nèi)，但是對(duì)于虛擬電廠(chǎng)來(lái)說(shuō)，這些參數(shù)可以或大

22、或小的修改，而不必移除發(fā)電機(jī)來(lái)重新輸入不同的參數(shù)。除了上述的框，程序內(nèi)的流動(dòng)是從左到右。最左邊的六個(gè)按鈕允許投入新的發(fā)電機(jī)，移除發(fā)電機(jī)，并且保存和下載發(fā)電機(jī)配置。中間的四個(gè)按鈕允許輸入預(yù)測(cè)數(shù)據(jù)來(lái)完成方案的主要目的。它們被分成兩組然后再分成兩組。首先，他們根據(jù)用戶(hù)是需要運(yùn)行單一預(yù)測(cè)還是需要運(yùn)行一系列預(yù)測(cè)將用戶(hù)分成兩組。這種分組是垂直的，最左邊的按鈕只處理單一的預(yù)測(cè)。第二次分組是根據(jù)用戶(hù)希望運(yùn)行優(yōu)化版本還是非優(yōu)化版本，這可以作為一個(gè)有用的示

23、范工具，來(lái)顯示運(yùn)行優(yōu)化過(guò)程節(jié)省了時(shí)間。兩種輸出是明顯不同的，這種分組是橫向的，非優(yōu)化處理時(shí)使用最頂端的按鈕。稍后在討論優(yōu)化本身。窗口的最</p><p><b>　　后端</b></p><p>　　雖然前端是非常簡(jiǎn)單直接的，但它和它的進(jìn)程完全建立在后端正確的架構(gòu)上。包括圖形用戶(hù)界面接口、后端是一個(gè)單層。圖5 可以看出后端的布局，該功能可以分為兩大類(lèi)，即數(shù)據(jù)管理和數(shù)據(jù)

24、處理。雖然數(shù)據(jù)處理涉及動(dòng)態(tài)的數(shù)據(jù)管理，但還是有必要區(qū)分兩者，這樣才可以實(shí)現(xiàn)代碼重用。雖然數(shù)據(jù)處理比較復(fù)雜，但它是建立在正確的數(shù)據(jù)管理上。數(shù)據(jù)管理功能作為后端部分運(yùn)行的第一個(gè)例程，在這里數(shù)據(jù)結(jié)構(gòu)需要初始化。管理者也需要處理發(fā)電機(jī)的添加、移除、選擇以及虛擬電廠(chǎng)配置的下載和保存。對(duì)于預(yù)測(cè)數(shù)據(jù)，數(shù)據(jù)管理職能包括讀取這些數(shù)據(jù)序列表，而這些序列表本身就包括讀取和存儲(chǔ)預(yù)測(cè)數(shù)據(jù)。最后一點(diǎn)，數(shù)據(jù)管理功能能將輸出的數(shù)據(jù)呈現(xiàn)在屏幕上。但是，管理職能除了組織數(shù)

25、據(jù)不做其他任何事。管理和處理之間的銜接數(shù)據(jù)處理功能的切入點(diǎn)，通過(guò)數(shù)據(jù)收集處理功能收集輸出數(shù)據(jù)。數(shù)據(jù)處理功能處理入口處給它的數(shù)據(jù)，但是這個(gè)數(shù)據(jù)是沒(méi)有形式的，處理功能旁邊是一個(gè)收集處理器，能夠進(jìn)行數(shù)據(jù)管理。每一個(gè)發(fā)電機(jī)型號(hào)都有一級(jí)，一旦初始化，這一級(jí)能夠掌握有關(guān)發(fā)電機(jī)的參數(shù)。此外，每一級(jí)都可以建立自己的輸出功率圖，即瞬時(shí)功率和平均功率。更復(fù)雜的是建立這些圖的數(shù)學(xué)方面，也就是這些圖是由位于他們</p><p><

26、b>　　模型輸出</b></p><p>　　模型和程序有更易于觀(guān)察的特點(diǎn)，它們中的一些比其它的更直觀(guān)。然而，一般情況下，可以說(shuō)， 虛擬電廠(chǎng)單位數(shù)量增加有利于相對(duì)的減少平均功率的變化。因?yàn)樗岣吡穗娫吹目煽啃?，這對(duì)于系統(tǒng)運(yùn)營(yíng)商來(lái)說(shuō)是有益的。每種技術(shù)也有相關(guān)的輸出模式。例如，光伏組件只在白天進(jìn)行預(yù)期輸出，而CHP對(duì)光伏發(fā)電顯示相反的操作，僅在較涼爽的時(shí)候進(jìn)行操作。這些技術(shù)的預(yù)期性發(fā)電具有季節(jié)性，光

