201238232 六、發明說明： 【發明所屬之技術領域】 本發明係關於-種修正型模糊？1控制器之馬達驅動 ，，特別是-種可動_整更新因子論域之切換式磁阻“$ 二達之轉速及轉矩漣波’以提升馬達之控制性二 【先前技術】 ^歡錢展，使得傳統工㈣轉剌面 =技自動化生產邁進。隨著自動化生產設備需求日益增加=201238232 VI. Description of the invention: [Technical field to which the invention pertains] The present invention relates to a modified type of blurring? 1 controller motor drive, in particular - a kind of movable _ whole update factor domain switching reluctance "$ two speed and torque chopping" to improve the controllability of the motor [previous technology] ^ happy money Exhibition, making traditional workers (four) turn to the surface = technical automation production. With the increasing demand for automated production equipment =
鍵性角色—因其為生產設備的主要驅動U ^來，σΑ—方面努力研發新型的電動機結構，以期得 機械能轉換；另—方面乃就現有的電動機結構研究出 „入Γΐ财法與設計蚊適合鶴電路，崎電域的性能可 付合自動化生產設備之伺服控制需求。 此了 切換式磁阻馬達有許多的伽，如轉子無繞組使得結 且堅固耐用’可操作在高溫或具混塵的惡劣環境中 二 因使用非永磁式材觸以取縣g，價格低廉，具高轉舉二 二ί維護，具容忍故障強健性與可靠度，廣速度範圍下具 效率、以及沒有滑差的問題等。但因為雙凸極 = =導致馬達在激磁運轉時會產生較高的轉矩漣波 ^日’因此應用的範圍受到限制，通常應用在需^ ^但精ΐ要求不甚高的場合。但隨著電力電子、㈣控制方法㊉ ，、南速運算與強Α計算能力之微控㈣的。磁 ，達所遭遇的_，逐漸有專家學者提出改善 式磁阻馬達_可被用於如値驅動、電動車推進、與 啟動發電鮮各種需高性能操控的應用巾。^驅動 需具· ___、、最小_魏^== 201238232 和低速不振盪、以及高強健性等；但由於其磁路特性之高度非線 性與為得高效率運轉之高磁飽和區雜，以及磁賴、轉矩、轉 子，度間之高度耦合關係，使得要獲得合適的SRM操控特性變得报 不容易。 ,到目前為止已證明比例、積分、微分(PID)控制器於許多的工 f驅動應財林錯的難性能—基於受控體(plant)之精確數 二模型與系統操作人員之實際經驗，PI、PD、或PID控制器藉由適 二调，、,Ki、Kd三個主要控制參數，可有效補償驅動系統之變 動，但是衫統操作條件變更或是控制參數漂移，比例、積分、 =控财統之性能可能會嚴重的變差，而要_原本優異控制 f p’nTf垂重新調整KP、Kl、_制參數，但對不同的受控體而 是相#糾且乏味的工作。為了要達到所要求 碹:批㈣工風此’傳統的PID控制之控制法則必須依賴所建立的精 立i工!#受控11之正確可分析模型不確定或很難建 控制技術如模糊控制11、類神經網路、或基因演算 ♦續二或6何達成較麵控制性能。智*型控制法則會 二二ϋϊ人類專家經驗，以獲得受控體之最佳控制性能。 制二則棘拖t控制(fuzzy logic。。恤。1’ FLC)為可將語意控 為之最有力嫩之—，㈣Lc已被認 = 非線性且無法精準或不容易建立其正確數學 相較於傳統的 由於模糊控制架 fix乃設計—因為它具適應性與有效性，因此 =應_ ’附減少晴狀硬體與/貫際研九 啟發式規縣中。易將人類專家之實際經驗合併顺制器之 參、圖1…曰不-習知自調適模糊邏輯控制器⑽之基本 201238232 架構，其具有一第一正規化單元110、一第二正規化單元12〇、— 模糊化單元130、一推論引擎140、一解模糊化單元150、一規則庫 160(其包含一第一規則庫丨6丨及一α規則庫162)、一資料庫no、以 及一輸出增益單元180。 第一正規化單元110係用以使轉速誤差e正規化為如（= Gee) ’其中e(k) = r(k)-y(k) ’ r與y分別為參考命令和受控體輸出， k代表目則時間點，Ge為第一比例因子。 第二正規化單元120係用以使轉速誤差的變化量Ae正規化為 △eN (= G〜Ae)，其中GAe為第二比例因子，而△e(k) = e(k) _ e(k-1) =y(k-1) - y(k)，當r(k) = r(k-1)時，其中 k代表目前時間點，k-1代表前一個時間點。 模糊化單元130、推論引擎140、解模糊化單元150、規則庫 160、及資料庫170係用以依eN及Λθν產生增益更新因子正規化 的輸出量Διιμ。 ' 曰輸出增益單元180係用以依α及ΛυΝ產生控制信號ui變化增 量Au = (aG^u)AuN，其中為一輸出比例因子，且u(k) _ u(k-1)+Au(k)。 ~ 相較於傳統模糊類PI控制器，自調適模糊邏輯控制器1〇〇多了 該a規則庫162。模糊控制器具適應性之要求是必須的，有三種常 用的方法可使模糊控制器具適應性’即輸人或輸出比例因子 (scaling — SF)調變、歸屬函數(MF)重定義或位移、和控 制規則修正。-自觸模糊邏輯控制!!可藉由變更其歸屬函數二 比例因子、或兩者，來微調一良好工作之控制器。 自調適模糊邏輯控制器⑽其輸人與ς出變數歸屬函數之論 域皆被定義於正規化定義域(normalizeddomain)[内 所示，語意值NB、_、NS、ZE、PS、PM、PB分別代表口、 負的令、負的小、零、正的大、正的中、正的大。另—方面，增 201238232 );論域被正規化於_， 代表零、报小、小、;' S、SB、MB、B、VB分別 端兩個模糊集人使用描n垂、大、很大。其中除了最外 與相鄰函數5幢疊的屬角^ ’其他歸屬函數選用等基底並 單元化單元110、第二正規化單元120、及輸出增益 適模糊1 =的^色_於傳統控制器之增益單元，因而對自調 湘⑽之穩定度和操控性能具有決定性的影塑 二控制器丨_兩個正規化輸入(βΝ，△〜)和‘ 糊f紋義於正規化錢域卜1, 1]内。在傳統的模 :制益中，正規化的輸出量(編)係藉由輸出比例因子 映射觸龍的實際輸出~賴，細在自調賴 ==器⑽中，實際的輸出量是來自於有效的比例因城 到’/亦即调整比例因子可以調變控制變數所對應 的娜域—此作法等效於調控傳統P|D控制器之Κρ、κ(、〜三參數。 „然而，如果Gau已選定，只需要調變口值就可達到調整整個控 ，器增益之目的。合適的Ge、‘、和Gau數值可來自於專家經驗 或對受控體相關知識之認知推導而來，或是透過試誤法(计^ _ error)不斷嘗試調整到最佳可接受操作性能而得到。用來計算控 制器輸出Δυ和增益更新因子〇的規則庫如圖4和圖5所示。這^言^ 計於二維相位平面(細eplane)上最常使用之規則庫，圖4和圖5 内之控制規則是基於受控體之步級響應（stepresp〇nse)特性而建 立的，例如若輸出響應往下遠離命令值，則希望有一大的控制信 號來調整輸出朝向命令值；而當輸出響應朝命令值靠近且接近穩 態值，則需要一小的控制信號以縮短穩態時間和減少穩態誤差f △u和α的控制規則可描述為 FU;若e為Ε且Ae為ΑΕ，則Au為Λυ。Key role - because it is the main drive U ^ of the production equipment, σ Α - efforts to develop a new motor structure, in order to achieve mechanical energy conversion; another aspect is to study the existing motor structure Mosquito is suitable for crane circuit, and the performance of Saki Electric can meet the servo control requirements of automated production equipment. This switched reluctance motor has many gamma, such as rotor without winding, which makes the junction strong and durable. It can be operated at high temperature or mixed. In the harsh environment of dust, the use of non-permanent magnets touches the county g, the price is low, the maintenance is high, the maintenance is robust, the reliability is high, the speed is high, and the speed is not smooth. Poor problems, etc. But because the double salient pole = = causes the motor to produce a higher torque ripple during the excitation operation, so the scope of application is limited, usually applied in the need of ^ ^ but the precision requirements are not very high Occasionally, but with the power electronics, (four) control method ten, the south speed calculation and the powerful control