1.前言

    蓄電(dian)池的(de)(de)充電(dian)控制系統(tong)是(shi)個非線性[1]的(de)(de)、時(shi)變的(de)(de)、有(you)干擾的(de)(de)、具有(you)純滯后的(de)(de)控制系統(tong)[2]。在充放電(dian)過(guo)程(cheng)中涉(she)及到很多參數，如充電(dian)率、最(zui)大允許充電(dian)電(dian)流(liu)、內(nei)阻、出(chu)氣點電(dian)壓(ya)、溫度、壽命等(deng)。

   傳統的(de)(de)(de)控(kong)(kong)(kong)制(zhi)(zhi)系(xi)統是(shi)(shi)建(jian)立在被(bei)控(kong)(kong)(kong)對(dui)(dui)象精(jing)確數(shu)學(xue)(xue)模型基(ji)礎上的(de)(de)(de)，如果被(bei)控(kong)(kong)(kong)對(dui)(dui)象的(de)(de)(de)數(shu)學(xue)(xue)模型很(hen)復雜(za)或者數(shu)學(xue)(xue)模型無法建(jian)立，控(kong)(kong)(kong)制(zhi)(zhi)系(xi)統就較(jiao)難實現。蓄電(dian)(dian)池充(chong)電(dian)(dian)正(zheng)是(shi)(shi)屬于(yu)這種情況，由于(yu)蓄電(dian)(dian)池的(de)(de)(de)充(chong)電(dian)(dian)過(guo)程(cheng)有(you)自己(ji)獨特的(de)(de)(de)物(wu)理化學(xue)(xue)規律(lv)，因(yin)此(ci)考慮采用模糊控(kong)(kong)(kong)制(zhi)(zhi)技術(shu)來(lai)進行蓄電(dian)(dian)池的(de)(de)(de)充(chong)電(dian)(dian)控(kong)(kong)(kong)制(zhi)(zhi) [3]。

   模糊(hu)控(kong)(kong)制[4]是(shi)以模糊(hu)集合理(li)論為基礎的(de)(de)控(kong)(kong)制手段(duan)，它是(shi)模糊(hu)系統(tong)理(li)論、模糊(hu)技(ji)術與(yu)自動控(kong)(kong)制技(ji)術相結(jie)合的(de)(de)產物,出發點是(shi)操作人(ren)員的(de)(de)控(kong)(kong)制經(jing)驗或相關專家的(de)(de)知識，在設計中(zhong)不需(xu)要(yao)建立被控(kong)(kong)對象的(de)(de)精確數學模型(xing)[5]，因而使得(de)控(kong)(kong)制機理(li)和策(ce)略易于接受(shou)與(yu)理(li)解(jie)，設計簡單(dan)，便于應用。

2.模糊控(kong)制器的結構及(ji)算(suan)法

   基于微控制(zhi)(zhi)器(qi)組成的模糊(hu)控制(zhi)(zhi)器(qi)包括模糊(hu)化接口、決(jue)策邏(luo)輯(ji)、知識(shi)庫和(he)反模糊(hu)化接口四個部分，如(ru)圖1所示。

圖1模糊控(kong)制器的結構框圖

Fig.1 The structure diagram of fuzzy controller

   在進行模糊控制算法(fa)設計時，須首先考慮智能充(chong)(chong)電系統的技術(shu)要求，體現智能充(chong)(chong)電控制技術(shu)的優勢，解(jie)決(jue)(jue)長期困擾蓄電池裝備中的效率和壽(shou)命問(wen)題，所以應當(dang)提高智能充(chong)(chong)電控制中對控制量的精(jing)度；其(qi)次(ci)，模糊控制算法(fa)必須高速可靠，對外部檢測得到的物理量要有(you)非常快的反應速度。采用新型大存(cun)儲容量的微控制器(qi)解(jie)決(jue)(jue)使(shi)用查(cha)表法(fa)所帶來的需要大容量內存(cun)的問(wen)題。

3 模糊(hu)控制器的設計

3.1雙輸(shu)入單輸(shu)出模糊控制(zhi)器

    常見的模糊控制(zhi)器[6]根據輸(shu)(shu)(shu)入(ru)(ru)與(yu)輸(shu)(shu)(shu)出個數分(fen)為單(dan)輸(shu)(shu)(shu)入(ru)(ru)單(dan)輸(shu)(shu)(shu)出、雙(shuang)輸(shu)(shu)(shu)入(ru)(ru)單(dan)輸(shu)(shu)(shu)出、多(duo)輸(shu)(shu)(shu)入(ru)(ru)單(dan)輸(shu)(shu)(shu)出和雙(shuang)輸(shu)(shu)(shu)入(ru)(ru)多(duo)輸(shu)(shu)(shu)出等(deng)幾(ji)種。平時(shi)應用較多(duo)的是雙(shuang)輸(shu)(shu)(shu)入(ru)(ru)單(dan)輸(shu)(shu)(shu)出模糊控制(zhi)器。

圖2 雙輸入單輸出(chu)模糊控制器方框圖

 Fig.2 The rectangular diagram of fuzzy cONtroller witch has two input units and one output unit