27、伏發(fā)電在夏天較有優(yōu)勢(shì)，CHP則在冬季的幾個(gè)月內(nèi)優(yōu)勢(shì)更明顯。風(fēng)力渦輪機(jī)根據(jù)前一天的研究可以預(yù)測(cè)顯示整天的輸出。但是，程序使用的模型有其局限性。這種模型假設(shè)預(yù)測(cè)數(shù)據(jù)是完全正確的，該方法只有在預(yù)測(cè)時(shí)間內(nèi)推出了自己的問(wèn)題，并沒(méi)有完全準(zhǔn)確的預(yù)測(cè)數(shù)據(jù)。由于模型沒(méi)有考慮到這種不確定性，所以系統(tǒng)操作對(duì)未來(lái)的預(yù)測(cè)不能獲得有益的啟示。該模型仍然會(huì)為預(yù)測(cè)產(chǎn)生一個(gè)近似的數(shù)據(jù)，但是由于不確定性就不能提供變化度。</p><p><b

28、>　　案例研究 </b></p><p>　　案例研究給出了三臺(tái)發(fā)電機(jī)設(shè)置，機(jī)風(fēng)電、光伏和熱電聯(lián)產(chǎn)。額定功率為1.5千瓦的風(fēng)力發(fā)電機(jī)組在10米的高度，切入速度為4MS-1，額定轉(zhuǎn)速為12MS-1，切除速度為25MS-1，站點(diǎn)粗糙度為0.7；最大額定管道儲(chǔ)運(yùn)為1.5千瓦的光伏陣列，表面方位角為180°，溫度系數(shù)為0.005，傾斜角為10°。微熱電聯(lián)產(chǎn)，鍋爐熱電力轉(zhuǎn)化率為0.3，

29、額定功率為10千瓦，恒熱點(diǎn)為19°-20°，安裝在建筑物的熱容量為800JK-1，熱阻為0.005千瓦-1.首先,機(jī)組孤立。瞬時(shí)風(fēng)力，光伏發(fā)電和熱電聯(lián)產(chǎn)對(duì)天氣預(yù)報(bào)測(cè)試，風(fēng)速為15ms-1，輻照為800wm-2,太陽(yáng)方位角為160°，太陽(yáng)高度角為60°，溫度為8°，經(jīng)過(guò)900s的時(shí)間。蒙特卡羅模擬，使用相同發(fā)電機(jī)參數(shù)和氣象參數(shù)，測(cè)試同時(shí)進(jìn)行以驗(yàn)證其有效性。使用一百萬(wàn)的隨機(jī)樣本，風(fēng)，光伏和熱

30、電的誤差分別為0.010%,0.008%,0.002%，然后在上述相同條件下經(jīng)營(yíng)發(fā)電機(jī)組，蒙特卡洛模擬運(yùn)行，對(duì)一萬(wàn)個(gè)隨機(jī)點(diǎn)進(jìn)行比較預(yù)測(cè)，對(duì)組合模型給出的平均誤差是0.005%。這似乎比起上面說(shuō)的有輕微的下降，然而，熱電聯(lián)產(chǎn)機(jī)組比光伏和風(fēng)作為單獨(dú)數(shù)據(jù)點(diǎn)的數(shù)量多兩倍，平均錯(cuò)誤也最低</p><p><b>　　結(jié)論</b></p><p>　　本文概述了虛擬電廠(chǎng)模型，展示

31、了這樣一種視角，即如何應(yīng)用這個(gè)模型使用用戶(hù)友好型工具。在概率模型建設(shè)中，有三大技術(shù)已經(jīng)取了進(jìn)展性的討論：風(fēng)能，太陽(yáng)能，微熱電聯(lián)產(chǎn)，還有伴隨這些模型所開(kāi)發(fā)的工具，并把重點(diǎn)放在模型快速、易于掌握和易于使用上。而當(dāng)前模型的局限處就是無(wú)法結(jié)合不準(zhǔn)確性來(lái)預(yù)測(cè)數(shù)據(jù)。今后的工作主題將包括整合未來(lái)1個(gè)小時(shí)內(nèi)不精準(zhǔn)預(yù)測(cè)，和對(duì)復(fù)合虛擬電廠(chǎng)輸出的不精確預(yù)測(cè)的影響。</p><p>　　International Journal of