of the micro-control (four). Magnetic, up to the encounter _, gradually experts and scholars proposed improved reluctance horse _ can be used for such as 値 drive, electric vehicle propulsion, and start-up power generation of various applications requiring high-performance control. ^Drive requirements · ___,, minimum _Wei ^== 201238232 and low speed non-oscillation, and high robustness Etc.; however, due to the high nonlinearity of its magnetic circuit characteristics and the high magnetic saturation region for high efficiency operation, and the high coupling relationship between magnetic dependence, torque, rotor and degree, it is necessary to obtain suitable SRM control characteristics. It is not easy to report. So far, it has been proved that the proportional, integral, and differential (PID) controllers are difficult to perform in many work-failures based on the precise number two model and system operation of the controlled plant. The actual experience of personnel, PI, PD, or PID controller can effectively compensate for the changes of the drive system by appropriately adjusting the two main control parameters, Ki, Kd, but the operating conditions of the system change or the control parameters drift. The performance of the ratio, integral, and control system may be seriously deteriorated. However, the original excellent control f p'nTf re-adjusts the parameters of KP, Kl, and _, but for different controlled bodies, And boring work. In order to To the required 碹: batch (four) work wind this 'traditional PID control law must rely on the established fine work! #控制11's correct analyzable model is uncertain or difficult to build control techniques such as fuzzy control 11, Neural networks, or genetic algorithms ♦ Continued 2 or 6 to achieve better control performance. The intelligent control method will have two or two human expert experience to obtain the best control performance of the controlled body. Control (fuzzy logic. 1 'FLC) is the most powerful thing to control the language - (4) Lc has been recognized = nonlinear and can not be accurate or not easy to establish its correct mathematics compared to the traditional due to fuzzy control The frame fix is designed—because it is adaptable and effective, so = should be _ 'attached to reduce the fine hardware and / cross-section of the nine heuristics. It is easy to combine the practical experience of human experts with the parameters of the controller. Figure 1... The basic 201238232 architecture of the self-adaptive fuzzy logic controller (10) has a first normalization unit 110 and a second normalization unit. 12〇, a fuzzification unit 130, an inference engine 140, a defuzzification unit 150, a rule base 160 (which includes a first rule base 丨6丨 and an alpha rule base 162), a database no, and An output gain unit 180. The first normalization unit 110 is configured to normalize the rotational speed error e as (= Gee) 'where e(k) = r(k) - y(k) ' r and y are reference commands and controlled body outputs, respectively , k represents the time point of the target, and Ge is the first scale factor. The second normalization unit 120 is configured to normalize the amount of change Ae of the rotational speed error to ΔeN (= G 〜 Ae), where GAe is the second scale factor and Δe(k) = e(k) _ e( K-1) = y(k-1) - y(k), when r(k) = r(k-1), where k represents the current time point and k-1 represents the previous time point. The blurring unit 130, the inference engine 140, the defuzzification unit 150, the rule base 160, and the database 170 are configured to generate an output amount Διιμ normalized by the gain update factor according to eN and Λθν. The 曰 output gain unit 180 is used to generate a control signal ui variation increment Au = (aG^u)AuN according to α and ,, where is an output scale factor, and u(k) _ u(k-1)+Au (k). ~ Compared to the traditional fuzzy PI controller, the self-adaptive fuzzy logic controller 1 has more than the a rule base 162. The fuzzy controller has the requirement of adaptability. There are three commonly used methods to make the fuzzy controller adaptable' ie input or output scaling factor (scaling-SF) modulation, attribution function (MF) redefinition or displacement, and Control rule correction. - Self-touch fuzzy logic control!! Fine-tuning a good working controller by changing its attribution function two scale factor, or both. The self-adaptive fuzzy logic controller (10) has its domain of input and output variable attribution functions defined in the normalized domain [indicated, semantic values NB, _, NS, ZE, PS, PM, PB). It is the big, the medium, the big, the big, the big, the big, the big, the big. On the other hand, increase 201238232); the domain is normalized to _, representing zero, small, small,; 'S, SB, MB, B, VB respectively, two fuzzy sets of people use the description n vertical, large, very Big. In addition to the outermost and adjacent function 5 stacks of the corners ^ 'other attribution functions select the base and unitization unit 110, the second normalization unit 120, and the output gain is suitable for blur 1 = ^ color _ traditional controller The gain unit is thus decisive for the stability and handling performance of the self-tuning (10). Two normalized inputs (βΝ, △~) and 'paste f' are defined in the normalized money domain. , 1] inside. In the traditional mode: the benefits, the normalized output (edit) is based on the output scale factor mapping the actual output of the contact dragon ~ lie, fine in the self-adjusting == device (10), the actual output is from The effective ratio of the city to '/, that is, the adjustment of the scale factor can modulate the Naval domain corresponding to the control variable—this method is equivalent to the regulation of the traditional P|D controller Κρ, κ (, ~ three parameters. „ However, if Gau has been selected, and only needs to adjust the mouth value to achieve the purpose of adjusting the whole control and gain. The appropriate Ge, ', and Gau values can be derived from expert experience or cognitive derivation of the knowledge of the controlled body, or It is obtained by continuously trying to adjust to the best acceptable operating performance through trial and error (calculation _ error). The rule base used to calculate the controller output Δυ and the gain update factor 如图 is shown in Fig. 4 and Fig. 5. The rule rule is the most commonly used rule base on the two-dimensional phase plane (fine eplane). The control rules in Figure 4 and Figure 5 are based on the stepresp〇nse characteristics of the controlled body, for example. Output response down from the command value, I hope there is a big control The number adjusts the output toward the command value; and when the output response approaches the command value and approaches the steady state value, a control signal that requires a small control signal to shorten the steady state time and reduce the steady state error f Δu and α can be described as FU; if e is Ε and Ae is ΑΕ, then Au is Λυ.
Ra·右e為Ε且為，則α為殳。 201238232 數均ilLr適f糊邏輯控制器100其輸入變數e和^的歸屬函 _則庫各二值二=二因右此用來推_ 大多㈣τΛ條 (財98條㈣朗）。然而，在絕 和计篡制法必須實現於具有限記憶體空間 控㈣(MCUW，在降低成本財量下，控 p或MCU的性能是很大的挑戰。 換目的在於提出—種可動態調整更新因子論域之切 換式磁阻馬達控㈣，其增益更新因子之論域依轉速改變。 切拖2明之另—目的在於提出—種可動態調整更新因子論域之 切換式磁阻馬達控㈣，其增益更新因子具有3個語意值。 明之X-目的在於提Λ —種可麟調整更_子論域之 2式磁’達控㈣’其可降簡需之難酬數以簡化整個 控制器之複雜度及減輕微控制器之運算負荷。 /、，、達成上述之目的，一可動態調整更新因子論域之切換式磁 阻馬達控制ϋ乃被提出’其係以_轉賴差和—騎誤差變化量 2輸入賊電流命令變化量讀出魏，該控織具有:一 規化單元’侧以使轉速誤差正規化為—正規化轉速誤 差，一第二正規化單元，係用以使該轉速誤差變化量正規化為一 正規化轉速誤差變化量；—轉速適應控制參數產生單S，係用以 依该正規化轉速誤差及該正規化轉速誤差變化量產生一更新因子 及y正規化的電流命令變化量，其巾該更新因子之論域係依轉速 改變一隨轉速增加/降低而左移/右移；以及一輸出增益單元，係用 贿該更新因子與-輸出比例因子之乘積產生—增益，及依該增 盈放大该正規化的電流命令變化量，以產生該電流命令變化量。 οπ本發明控制策略的實現是藉由所提出的修正型模糊類ΡΙ控制 為來滿足變觀命令的要求。在速度控制器方面，本發明將模糊 修正因子與模糊汧控制器的規則庫結合與簡化，並導入模糊速度 7 201238232 二斋内’改善原本卿庫的複雜纽其在鶴時誤差過大的問 並，有即時自我調適之能力。模糊控制器之控制機制，提供 服系統非線性特性與降低轉矩漣波的能力，最後利用數位訊 =理器來實現所提出之控娜略，經由不同操作條件(速度追縱 瞬間負載’憂動）下之動態測試的響應結果，來驗證所提方法之可 行性與操作性能。 為使貝審查委員能進一步瞭解本發明之結構、特徵及其目 的，兹附以圖式及較佳具體實關之詳細說明如后。 【實施方式】 為了在低價且有限處理能力之MCU實現所用的控制法則，本 ^明先將重職在α酬㈣之麵控概職減少，然後進一步 簡化所設計的控㈣之記，隨需求和計算複雜度。 睛參照圖6，其繪示包含本發明可動態調整更新因子論域之切 換式磁阻馬達控制器600一較佳實施例之方塊圖。如圖6所示，可 動態調整更新因子論域之她式雜馬雜㈣_，哺速誤差 Werr和轉速誤差變化量為輸入變數及以電流命令變化量△丨* 為輸出變^具有-第—正規化單元、—第二正規化單元卿、 -模糊化單tl63G、-推論引擎64〇、一解模糊化單元咖、一規則 庫660(其包含—第一規則庫661及一 α「簡 、 ⑽、以及-輸出增益單元680。 貝料庫 第一正規化單元610係用以使轉速誤差werr正規化為⑴沉別（= ewerr) ’其中Werr(k) = w*(k) _ Wr(k) ’ ω*與吣分別為參考命令和受 控體輸出’ k代表目前時間點，Ge為第一比例因子。 