1.前言

    蓄電(dian)(dian)池的(de)充電(dian)(dian)控制(zhi)系(xi)統是個非線性[1]的(de)、時變的(de)、有干(gan)擾(rao)的(de)、具(ju)有純滯后的(de)控制(zhi)系(xi)統[2]。在充放電(dian)(dian)過程中涉(she)及到很多參(can)數，如充電(dian)(dian)率、最大允(yun)許充電(dian)(dian)電(dian)(dian)流、內阻、出(chu)氣點電(dian)(dian)壓、溫度、壽(shou)命等。

   傳統(tong)的(de)控(kong)制(zhi)系統(tong)是建(jian)立在被(bei)控(kong)對象(xiang)精確數學(xue)模(mo)(mo)型(xing)基礎(chu)上的(de)，如(ru)果被(bei)控(kong)對象(xiang)的(de)數學(xue)模(mo)(mo)型(xing)很復雜或者數學(xue)模(mo)(mo)型(xing)無法建(jian)立，控(kong)制(zhi)系統(tong)就較難實現。蓄(xu)(xu)電池充電正是屬(shu)于這種情況，由于蓄(xu)(xu)電池的(de)充電過程有自己(ji)獨特的(de)物理化學(xue)規律，因此考慮(lv)采用(yong)模(mo)(mo)糊控(kong)制(zhi)技(ji)術來進行蓄(xu)(xu)電池的(de)充電控(kong)制(zhi) [3]。

   模(mo)糊控制[4]是(shi)(shi)以模(mo)糊集合(he)理(li)論(lun)為基礎的控制手段(duan)，它是(shi)(shi)模(mo)糊系統理(li)論(lun)、模(mo)糊技(ji)術與(yu)自(zi)動控制技(ji)術相結合(he)的產物,出發點(dian)是(shi)(shi)操作(zuo)人(ren)員(yuan)的控制經(jing)驗或相關專家的知(zhi)識，在設計中不(bu)需要建立被控對象的精確(que)數(shu)學模(mo)型[5]，因而使得控制機理(li)和策略易于(yu)接受與(yu)理(li)解，設計簡單(dan)，便(bian)于(yu)應用。

2.模糊(hu)控(kong)制(zhi)器的(de)結(jie)構及算法

   基(ji)于(yu)微控(kong)制器(qi)組(zu)成(cheng)的(de)模(mo)糊(hu)控(kong)制器(qi)包括(kuo)模(mo)糊(hu)化(hua)接口(kou)、決策(ce)邏輯、知識庫和反模(mo)糊(hu)化(hua)接口(kou)四個(ge)部分，如(ru)圖(tu)1所(suo)示。

圖(tu)1模糊(hu)控制器的結構框圖(tu)

Fig.1 The structure diagram of fuzzy controller

   在進行(xing)模(mo)糊控(kong)(kong)制(zhi)(zhi)算法設計(ji)時，須(xu)首先考慮(lv)智能充(chong)電(dian)(dian)系(xi)統的(de)技術要求(qiu)，體現智能充(chong)電(dian)(dian)控(kong)(kong)制(zhi)(zhi)技術的(de)優勢，解決長期(qi)困(kun)擾蓄(xu)電(dian)(dian)池裝備中的(de)效(xiao)率(lv)和壽(shou)命問題，所以應(ying)當提高(gao)智能充(chong)電(dian)(dian)控(kong)(kong)制(zhi)(zhi)中對(dui)控(kong)(kong)制(zhi)(zhi)量的(de)精度(du)；其次(ci)，模(mo)糊控(kong)(kong)制(zhi)(zhi)算法必(bi)須(xu)高(gao)速可靠，對(dui)外(wai)部檢測得到的(de)物理量要有(you)非常快的(de)反(fan)應(ying)速度(du)。采(cai)用新(xin)型大存(cun)(cun)儲容(rong)量的(de)微控(kong)(kong)制(zhi)(zhi)器解決使用查表(biao)法所帶來的(de)需要大容(rong)量內(nei)存(cun)(cun)的(de)問題。

3 模糊控制器的(de)設計(ji)

3.1雙輸(shu)入單輸(shu)出模糊控制器

    常見的模(mo)糊控(kong)制器[6]根據輸(shu)(shu)入(ru)(ru)(ru)與輸(shu)(shu)出(chu)(chu)(chu)個數(shu)分為單輸(shu)(shu)入(ru)(ru)(ru)單輸(shu)(shu)出(chu)(chu)(chu)、雙(shuang)輸(shu)(shu)入(ru)(ru)(ru)單輸(shu)(shu)出(chu)(chu)(chu)、多輸(shu)(shu)入(ru)(ru)(ru)單輸(shu)(shu)出(chu)(chu)(chu)和雙(shuang)輸(shu)(shu)入(ru)(ru)(ru)多輸(shu)(shu)出(chu)(chu)(chu)等幾種。平時應用較(jiao)多的是雙(shuang)輸(shu)(shu)入(ru)(ru)(ru)單輸(shu)(shu)出(chu)(chu)(chu)模(mo)糊控(kong)制器。