32、Electrical Engineering Education 46/4</p><p>　　Characterising virtual power plants</p><p>　　Guy Newman and Joseph Mutale</p><p>　　School of Electrical and Electronic Engineering, Un

33、iversity of Manchester, Manchester, UK</p><p>　　E-mail: g.newman-3@student.manchester.ac.uk</p><p>　　Abstract The use of small-scale generation based on disparate new and renewable technologies

34、is</p><p>　　becoming more prevalent due to the imperative to reduce greenhouse gas emissions that are thought</p><p>　　to be the chief cause of climate change. As penetration of these distribute

35、d elements increases,</p><p>　　displacing central generation, there is a growing need to understand the composite behaviour of</p><p>　　groups of Distributed Generation (DG) devices, also known

36、as Virtual Power Plants (VPPs). This paper</p><p>　　presents an overview of the mathematical modelling of the VPP leading to the development of a</p><p>　　user-friendly tool that can be used as

37、a power system engineering teaching aid to demonstrate the</p><p>　　characteristics of VPPs.</p><p>　　Keywords distributed generation; micro CHP; solar PV; virtual power plant; wind generation&l

38、t;/p><p>　　The growth of environmental awareness in society is putting pressure on the electric</p><p>　　power generation business to reduce CO2 emissions, thought to be the chief cause</p>

39、<p>　　of climate change. This is leading to new methods to generate power. Some of these</p><p>　　methods build upon existing green technologies and expand their size, such as new</p><p>　　

40、large-scale offshore wind farms, while other methods employ newer technologies </p><p>　　such as fuel cells. The shift is driven by a combination of factors ranging from</p><p>　　increased aware

41、ness of climate change due to power generation from fossil fuels</p><p>　　through to concerns about long-term security of oil supplies. Aside from energy</p><p>　　trading companies, small-scale

42、Distributed Generation (DG) technologies are available</p><p>　　for commercial and residential buildings, which can offer similar green credentials</p><p>　　for a smaller scale of power generati

43、on. Of course, small-scale conventional</p><p>　　fossil fuel based generation can also be installed for back-up purposes.</p><p>　　Environmentally, the installation of green technologies such as

44、 photovoltaic (PV),</p><p>　　wind and micro combined heat and power (μCHP, or microCHP) is benefi cial in</p><p>　　that it reduces the electric power drawn from the grid, thus reducing the carbo

45、n</p><p>　　dioxide released from conventional electricity generation. If installed in large enough</p><p>　　quantities, the annual electrical power produced by these technologies can equal or<

46、;/p><p>　　become greater than the annual power usage of the building, or site. For a building</p><p>　　with this quantity of DG installed it would be logical to suggest that the site could</p&g

47、t;<p>　　be removed from the grid entirely and become self-sustaining, if it were not for the</p><p>　　intermittency of the technologies involved. PV arrays are dormant at night, wind</p><p&

48、gt;　　turbines are a slave to the wind speed, and microCHP is a slave to the site heating</p><p>　　requirements. To become truly off-grid requires the installation of energy storage</p><p>　　devi

49、ces, which is costly, so many sites simply feed power back into the grid when</p><p>　　they produce more than they consume, and take power out of the grid when they</p><p>　　consume more than th

50、ey produce. The reliance on fossil fuels (for conventional</p><p>　　generation) is, therefore, maintained.</p><p>　　The approach to modelling power networks with DG has been to assume the DG<

51、/p><p>　　as a load offset.1 This assumption is acceptable to a point; however, the assumption</p><p>　　fails if the small-scale generation outstrips the load, and leads to a reduction of</p>

52、<p>　　controllability of the network. At this point the fl ow of power through the distribution</p><p>　　network can alter signifi cantly, affecting bus voltages within the power network.</p>&l

53、t;p>　　Most importantly, the physical hardware of the network is installed to control power</p><p>　　fl ow for a range of conditions, but these conditions are usually entirely unidirectional,</p>&l

54、t;p>　　from the remote power station to the consumer. The reversal of power fl ow</p><p>　　is beyond the scope of the hardware to control, so the hardware must be changed or</p><p>　　modifi ed

55、 to enable bidirectional power fl ow. On a smaller scale, the fl ow of power</p><p>　　within the low voltage (LV) network is adjusted by the reversal of power, affecting</p><p>　　the line voltag