第二^規化單元620係用以使轉速誤差變化量^⑴时正規化為 △⑴errN (= GMAWerr)，其中〇Ae為第二比例因子，而=Ra·right e is Ε and is, then α is 殳. 201238232 Number average ilLr suitable for the paste logic controller 100 its input variables e and ^ attribution letter _ then the library each two values two = two due to this right to push _ most (four) τ Λ (Cai 98 (four) lang). However, in the absolute and fixed method, it must be realized in the limited memory space control (4) (MCUW, under the cost reduction, controlling the performance of p or MCU is a big challenge. The purpose of the change is to propose a kind of dynamic adjustment The switched-type reluctance motor control of the update factor domain (4), the domain of the gain update factor changes according to the rotational speed. The other is to propose a switchable reluctance motor control that can dynamically adjust the update factor domain (4) The gain update factor has three semantic values. The X-purpose is to improve the 可 种 种 调整 调整 调整 调整 _ _ _ 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 子 磁 磁The complexity of the device and the reduction of the computational load of the microcontroller. /,,, to achieve the above purpose, a switchable reluctance motor control that can dynamically adjust the update factor domain is proposed to be - riding error variation 2 input thief current command change amount read Wei, the control weaving has: a regulatory unit 'side to normalize the rotational speed error to - normalized rotational speed error, a second normalization unit, is used Make the speed error change The normalization is a normalized rotational speed error variation; the rotational speed adaptive control parameter generates a single S, which is used to generate an update factor and a y normalized current command change according to the normalized rotational speed error and the normalized rotational speed error variation. The coverage factor of the update factor is changed to the left/right shift according to the rotation speed increase/decrease according to the rotation speed; and an output gain unit is generated by multiplying the update factor and the output scale factor by the bribe, and The normalized current command variation is amplified according to the gain to generate the current command variation. The implementation of the control strategy of the present invention is achieved by the proposed modified fuzzy class control to satisfy the requirement of the change command. In terms of the speed controller, the present invention combines and simplifies the fuzzy correction factor with the rule base of the fuzzy 汧 controller, and introduces the fuzzy speed 7 201238232, and improves the complexity of the original corpus in the crane. , has the ability to adapt to the moment. The control mechanism of the fuzzy controller provides the nonlinear characteristics of the service system and the ability to reduce the torque ripple. Finally The digital control is used to implement the proposed control, and the feasibility and operational performance of the proposed method are verified by the response results of dynamic tests under different operating conditions (speed tracking instantaneous load 'worry'). The structure, features and objects of the present invention can be further understood by the members of the review board, and the detailed descriptions of the drawings and the preferred embodiments are as follows. [Embodiment] For use in an MCU with low cost and limited processing capability The control law, this first Ming will reduce the number of positions in the alpha (4), and then further simplify the design of the control (4), with the needs and computational complexity. Eyes with reference to Figure 6, which includes the present invention A block diagram of a preferred embodiment of the switched reluctance motor controller 600 of the update factor domain can be dynamically adjusted. As shown in FIG. 6, the update factor domain can be dynamically adjusted to her type of mixed horse (4) _, the feeding speed error Werr And the change amount of the rotational speed error is the input variable and the current command change amount Δ丨* is the output change ^ has - the first normalization unit, the second normalization unit, the - fuzzy single tl63G, the inference engine 64 〇 A defuzzification unit coffee, a rule base 660 (which comprises - a first rule base 661 and a α 'Jane, ⑽, and - an output unit 680 gain. The first normalization unit 610 is used to normalize the rotational speed error werr to (1) sinking (= ewerr) 'where Werr(k) = w*(k) _ Wr(k) ' ω* and 吣 respectively The reference command and the controlled body output 'k' represent the current time point and Ge is the first scale factor. The second control unit 620 is configured to normalize the rotational speed error variation ^(1) to Δ(1)errN (= GMAWerr), where 〇Ae is the second scale factor, and =
