201216594 六、發明說明： 【發明所屬之技術領域】 本發明係關於一種永磁同步風力發電系統，詳言之，係 關於一種利用智慧型最大功率追縱器之永磁同步風力發電 系統。 【先前技術】 習知風力發電系統可利用功率電子轉換器以定速或可變 速度操作。由於可變速發電可以在所有風速下達到最大效 率，以增進輸出能量及減少電壓閃爍問題，故可變速發電 常為業界使用。許多發電機之研究及實務上風力發電機為 具有繞組式轉子或鼠籠式轉子之感應機。最近，永磁同步 發電機之應用逐漸增加。具有高效率及高可控性之高功能 及可變速發電可利用永磁同步發電機達成。習知的研究專 /主在二種最大風力控制方法’其為：尖端速度比控制（tip-speed ratio， TSR) 、 功率 信號迴 授控制 （p〇wer signai feedback，PSF)及爬坡法控制（hill-ciimb searching，HCS)。 參考圖1 ’其顯示尖端速度比與功率係數之關係示意 圖。尖端速度比控制在於控制風力機轉子速度以保持一最 佳尖端速度比》功率信號迴授需要知道風力機的最大功率 曲線及經由其控制機制追縱其曲線。在習知的風力發電最 大功率點追蹤策略中，尖端速度比控制方法因難以取得風 速及風力機速度’故其使用受到限制β許多習知的風力發 電最大功率點追蹤策略係藉由使用風力機最大功率曲線以 減少測量’但仍須知道風力機的特性。習知爬坡法控制係 149313.doc 201216594 連續地搜尋風力機的尖端輸出功率。比較上，因簡單性及 系統特性的獨立性，爬坡法控制之風力發電最大功率點追 蹤方法較受歡迎。 因此，有必要提供-種創新且具進步性的利用智慧型最 大功率追蹤器之永磁同步風力發電系統，以解決上述問 題。 。 【發明内容】 本發明提供-種利用智慧型最大功率追蹤器之永磁同步 風力發電系統，包括：一風力機、一永磁同步發電機、一 轉換器（Converter)、一反流器（Inverter)及一智慧型最大功 率追縱器。該永磁同步發電機用以接收該風力機之機械 能，並輸出三相交流電能。該轉換器用以將該三相交流電 能轉換為直流電。該反流器用以將該直流電轉換為交流 電。該智慧型最大功率追蹤器包括一爬坡控制電路、一 Wilcoxon徑向基底類神經網路及一電流控制器。該爬坡控 φ 制電路用以依據該直流電之一實際直流電壓及一實際直流 電流，於一最大功率曲線對應計算一設定直流電壓。該 Wilcoxon徑向基底類神經網路用以依據該實際直流電壓及 $設定直流電壓，計算-命令電流。該電流控制器用以依 據該交流電之一實際交流電流及該命令電流，輸出一控制 值至該反流器^ 本發明利用該爬坡控制電路及該WHc〇x〇n徑向基底類神 經網路，可達到良好控制效果，並且本發明之系統不需升 壓型（Boost)轉換器及偵測發電機之轉速，可降低系統成 1493I3.doc 201216594 本。本發明之永磁同步風力發電系統可實現變速運轉，及 控制風力機保持在最佳尖端速度比及最大功率係數附近運 行，以使風能獲得較高能量轉換效率，明顯提高發電量。 【實施方式】 參考圖2’其顯示本發明利用智慧型最大功率追蹤器之 永磁同步風力發電系統之電路方塊示意圖。本發明利用智 慧型最大功率追蹤器之永磁同步風力發電系統2〇包括：一 風力機21、一永磁同步發電機22(PMSG)、一轉換器 23(Converter)、一反流器25(Inverter)及一智慧型最大功率 追蹤器26。該永磁同步發電機22用以接收該風力機21之機 械能，並輸出三相交流電能。 該轉換器23用以將該三相交流電能轉換為直流電。在本 實施例中’該轉換器23包括複數個二極體（例如：六個二 極體），組成為一個三相全波整流電路。該反流器25用以 將該直流電轉換為交流電。在本實施例中，該反流器25包 括複數個反流單元（例如：四個反流單元），每一反流單元 具有一電晶體及一個二極體。 該智慧型最大功率追蹤器26包括一爬坡控制電路261、 一 Wilcoxon徑向基底類神經網路262及一電流控制器263 » 該爬坡控制電路261用以依據該直流電之一實際直流電壓 Vdc及一實際直流電流Idc，於一最大功率曲線對應計算一 設定直流電壓G。該Wilcoxon徑向基底類神經網路262用 以依據該實際直流電壓Vd(：及該設定直流電壓ρς，計算一 命令電流Id。該電流控制器263用以依據該交流電之一實際 149313.doc 201216594 交流電流！及該命令電流Id，輸出一控制值至該反流器25。 在本實施例巾1¾爬坡控制電路26 i依據該實際直流電 壓Vdc及該實際直流電流Id。計算得—直流功率Pd。，該直流 功率Pde近似於該最大功率曲線之—機械功率匕。參考圖 3_’立其顯示複數個最大功率曲線及其對應之最佳操作點之 不思圖’其中風速Ul<u2<u3<u4。依據該最大功率曲線之 該機械功率Pm與該設定直流電壓〇關係，對應計算該設 疋直流電壓。為得到最大功率，最佳的該設定直流電壓 L必須利用爬坡法即時搜尋。