圖(tu)(tu)2 雙(shuang)輸(shu)入單輸(shu)出模糊控制(zhi)器方框(kuang)圖(tu)(tu)

 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)單輸(shu)出模(mo)糊(hu)(hu)控(kong)制(zhi)(zhi)器的(de)方(fang)框圖(tu)。其中屬(shu)于(yu)(yu)論域(yu)(yu)X的(de)模(mo)糊(hu)(hu)集(ji)合 取自(zi)系(xi)統誤(wu)差(cha)e的(de)模(mo)糊(hu)(hu)化，屬(shu)于(yu)(yu)論域(yu)(yu)Y的(de)模(mo)糊(hu)(hu)集(ji)合 取自(zi)系(xi)統誤(wu)差(cha)變化率(lv) 的(de)模(mo)糊(hu)(hu)化，二者構成(cheng)模(mo)糊(hu)(hu)控(kong)制(zhi)(zhi)器的(de)二維輸(shu)入(ru)；屬(shu)于(yu)(yu)論域(yu)(yu)Z的(de)模(mo)糊(hu)(hu)集(ji)合 是反映控(kong)制(zhi)(zhi)量變化的(de)模(mo)糊(hu)(hu)控(kong)制(zhi)(zhi)器的(de)一(yi)維輸(shu)出，模(mo)糊(hu)(hu)控(kong)制(zhi)(zhi)器的(de)控(kong)制(zhi)(zhi)規則(ze)[7]通常由模(mo)糊(hu)(hu)條件語句if    and    then  來表達。

3.2精確量的模(mo)糊化

   模糊(hu)(hu)控(kong)制系(xi)統中的(de)被控(kong)對象狀態(tai)變化(hua)是(shi)(shi)連(lian)續的(de)，系(xi)統給定(ding)值也是(shi)(shi)連(lian)續的(de),反映到模糊(hu)(hu)控(kong)制器(qi)輸(shu)入(ru)(ru)端的(de)輸(shu)入(ru)(ru)量也必然是(shi)(shi)連(lian)續的(de)。但模糊(hu)(hu)控(kong)制器(qi)由計算(suan)機構(gou)成，它只(zhi)能執行離(li)(li)(li)散處理[7]。因此(ci)模糊(hu)(hu)控(kong)制器(qi)要求輸(shu)入(ru)(ru)量是(shi)(shi)離(li)(li)(li)散模糊(hu)(hu)量，即論(lun)域是(shi)(shi)離(li)(li)(li)散的(de)。對連(lian)續論(lun)域要進行離(li)(li)(li)散化(hua)。連(lian)續論(lun)域[8]經過量化(hua)后(hou)就成為一個離(li)(li)(li)散論(lun)域[8]。

設有連(lian)續(xu)論(lun)域[a，b]，而量化之后的離散論(lun)域為(wei) ，亦即(ji)將連(lian)續(xu)論(lun)域分為(wei)2n段，則存(cun)在系數

                                                                   （1）

   隨后，在求(qiu)出每條(tiao)規則(ze)的(de)強(qiang)度(du)之后，把相互矛盾(dun)的(de)規則(ze)中(zhong)強(qiang)度(du)較小的(de)舍去；把相同規則(ze)合(he)成(cheng)一條(tiao)規則(ze)，得到最后控制規則(ze)基。

3.3.2根據學習算法(fa)生成控制規則(ze)

   對被(bei)控(kong)(kong)對象(xiang)執行手動控(kong)(kong)制(zhi)所得到的(de)(de)控(kong)(kong)制(zhi)規(gui)則(ze)是較粗糙(cao)的(de)(de)，有時還可能會(hui)出現(xian)控(kong)(kong)制(zhi)死(si)區，一個控(kong)(kong)制(zhi)規(gui)則(ze)表中會(hui)出現(xian)空項,這是不能滿(man)(man)足實際控(kong)(kong)制(zhi)要求的(de)(de)。為了取得更滿(man)(man)意(yi)的(de)(de)控(kong)(kong)制(zhi)效(xiao)果，可以(yi)對原始的(de)(de)控(kong)(kong)制(zhi)規(gui)則(ze)進行改進。這時，應(ying)以(yi)粗糙(cao)的(de)(de)控(kong)(kong)制(zhi)規(gui)則(ze)為基礎(chu)，通過仿(fang)真實驗(yan)和(he)系統調(diao)試加以(yi)完善(shan)。