Werr(k)-Werr(k-1) ； 目前時間點，間r)~ 201238232 模糊化單元630、推論引擎64〇 '解模糊化單元65〇 60 及貧料庫67〇係用以形成一轉速適應控制參數產生單元、 依⑴*、uwN及產生更新因子及正規化的輸 中％之論域係依轉速改變—隨轉速增加/降低而左移/右移。N、 量單^咖以依。「及△丨*N產生控制嫌之變化增 其中為—輸出比例因子，，= 明如^轉速聽㈣參數產生單元所·崎屬函數和規則庫說 9所屬甘函數：控制器輸出ΔΓ之歸屬函數與AU的定義相同（如圖 域範圍被正規化在卜1, 1] (A)内，輸入變數^和 的^屬，朗樣被定義在[_!，彳](_區肋。根據馬達速度響應 口匕的、乡考命令之間關係’⑴时和^⑴抓的歸屬函數可被簡化成三 類’即速度誤差（或誤差變量）為負(Ν)、零(ΖΕ)、或正(ρ)，因此選 擇此二健糊集合作為輸入變數之歸屬函數，如圖7所示。同理更 新因子α「也可被分成三類，即小(s)'中(Μ)、大⑹，基於對系統 響應的直覺分析，所推導出的控制規則如圖8所示，而且為了獲得 好的控制解析度和對命令設定點（货伽丨时，印)變動時提供較佳 的適應性’ α「的論域會根據速度命令加以調變，例如，速度命令範 圍在24QQ-3_ipm時’其的論域為[◦，◦ 6]，當速度命令每減少 600rpm ’其論域往右⑷方向位狐丨，因此速度命令範圍在 0-600rpm時’其％的論域為[〇 41 〇]，如圖9所示。基於輸出比例 因子arG〜之連續調變，控制器的輸出將隨著速度命令範圍減少而 增加’如此可以克服只用固定輸出比例因子控制所遇到的輸出響 應不易調控的問題。 規則-庫推導.用來推導控制器輸出變化增量(△「)之規則庫， 與圖4所tf相同—其具有49條控概貞彳，此處重點無設計推導更 新因子ar之規則庫’為了要減少圖5之控制規則數，有必要對圖ι〇 201238232 所示之馬達步級祕響應行為做詳細觀察，由關可看出， 之速度響應可被劃分為四個操作區RrR|v和兩組特殊點： (加讀的‘胃_，叫和峰_3,35)，類似於非線性系統之相 面分析，關之時域響應在e·/^狀態空間平面之映射如圖_ 不。當速度響應曲線通過不同操作區時，e和Ae的正、負號會隨 著改變’例如，在R—R1V區，（e^e)的符號分別為㈠和(+ : 此兩區域之響應狀態意味著目前馬達速度不僅朝上㈣或朝下 (Rlv)，離速度命令’而且正往遠離命令方向前進，在此情況下， 可動‘悲，整更新因子論域之切換式磁阻馬達控制器_應該提供 大的增益(arG〜)以防止更差速度響應發生，此 若誤差和誤差變化量均為負或均為正，咖為大⑻/了貫見為· 另一方面，當速度響應接近速度命令且正朝著命令前進，此 情況下速度曲線位於RAR·，且(e，Ae)的正、負符號是相反的， 在此時可動態調整更新因子論域之切換式磁阻馬達控制器6〇〇應 该要減少增益以避免造成太大的超越量或過低量__⑽)並 減少安定時間，此控制法則可實現為： 若誤差和誤差變化量互為異號，則ar為小（S)。 此外，虽響應接近穩定狀態，亦即e或Ae為零，也就是速度 響應剛好到達或離開設定點，且正好快速要遠離或接近設定點， 在此情況下，提供適中的增益將可避免超越量或過低量，此種在 ^度命令附近之增益_方式，也可縣造·應的紐並增加 二應到達穩態之收斂率’此控概則可實現為： 右誤差和誤差變化量有一為零，則ar為中（M)。 ，後在穩態時，可動態調整更新因子論域之切換式磁阻馬達 ，制益刪應雜錄小增益’⑽免造成在設定簡近之顫動 jhattering)問題，此情況之控制法則可實現為： 若誤^和誤差變化量均為零’則ar為小（S)。 這些控制規則的推導可用於a規則庫162和〜簡化規則庫662 201238232 的演繹’而且基於對馬達響應行為和〇規則庫162架構(圖5)的瞭解 和推論，所提出的簡化規則庫662架構如圖8所示。 增盈凋控機制（gain tuning strategy):不具比例增益調控 機制之模_P丨㈣H有—缺點，即#控被設計完成，其^ 入=輸出變數之論域範圍是固定的，當系統趨近於穩態時，固定 的定義域會導致纟、制達觀之安定時間拉長，且在預設的速度 ^近會造成振i ’這是由於當速度誤差變小時控㈣解析度ς 夠而無法適時反應所造成。為了要得到滿意的驅動性能，控制器 的論域範圍應該要根據操作點來做調整，因此我們提出透過可連 續非線性的罐修正因子來賴比綱益， 出變數論域範圍之模糊控制器，此處我們把重點放在=二3 變，因為它等效於一般控制器之增益，而且對自調適模‘ 類卩丨控制_言’輸出_增益的調整扮演著重要角色，由於 性能有減的影響。所提出的模糊控繼之自調 a·輸入與輸出比例因子變動效應：比例因子調變是最 提升模糊控制器性能的方法之一，比例因子的設計，特別 =例因子’對模糊控制器是很重要的，因為它們會左右控制器性 ^以圖1所示之自調適模糊控制器1〇〇為例，藉由比例增益將e、 =、和Διι轉換為eN、△〜、和ΔυΝ，意味著它們是由實際 =轉換到卜1，彳则，而此轉換被稱為正規化(⑽醜1丨她加）1 二比例因子的效麟效於擴展或縮小輸人與輸出變數之實^ 二下比例因子和實際論域間之關係和它們對SRM響應的效應^ ^若誤差量比先前縣值大，為了_上升_並減少解 ^ e和的值應該被調小，如此可保持系統穩定度並使控 ^平滑。另-方面’應指定較大的值給以加速魏響癖^ 右*差量比先前誤差值小，為了縮短安定時間並增加控制解析户） 11 201238232 和靈敏度，Ge#Ae的值應該被調大，在此同時，應指定較小的‘ 值以避統大的朗量和紐，也可賊速到麵定狀態。 b.自调適機制：為了要達成自調適目的，適紐模糊控制器 之輸出必須是可職的，目前仍㈣效且有祕方法可用來作模 糊控制器之調變以得到最佳的輸出響應，因為要決定可調朱數（比 =子、歸屬函數、控制酬）的最佳值，必須要有受控體的精確 數子核型之相關知識·，並且由3小節之測試結果得知，輪出比姻 子之變動對系統響應性能的影響比輸入比例因子之變動來得大， 子，變將會大大增加系統的複雜度和微控 制為之计异負何罝，因此本發明所提出的可動態調整更新因子論 $之切換式磁阻馬達控制器_僅作輸出比例因子之調變。其調變 機制如下所述： 〃 紅為K衫作增益調變），而4^、和^初始操Werr(k)-Werr(k-1) ; current time point, between r)~ 201238232 fuzzification unit 630, inference engine 64〇' defuzzification unit 65〇60 and poor material bank 67〇 are used to form a rotation speed The adaptive control parameter generation unit, the domain according to (1)*, uwN, and the generation of the update factor and the normalized input are changed according to the rotational speed—the left/right shift as the rotational speed increases/decreases. N, the amount of money ^ coffee to rely. "And △ 丨 * N produces a change in control suspicion, which is - output scale factor, = = Mingru ^ rpm listening (four) parameter generation unit, Sakae function and rule library, 9 belong to Gan function: controller output ΔΓ attribution The function is the same as the definition of AU (as shown in the figure domain is normalized in Bu 1,1) (A), the input variable ^ and ^ are genus, the Lang sample is defined in [_!, 彳] (_ zone rib. According to The relationship between the motor speed response response and the township test command '(1) and ^(1) can be simplified into three categories, ie the speed error (or error variable) is negative (Ν), zero (ΖΕ), or Positive (ρ), therefore, the two sets of the two pastes are selected as the attribution function of the input variables, as shown in Fig. 7. The same reason update factor α "can also be divided into three categories, namely small (s)' (Μ), large (6) Based on the intuitive analysis of the system response, the derived control rules are shown in Figure 8, and provide better adaptation in order to obtain good control resolution and change the command set point (printing). The domain of the 'α' will be modulated according to the speed command, for example, when the speed command range is 24QQ-3_ipm The domain is [◦, ◦ 6], when the speed command is reduced by 600 rpm, the domain is oriented to