56、e profi le. Zero transfer of power from the distribution network to the</p><p>　　LV network does not guarantee that the LV lines are not carrying power,</p><p><b>　　anymore.</b></

57、p><p>　　The virtual power plant (VPP) is an artifi cial layer placed between the system</p><p>　　operator and the DG user. It is formed out of individual DG units, and co-ordinates</p><p

58、>　　the actions of the units as a whole (where this is technologically possible), rather</p><p>　　than leaving units to govern themselves individually. In this way it acts as a facilitator,</p><

59、p>　　as the network operator can instruct the VPP operator to control the output of</p><p>　　the plant, requiring more or less power output. It also acts to amalgamate the power</p><p>　　outpu

60、t from the component generators. This is useful for owners of intermittent DG,</p><p>　　as the amalgamation of devices acts to reduce the complexity of their output power,</p><p>　　which is stoc

61、hastic in nature to predict.</p><p>　　Challenge of predicting the output of a VPP</p><p>　　The primary problem with predicting the output power from microCHP, solar PV</p><p>　　and

62、wind is that these devices are driven by unpredictable sources. Accurate prediction</p><p>　　schemes for wind speed, solar irradiance and temperature are all advanced topics</p><p>　　in meteorol

63、ogical studies, but they are complex problems. In addition to the problems</p><p>　　of accurate forecasting, there are local variations which are beneath the scale</p><p>　　of the forecast. Over

64、all, the problem is of sizeable magnitude. In this work we focus</p><p>　　on three DG technologies, namely wind, microCHP and solar power. In order to</p><p>　　derive aggregate models for these

65、technologies in the form of a VPP, it is fi rst necessary</p><p>　　to understand the characteristics of each technology.</p><p>　　Wind power basics</p><p>　　The wind speed at ground

66、 level is, as mentioned above, characterised by a degree</p><p>　　of uncertainty in the large-scale forecasting and also by localised turbulence below</p><p>　　the scale of the forecasted data.

67、It is assumed in the following work that there is no</p><p>　　forecasting error, although this is not the case. This assumption is made to simplify</p><p>　　the computations. The remaining uncer

68、tainty in the wind speed is dependent on</p><p>　　localised variations, and this can be modelled. In Ref. 2, the average wind speed for</p><p>　　a given altitude in given as eqn (1). The standar

69、d deviation due to the geometry of</p><p>　　the landscape is determined from eqn (2), also from Ref. 2. Broadly speaking, the</p><p>　　higher the measurement point, the greater and smoother the

70、wind speed will be.</p><p>　　Equations 1 and 2 provide the mean speed, and the standard deviation of the wind</p><p>　　speed, respectively.</p><p>　　where z is the height of interes

71、t, z0 is the roughness length, h is the measurement</p><p>　　height, and U is the wind speed.</p><p><b>　　?</b></p><p>　　where σ is the standard deviation.</p>&l

72、t;p>　　No distribution function is noted for accurately representing small time period</p><p>　　measurements, so the uniform distribution was chosen for use with the data, as this</p><p>　　ass

73、umes no bias of distribution. With the uniform distribution, the mean wind speed</p><p>　　and the standard deviation, the wind speed probability distribution can be evaluated.</p><p>　　This dist

74、ribution must be combined with the turbine power curve in order to produce</p><p>　　a useful output. An ideal power curve is shown in Fig. 1. Using statistical theory, it</p><p>　　is possible to

75、 combine the two curves to produce a single probabilistic power output.</p><p>　　Although this can be done for a single unit, it is impractical to manage the generator</p><p>　　outputs this way,

76、 as they must be combined into a single output for the VPP. Using</p><p>　　continuous data is possible, but it is more convenient for the process to use discrete</p><p>　　data. Therefore the pro

77、babilistic output is stored in a discrete form for the wind turbines,</p><p>　　PV arrays, μCHP generators and the VPP.</p><p>　　Fig. 1 Wind speed power curve.</p><p>　　Solar power b

78、asics</p><p>　　The solar irradiance at ground level is also characterised by a degree of uncertainty.</p><p>　　For irradiance, the largest uncertainty is due to the variable nature of clouds, al

79、though</p><p>　　some structural shadowing may be involved, depending on the placement of the PV</p><p>　　array. Whilst the larger cloud formations are visible from space and can be predicted,<

80、;/p><p>　　smaller cloud formations are less easily noticed and introduce a degree of error into</p><p>　　forecasted irradiance. Reference 3 gives the Meteosat 8 satellite a resolution of 1 km2</