    圖(tu)2是雙輸(shu)入(ru)單(dan)輸(shu)出模(mo)(mo)糊(hu)控制(zhi)(zhi)器(qi)的(de)(de)方框圖(tu)。其中屬(shu)于(yu)論域(yu)X的(de)(de)模(mo)(mo)糊(hu)集合(he) 取自系(xi)統(tong)誤(wu)差e的(de)(de)模(mo)(mo)糊(hu)化(hua)，屬(shu)于(yu)論域(yu)Y的(de)(de)模(mo)(mo)糊(hu)集合(he) 取自系(xi)統(tong)誤(wu)差變(bian)化(hua)率 的(de)(de)模(mo)(mo)糊(hu)化(hua)，二者構成模(mo)(mo)糊(hu)控制(zhi)(zhi)器(qi)的(de)(de)二維輸(shu)入(ru)；屬(shu)于(yu)論域(yu)Z的(de)(de)模(mo)(mo)糊(hu)集合(he) 是反映控制(zhi)(zhi)量變(bian)化(hua)的(de)(de)模(mo)(mo)糊(hu)控制(zhi)(zhi)器(qi)的(de)(de)一維輸(shu)出，模(mo)(mo)糊(hu)控制(zhi)(zhi)器(qi)的(de)(de)控制(zhi)(zhi)規(gui)則[7]通常由模(mo)(mo)糊(hu)條件語句if    and    then  來表達。

3.2精確量的模糊化

   模(mo)(mo)(mo)糊控制系統(tong)中的被(bei)控對象狀態(tai)變(bian)化(hua)是連(lian)續的，系統(tong)給定值也(ye)是連(lian)續的,反映到模(mo)(mo)(mo)糊控制器(qi)輸入端的輸入量(liang)(liang)也(ye)必然是連(lian)續的。但模(mo)(mo)(mo)糊控制器(qi)由計算機(ji)構成，它(ta)只能執行(xing)離(li)散(san)處理[7]。因此(ci)模(mo)(mo)(mo)糊控制器(qi)要求輸入量(liang)(liang)是離(li)散(san)模(mo)(mo)(mo)糊量(liang)(liang)，即(ji)論域(yu)(yu)(yu)是離(li)散(san)的。對連(lian)續論域(yu)(yu)(yu)要進(jin)行(xing)離(li)散(san)化(hua)。連(lian)續論域(yu)(yu)(yu)[8]經過量(liang)(liang)化(hua)后就成為一(yi)個離(li)散(san)論域(yu)(yu)(yu)[8]。

設有連(lian)(lian)續論域(yu)[a，b]，而量化之后的離散論域(yu)為(wei) ，亦即將連(lian)(lian)續論域(yu)分為(wei)2n段，則存在(zai)系(xi)數

                                                                   （1）

   隨后(hou)，在求(qiu)出每條規(gui)則(ze)的(de)強(qiang)度(du)之(zhi)后(hou)，把相互矛盾的(de)規(gui)則(ze)中強(qiang)度(du)較(jiao)小的(de)舍去(qu)；把相同規(gui)則(ze)合成(cheng)一(yi)條規(gui)則(ze)，得到最后(hou)控制規(gui)則(ze)基。

3.3.2根據學習(xi)算法生成控(kong)制規則

   對被控(kong)(kong)(kong)(kong)對象執行手動控(kong)(kong)(kong)(kong)制(zhi)(zhi)所得到(dao)的控(kong)(kong)(kong)(kong)制(zhi)(zhi)規則是較粗(cu)糙的，有時(shi)還可(ke)能會出(chu)(chu)現(xian)控(kong)(kong)(kong)(kong)制(zhi)(zhi)死區，一(yi)個控(kong)(kong)(kong)(kong)制(zhi)(zhi)規則表中會出(chu)(chu)現(xian)空(kong)項,這是不能滿(man)足實際控(kong)(kong)(kong)(kong)制(zhi)(zhi)要(yao)求的。為了取得更滿(man)意的控(kong)(kong)(kong)(kong)制(zhi)(zhi)效果(guo)，可(ke)以(yi)對原始(shi)的控(kong)(kong)(kong)(kong)制(zhi)(zhi)規則進(jin)行改進(jin)。這時(shi)，應(ying)以(yi)粗(cu)糙的控(kong)(kong)(kong)(kong)制(zhi)(zhi)規則為基礎，通過仿真(zhen)實驗和(he)系統調(diao)試加以(yi)完善(shan)。