the right (4) direction, and therefore the speed command range is 0-600 rpm 'the domain of % is [〇41 〇], As shown in Figure 9. Based on the continuous modulation of the output scaling factor arG~, the output of the controller will increase as the speed command range decreases. This can overcome the difficulty of regulating the output response encountered only by the fixed output scaling factor control. Problem. The rule-library derivation. The rule base used to derive the controller output change increment (△ ") is the same as the tf in Figure 4 - it has 49 control profiles, where there is no design derivation update factor ar Rule Base 'In order to reduce the number of control rules in Figure 5, it is necessary to make a detailed observation of the motor step-level response behavior shown in Figure ι〇201238232. As can be seen, the speed response can be divided into four operating areas. RrR|v and two special points: (added 'stomach _, called and peak _3, 35'), similar to the phase analysis of nonlinear systems, the time domain response is in the e·/^ state space plane The mapping is shown in Figure _ No. When the speed response curve passes through different operating areas The positive and negative signs of e and Ae will change with 'for example, in the R-R1V area, the signs of (e^e) are (1) and (+: the response state of the two regions means that the current motor speed is not only facing upwards. (4) or downward (Rlv), away from the speed command 'and moving forward away from the command direction. In this case, the movable 'sorrow, the whole renewed factor domain of the switched reluctance motor controller _ should provide a large gain (arG ~) to prevent a worse speed response, if the error and the amount of error change are both negative or positive, the coffee is large (8) / the consistent view is · On the other hand, when the speed response is close to the speed command and is moving toward the command In this case, the speed curve is located at RAR·, and the positive and negative signs of (e, Ae) are opposite. At this time, the switched reluctance motor controller 6 that can dynamically adjust the update factor domain should be reduced. The gain can be avoided to avoid too much overshoot or too low __(10)) and reduce the settling time. This control law can be implemented as follows: If the error and the error variation are different from each other, ar is small (S). In addition, although the response is close to steady state, ie, e or Ae is zero, that is, the speed response just reaches or leaves the set point, and just needs to move away from or close to the set point quickly, in this case, providing a moderate gain will avoid exceeding The amount or the amount of the gain, the gain _ mode near the ^ degree command, can also be made by the county and should increase the convergence rate of the steady state. The control principle can be realized as: right error and error change If the quantity has zero, then ar is medium (M). After the steady state, the switching reluctance motor of the update factor domain can be dynamically adjusted, and the problem of the small gain of the miscellaneous recording can not be caused by the hysteresis (10), and the control law of this case can be realized. To: If the error and the amount of error change are both zero, then ar is small (S). The derivation of these control rules can be used for the deduction of a rule base 162 and the simplified rule base 662 201238232 and based on the understanding and inference of the motor response behavior and the rule base library 162 architecture (Fig. 5), the proposed simplified rule base 662 architecture As shown in Figure 8. Gain tuning strategy: The model with no proportional gain control mechanism _P丨(4)H has the disadvantage, that is, the #control is designed to be completed, and the range of the field of the input variable is fixed. When the steady state is near, the fixed domain will cause the stability time of the 纟 and 制 拉 to be elongated, and the vibration will be caused by the preset speed ^. This is because the speed error becomes smaller (four) resolution is enough. Unable to react in time. In order to obtain satisfactory driving performance, the controller's domain range should be adjusted according to the operating point. Therefore, we propose a fuzzy controller with a variable nonlinearity through a continuously nonlinear tank correction factor. We