4.模糊智能充電(dian)系統的工(gong)作(zuo)原(yuan)理(li)及結構

   智能充電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)系統主要由充電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)源(yuan)和單(dan)片(pian)機控(kong)制(zhi)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)兩部分(fen)組成(cheng)。220V的(de)(de)(de)(de)交流(liu)(liu)市電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)經(jing)整流(liu)(liu)濾波(bo)(bo)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)變(bian)為(wei)(wei)脈動的(de)(de)(de)(de)310V高壓(ya)(ya)直(zhi)流(liu)(liu)。然后經(jing)DC-DC變(bian)換電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)（脈沖功(gong)率變(bian)壓(ya)(ya)器）變(bian)為(wei)(wei)充電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)所需的(de)(de)(de)(de)60V直(zhi)流(liu)(liu)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)。為(wei)(wei)了(le)保(bao)證輸(shu)出電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)的(de)(de)(de)(de)穩定性，采用了(le)UC3842對(dui)60V直(zhi)流(liu)(liu)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)進行穩壓(ya)(ya)。二(er)(er)次斬(zhan)(zhan)波(bo)(bo)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)主要由MOSFET管、電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)感(gan)、電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)容和二(er)(er)極管組成(cheng)，輸(shu)出24-36V的(de)(de)(de)(de)充電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)。控(kong)制(zhi)部分(fen)采用C504單(dan)片(pian)機，通過對(dui)蓄(xu)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)池端電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)信(xin)號的(de)(de)(de)(de)采集、分(fen)析處理、模(mo)糊推理[8]、模(mo)糊決策等，控(kong)制(zhi)二(er)(er)次斬(zhan)(zhan)波(bo)(bo)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)路(lu)(lu)中的(de)(de)(de)(de)MOSFET管的(de)(de)(de)(de)通斷時間來控(kong)制(zhi)充電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)。控(kong)制(zhi)部分(fen)還包(bao)括對(dui)電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)流(liu)(liu)和溫度的(de)(de)(de)(de)采集以及電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)壓(ya)(ya)和電(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)(dian)流(liu)(liu)的(de)(de)(de)(de)顯(xian)示。總(zong)體結構如圖(tu)3所示。

圖3 智能充電系統總體結構框圖

5結論

  蓄(xu)電(dian)(dian)(dian)池的(de)(de)(de)(de)充放(fang)(fang)電(dian)(dian)(dian)過程是一個復(fu)雜的(de)(de)(de)(de)過程，要用精確數學模型對蓄(xu)電(dian)(dian)(dian)池充電(dian)(dian)(dian)的(de)(de)(de)(de)控(kong)制(zhi)則有(you)相當的(de)(de)(de)(de)難(nan)度。蓄(xu)電(dian)(dian)(dian)池的(de)(de)(de)(de)充電(dian)(dian)(dian)控(kong)制(zhi)系統是個非線性(xing)的(de)(de)(de)(de)、時變的(de)(de)(de)(de)、有(you)干擾的(de)(de)(de)(de)、具有(you)純滯后的(de)(de)(de)(de)控(kong)制(zhi)系統，在充放(fang)(fang)電(dian)(dian)(dian)過程中涉及到很多(duo)參數，如充電(dian)(dian)(dian)率、最大允(yun)許(xu)充電(dian)(dian)(dian)電(dian)(dian)(dian)流、內阻、出氣點(dian)電(dian)(dian)(dian)壓、溫(wen)度、壽命等(deng)。

   作(zuo)者(zhe)創新點為：

(1) 隸屬函(han)(han)數 的(de)形狀，對控制效果影響較(jiao)(jiao)大(da)。窄型隸屬函(han)(han)數，反映模(mo)(mo)糊(hu)(hu)集(ji)(ji)合具有高(gao)分辨(bian)特性。如果系(xi)(xi)統誤(wu)(wu)差(cha)，采用高(gao)分辨(bian)率(lv)模(mo)(mo)糊(hu)(hu)集(ji)(ji)合，則誤(wu)(wu)差(cha)控制的(de)靈敏度(du)就會提高(gao)。在(zai)系(xi)(xi)統誤(wu)(wu)差(cha)較(jiao)(jiao)大(da)的(de)范(fan)圍內，采用具有低分辨(bian)率(lv)隸屬函(han)(han)數的(de)模(mo)(mo)糊(hu)(hu)集(ji)(ji)合；而在(zai)系(xi)(xi)統誤(wu)(wu)差(cha)較(jiao)(jiao)小，或接近于零(ling)時(shi)，宜(yi)采用具有高(gao)分辨(bian)率(lv)隸屬函(han)(han)數的(de)模(mo)(mo)糊(hu)(hu)集(ji)(ji)合。

(2)在定義(yi)某一(yi)語(yu)(yu)言變量，如(ru)誤差(cha)、誤差(cha)變化率(lv)和控制量變化的(de)(de)全部集合時，要考慮(lv)其對(dui)論域[-n,+n]的(de)(de)覆蓋(gai)程度，語(yu)(yu)言變量的(de)(de)全部模(mo)糊(hu)集合所(suo)包(bao)含(han)的(de)(de)非零隸屬度對(dui)應的(de)(de)論域元素個數，應是模(mo)糊(hu)集合總(zong)數的(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。