4.模(mo)糊智能充(chong)電系統的工作原理及(ji)結構

   智能充(chong)(chong)電(dian)(dian)(dian)(dian)(dian)(dian)系統(tong)主要(yao)由充(chong)(chong)電(dian)(dian)(dian)(dian)(dian)(dian)電(dian)(dian)(dian)(dian)(dian)(dian)源(yuan)和單片機控制(zhi)電(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)兩部分組成。220V的(de)(de)(de)交流(liu)(liu)市電(dian)(dian)(dian)(dian)(dian)(dian)經整流(liu)(liu)濾(lv)波(bo)電(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)變為(wei)(wei)脈(mo)動的(de)(de)(de)310V高壓(ya)(ya)直流(liu)(liu)。然后(hou)經DC-DC變換(huan)電(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)（脈(mo)沖(chong)功(gong)率變壓(ya)(ya)器）變為(wei)(wei)充(chong)(chong)電(dian)(dian)(dian)(dian)(dian)(dian)所(suo)需的(de)(de)(de)60V直流(liu)(liu)電(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)。為(wei)(wei)了(le)保證(zheng)輸出電(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)的(de)(de)(de)穩定性，采(cai)用了(le)UC3842對(dui)60V直流(liu)(liu)電(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)進行穩壓(ya)(ya)。二(er)次斬波(bo)電(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)主要(yao)由MOSFET管(guan)(guan)、電(dian)(dian)(dian)(dian)(dian)(dian)感、電(dian)(dian)(dian)(dian)(dian)(dian)容和二(er)極管(guan)(guan)組成，輸出24-36V的(de)(de)(de)充(chong)(chong)電(dian)(dian)(dian)(dian)(dian)(dian)電(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)。控制(zhi)部分采(cai)用C504單片機，通過對(dui)蓄(xu)電(dian)(dian)(dian)(dian)(dian)(dian)池端電(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)信(xin)號的(de)(de)(de)采(cai)集、分析處理(li)、模糊推理(li)[8]、模糊決策(ce)等，控制(zhi)二(er)次斬波(bo)電(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)中的(de)(de)(de)MOSFET管(guan)(guan)的(de)(de)(de)通斷時間來(lai)控制(zhi)充(chong)(chong)電(dian)(dian)(dian)(dian)(dian)(dian)電(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)。控制(zhi)部分還(huan)包括(kuo)對(dui)電(dian)(dian)(dian)(dian)(dian)(dian)流(liu)(liu)和溫度的(de)(de)(de)采(cai)集以及(ji)電(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)和電(dian)(dian)(dian)(dian)(dian)(dian)流(liu)(liu)的(de)(de)(de)顯(xian)示。總(zong)體結構如圖(tu)3所(suo)示。

圖3 智能充電系統總體結構框圖

5結論

  蓄電(dian)(dian)池(chi)(chi)的(de)充(chong)放電(dian)(dian)過程(cheng)是(shi)(shi)一個復雜的(de)過程(cheng)，要用(yong)精確數學模型對蓄電(dian)(dian)池(chi)(chi)充(chong)電(dian)(dian)的(de)控(kong)制(zhi)則(ze)有相當的(de)難度(du)。蓄電(dian)(dian)池(chi)(chi)的(de)充(chong)電(dian)(dian)控(kong)制(zhi)系統是(shi)(shi)個非(fei)線性的(de)、時(shi)變的(de)、有干(gan)擾的(de)、具有純(chun)滯(zhi)后的(de)控(kong)制(zhi)系統，在充(chong)放電(dian)(dian)過程(cheng)中涉及到很多(duo)參數，如充(chong)電(dian)(dian)率(lv)、最大允許(xu)充(chong)電(dian)(dian)電(dian)(dian)流、內阻、出氣點電(dian)(dian)壓、溫(wen)度(du)、壽命等。

   作者創(chuang)新(xin)點為：

(1) 隸屬函(han)數(shu)(shu) 的形狀，對控制效果影響較大。窄型(xing)隸屬函(han)數(shu)(shu)，反映模糊集(ji)(ji)(ji)合(he)具有(you)高(gao)分辨特性。如果系(xi)統(tong)誤(wu)差，采(cai)用高(gao)分辨率模糊集(ji)(ji)(ji)合(he)，則(ze)誤(wu)差控制的靈敏度(du)就會提高(gao)。在系(xi)統(tong)誤(wu)差較大的范圍內，采(cai)用具有(you)低分辨率隸屬函(han)數(shu)(shu)的模糊集(ji)(ji)(ji)合(he)；而(er)在系(xi)統(tong)誤(wu)差較小，或(huo)接近于零時，宜采(cai)用具有(you)高(gao)分辨率隸屬函(han)數(shu)(shu)的模糊集(ji)(ji)(ji)合(he)。

(2)在(zai)定義某一語言變(bian)量(liang)，如誤差、誤差變(bian)化(hua)率和(he)控制量(liang)變(bian)化(hua)的(de)(de)全部集合時(shi)，要考慮其對(dui)(dui)論域(yu)[-n,+n]的(de)(de)覆蓋(gai)程度，語言變(bian)量(liang)的(de)(de)全部模(mo)糊集合所包含的(de)(de)非(fei)零隸屬度對(dui)(dui)應的(de)(de)論域(yu)元素個數(shu)，應是模(mo)糊集合總(zong)數(shu)的(de)(de)3-4倍。

（3）查表法作為模糊控制算法有表格結構單一，修改繁瑣，缺乏靈活性的缺點。針對使用查表法作為模糊控制算法暴露的缺點，在硬件設計中與以補償，加入了一片X5045電可擦除芯片，將模糊控制表格中的數據存儲于微控制器外部存儲空間中，基本上克服了這個缺點。