focus on =2 and 3, because it is equivalent to the gain of the general controller, and plays an important role in the adjustment of the self-adjusting mode 'class 卩丨 control _ 』 'output _ gain, due to performance degradation influences. The proposed fuzzy control follows the self-adjustment a· input and output scale factor variation effect: scale factor modulation is one of the methods to improve the performance of fuzzy controllers. The design of scale factor, especially = example factor 'is for fuzzy controller Very important, because they will control the controller. Take the self-adaptive fuzzy controller 1 shown in Figure 1 as an example, and convert e, =, and Διι to eN, Δ~, and ΔυΝ by proportional gain. It means that they are converted from actual = to 1, and this conversion is called normalization ((10) ugly 1 丨 she plus) 1 two scale factor effect to expand or reduce the input and output variables ^ The relationship between the scale factor and the actual domain and their effect on the SRM response ^ ^ If the error is larger than the previous county value, the value of _ _ and _ _ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ ̄ System stability and smooth control. Another-aspect' should specify a larger value to speed up Wei Xiang癖^ Right* difference is smaller than the previous error value, in order to shorten the stability time and increase the control resolution) 11 201238232 and sensitivity, the value of Ge#Ae should be adjusted Large, at the same time, you should specify a smaller 'value to avoid the large amount of Lang and New Zealand, but also thief speed to face state. b. Self-adjustment mechanism: In order to achieve self-adjustment, the output of the adaptive fuzzy controller must be workable, and still has a (four) effect and a secret method that can be used as a fuzzy controller to obtain the best output. In response, because you want to determine the optimal value of the adjustable number (ratio = child, attribution function, control fee), you must have the knowledge of the exact number of karyotypes of the controlled body, and the test result of 3 bars is obtained. It is known that the effect of the change of the rotation on the response performance of the system is greater than the change of the input scale factor. The change of the system will greatly increase the complexity of the system and the micro control will be different. Therefore, the present invention The proposed switchable reluctance motor controller that can dynamically adjust the update factor theory is only used as the modulation factor of the output scale factor. The modulation mechanism is as follows: 〃 Red for K-shirt for gain modulation), and 4^, and ^ initial operation
Ge 1/werr，max(或 | 1/〇)时_ | )，=V△⑴阶扉， ’以使正】規化的輸人、輸出變數能涵蓋正規化定義 -[，]或«[emin’emax]，以提升控制解析度和規則庫之利。 =2) Ge、GAe、和的值保持在step【之初始值，接 =新因子’在此步驟控制器之輸出和更新因子可以表示如下:Ge 1/werr, max (or | 1 / 〇) _ | ), = V △ (1) order 扉, 'to make positive' of the input, output variables can cover the formal definition - [,] or « [ Emin'emax] to improve control resolution and rule base benefits. =2) The values of Ge, GAe, and are kept at the initial value of step [connected = new factor]. The output and update factors of the controller in this step can be expressed as follows:
Au(k) = ar(k)(kAuGAU)AuN 卜. a「(k)= far(e(k)，Ae(k)) ί義於e—Ae平面上之非線性函數—其_用圖8及 射-生之比例常數。此處設定^的值= ep的好幾倍’ kAu是一個經驗值，例如 下完全追縱命令值，則‘可設定為較小的二： 時間，應該娜~調大，在所提出的控麵構 12 201238232 值以中和更新因子只能位於[αι]範圍之效應，並且保證在 的超越量下仍保有快速的輸出響應。 士 J、Au(k) = ar(k)(kAuGAU)AuN 卜. a "(k)= far(e(k), Ae(k)) 非线性 a nonlinear function on the e-Ae plane - its _ 8 and the ratio of the ratio of the shoot-to-life. Here the value of ^ is set to several times ep 'kAu is an empirical value, for example, the full tracking command value, then 'can be set to a smaller two: time, should Na~ The adjustment is made in the proposed control plane 12 201238232 value to neutralize the update factor can only be in the range of [αι], and to ensure that the output response is still fast under the overshoot.