81、p><p>　　and 15 min, with an r.m.s. error of 20% at 15 min. Although no probability distribution</p><p>　　is mentioned in their work as being particularly well fi tted, they use</p><p>

82、　　the uniform distribution in their work and for consistency with the wind model, the</p><p>　　uniform distribution is chosen to represent this error. With this information, the</p><p>　　power o

83、utput for a PV array can be determined using models from the study</p><p>　　described in Ref. 4. Once again, this output is discrete in nature for ease of manipulation</p><p>　　and combination.&

84、lt;/p><p>　　MicroCHP basics</p><p>　　MicroCHP is more diffi cult to predict, as it requires more general data in order to</p><p>　　produce its output. The boiler in a microCHP unit is

85、fi red to heat a building, and the</p><p>　　waste heat is converted at some effi ciency into electrical energy. The building then</p><p>　　heats up until the thermostat is above the upper thresh

86、old, turning off the boiler for</p><p>　　heating purposes. Once the building has lost enough energy to fall beneath the lower</p><p>　　threshold of the thermostat, the boiler is turned on again.

87、</p><p>　　Given that forecasted data contains ambient temperature it is possible to determine</p><p>　　the likely duty cycle of the boiler provided that an educated prediction can be made</p&

88、gt;<p>　　of the building’s thermal parameters, such as its overall thermal resistance from</p><p>　　exterior to interior and heat capacity. By choosing appropriate values for the building’s</p>

89、<p>　　heat capacity and thermal resistance, and knowing other boiler parameters such</p><p>　　as thermal power rating and conversion effi ciency, it is possible to determine the</p><p>　　d

90、uty cycle and consequently determine the instantaneous stochastic output power of</p><p><b>　　microCHP.</b></p><p>　　Instantaneous and long-term power</p><p>　　While it

91、is important to determine the instantaneous power, as above, it must not be</p><p>　　forgotten that another important quantity is the average power over the prediction</p><p>　　period. The insta

92、ntaneous power provides the likely power generation for the VPP</p><p>　　at any point during the prediction time frame, which is different to the average</p><p>　　power throughout the prediction

93、 time frame, described as being the long-term</p><p>　　power. Whilst the instantaneous power is useful for determining fl ows in the network</p><p>　　during the time frame, the more static long-

94、term power is a more useful guide to the</p><p>　　available power for the time frame, and will show the overall predicted fl ow of</p><p><b>　　power.</b></p><p>　　The wi

95、nd turbine long-term power output can be generated using a simple mean</p><p>　　function. Although this is a simplifi cation, the period length chosen plays a key role</p><p>　　in determining ho

96、w inaccurate this method is. By choosing lengths which fall</p><p>　　between 10 and 60 min, the variance in the wind speeds is minimised. Whilst this</p><p>　　is no guarantee that the speeds wil

97、l exactly fi t the distribution, it maximises the</p><p>　　chance that it will.</p><p>　　The PV long-term power output can also be generated using a simple averaging</p><p>　　functi

98、on, due to the cloud formations which cause the initial inaccuracy. Whilst the</p><p>　　larger cloud formations which form the baseline forecast irradiance are easier to</p><p>　　predict, the sm

99、aller, fl eeting clouds cause inaccuracies which can be averaged out.</p><p>　　Due to their size and speed, they produce fl uctuations which are of a similar nature</p><p>　　to the duty cycle. W

100、hilst this produces errors in the instantaneous power, this behaviour</p><p>　　is smoothed out over longer time periods to produce a relatively accurate mean</p><p><b>　　result.</b>&

101、lt;/p><p>　　MicroCHP is more difficult to characterise. Due to the polarised on-or-off behaviour</p><p>　　of the generator, the output appears as a pulsed cycle. The probabilistic longterm</p>

102、;<p>　　power output obtained from this is entirely dependent on the boiler cycle</p><p>　　frequency and the observation period. Whilst the governing equations are lengthy</p><p>　　to deri

103、ve, they are not at all complicated, resolving into four distinct cases.</p><p>　　Combining the DG units</p><p>　　The combination of the probabilistic power output graphs to form an overall prob

104、abilistic</p><p>　　power output graph is, as mentioned above, greatly accelerated by using</p><p>　　discrete data. To add the probabilistic power outputs of two DG units using continuous</p&g

105、t;<p>　　data involves the use of a great many integrations, and whilst the process also</p><p>　　has to be done using discrete data, the discrete case resolves to simple multiplication</p><