St印3)若有需要則針對所要求的響應和特別操作 控制規則’如此各種操作情況下之適當的 值皆可由線上微調而得到。 口丁 圖13所示為包含本發明可絲調整更新因子論域之 阻馬達控制器600之SRM驅動系統方塊圖，其主要具有^ 1320、與換相邏輯判斷控制器漏）、一閘驅動器134卜其係財 搞合器隔離、一功率變頻器1350、與-SRM 1360—1為四相8/fi =模糊速度控㈣·包含有可動_整更新因子論域之切換 式阻馬達控制器咖，其係用以接受轉速誤差信號及轉速命人，、 U生四?電流命令信號’電流命令信號與由霍爾電流感測J ，測到之實際電流作比較制一電流誤差量，根據此^誤差量 =小，PI電流控制II131G產生所需之PWM_動信號，經 動益134G去驅動功率變的丨GBT開關^ 器·根據轉子角度、激磁導通角、與截止角來推導 的SRM 1360換相時機。激磁角度調變控制器酬根據所輸 ΐΓΪί差、電流、與轉子位置，藉由調控增量激磁導通 角”截止角來最小化SRM 1360轉矩漣波和能量轉換，為了簡 統的複雜度，所有控制器較佳為實現於一Dsp控制平台上。’、 ^案所揭示者，乃較佳實關，舉凡局部之變^修傅而源 於本案之技術思想而為熟習該項技藝之人所易於推知者，俱不 本案之專利權範脅。 、’’不上所陳，本案無論就目的、手段與功效，在在顯示兑週里 技術Ϊ徵，且其首先發明合於實用，巾在 專利要件，懇5青貴審查委員明察，並祈早曰賜予專利 社會，實感德便。 早嘉心 13 201238232 f圖式簡單說明】 =示二知自糊糊邏輯控制器之基本架構。 圖2、·，θ Tire、的歸屬函數。 圖3繪示增益更新因子α的歸屬函數。 圖4繪示Δυ的規則庫表格。 圖5繪示α的規則庫表格。 馬達示發明可動態調整更新因子論域之切換式磁阻 馬違控制益一較佳實施例之方塊圖。 之—較佳實施例。 函數本實=::棒24。〜 圖10繪示一馬達之步級響應。 圖11繪示圖7響應在e-^e平面之映射。 圖12繪示本發明α「簡化規則庫的架構。 示包含本發明可動態調整更新 之切換式磁阻 馬達控制器之SRM驅動系統方塊圖。 【主要元件符號說明】 自調適模糊邏輯控制器1〇〇 第—正規化單元11() 第二正規化單元120 模糊化單元13〇 推論引擎14〇 解模糊化單元15〇 規則庫160 第—規則庫161 α規則庫162 資料庫17〇 201238232 輸出增益單元180 可動態調整更新因子論域之切換式磁阻馬達控制器600 第一正規化單元610 第二正規化單元620 模糊化單元630 推論引擎640 解模糊化單元650 規則庫660 第一規則庫661 ar簡化規則庫662 資料庫670 輸出增益單元680 模糊速度控制器1300 PI電流控制器1310 激磁角度調變控制器1320 換相邏輯判斷控制器1330 閘驅動器1340 功率變頻器1350 SRM 1360 15St. 3) If necessary, the required response and special operation control rules. The appropriate values for such various operating conditions can be obtained by fine-tuning the line. FIG. 13 is a block diagram of an SRM driving system including a resistance motor controller 600 of the present invention, which mainly has a ^1320, and a commutation logic determination controller leak), a gate driver 134. Buqicai Caiqihe isolation, one power inverter 1350, and -SRM 1360-1 are four-phase 8/fi = fuzzy speed control (4) · Switched resistance motor controller with movable _ whole update factor domain , which is used to receive the speed error signal and the speed of the person, the U-four current command signal 'current command signal and the current measured by the Hall current sense J, the actual current measured to make a current error amount, according to this ^ Error amount = small, PI current control II131G generates the required PWM_motion signal, 动GBT switch that drives power change via 134G. ·SRM 1360 based on rotor angle, excitation conduction angle, and cutoff angle Commutation timing. The excitation angle modulation controller minimizes the SRM 1360 torque ripple and energy conversion by adjusting the incremental excitation conduction angle "offset angle" according to the input 差 difference, current, and rotor position. For simple complexity, All controllers are preferably implemented on a Dsp control platform. ', the case disclosed by the case is a better practice, and it is easy for those who are familiar with the skill from the technical idea of the case. The inferred person, the patent rights of the case are not threatened. ''Not on the Chen, the case, regardless of the purpose, means and efficacy, in the display of the technical levy in the week, and its first invention is practical, the towel in the patent The essentials, 恳5 Qinggui review committee clearly, and prayed to the patent society, the real sense of virtue. Early Jiaxin 13 201238232 f simple description] = the basic structure of the two-in-one logic controller. Figure 2 Fig. 3 shows the attribution function of the gain update factor α. Fig. 4 shows the rule base table of Δυ. Fig. 5 shows the rule base table of α. The motor shows that the invention can dynamically adjust the update factor domain. It A modified block diagram of a preferred embodiment of the present invention. A preferred embodiment. Function: =:: Rod 24. ~ Figure 10 illustrates the step response of a motor. 7 is a mapping of the response in the e-^e plane. Figure 12 is a block diagram showing the structure of the "simplified rule base" of the present invention. The block diagram of the SRM drive system including the switchable reluctance motor controller of the present invention which can be dynamically adjusted and updated is shown. Element Symbol Description] Self-adaptive fuzzy logic controller 1 〇〇 first - normalization unit 11 () second normalization unit 120 fuzzy unit 13 〇 inference engine 14 模糊 fuzzy unit 15 〇 rule library 160 - rule library 161 Alpha rule base 162 database 17〇201238232 output gain unit 180 switchable reluctance motor controller 600 that can dynamically adjust the update factor domain. First normalization unit 610 second normalization unit 620 fuzzy unit 630 inference engine 640 deblurring Unit 650 rule base 660 first rule base 661 ar simplified rule base 662 database 670 output gain unit 680 fuzzy speed controller 1300 PI current controller 1310 excitation angle modulation controller 1320 commutation logic judgment Controller 1330 Gate Driver 1340 Power Inverter 1350 SRM 1360 15