利用該爬坡控制電路261， 若該設定直流電壓ρς是隨著該機械功率匕之增加而增加， 則該設定直流電壓匕:之搜尋方向與該機械功率匕之增加方 向相同；反之，則搜尋方向為相反，例如：若風速之改變 為，則該設定直流電壓π之搜尋為 J — S — C — — ^。且該機械功率！^之增加量近似於該直流功 率Pde之增加量，故可在該直流功率Pde近似等於該機械功 率Pm及風力機慣量可降至最低之動態平衡操作點情形下， 執行該設定直流電壓<之搜尋。在動態情形下，該設定直 流電壓 <可保持且該Wilcoxon徑向基底類神經網路262可 即時調整負載電流，使得系統盡快達到其平衡點。 參考圖4’其顯示本發明之該WilcoxOI1徑向基底類神經 網路之階層示意圖。該Wilcoxon徑向基底類神經網路262 包括一輸入層、一隱藏層及一輸出層。其中該輸入層計算 該實際直流電壓及該設定直流電壓之一誤差函數，在該輸 入層之輸入為力⑴及尤^其中矸一匕-^^”及十^^在該輸 149313.doc 201216594 入層之節點(nodes)用以直接傳送輸入至下一層。亦即，對 於該輸入層之第!個節點，其輸入及輸出可如式⑴表示。 «以,(丨)=〇) y?\N) = f^{netf\N))=netf\N) .·_η ί-U (!)201216594 VI. Description of the Invention: [Technical Field] The present invention relates to a permanent magnet synchronous wind power generation system, and more particularly to a permanent magnet synchronous wind power generation system using a smart maximum power tracker. [Prior Art] A conventional wind power generation system can operate at a constant speed or a variable speed using a power electronic converter. Since variable speed power generation can achieve maximum efficiency at all wind speeds to increase output energy and reduce voltage flicker, variable speed power generation is often used in the industry. The research and practice of many generators is that wind turbines are induction machines with winding rotors or squirrel cage rotors. Recently, the application of permanent magnet synchronous generators has gradually increased. High-performance and variable-speed power generation with high efficiency and high controllability can be achieved with permanent magnet synchronous generators. The well-known research/mains are in the two largest wind control methods': tip-speed ratio (TSR), power signal feedback control (PSF) and hill climbing control (hill-ciimb searching, HCS). Referring to Figure 1 ', it shows a schematic diagram of the relationship between the tip speed ratio and the power factor. The tip speed ratio control is to control the wind turbine rotor speed to maintain an optimum tip speed ratio. The power signal feedback requires knowing the wind turbine's maximum power curve and tracking its curve via its control mechanism. In the conventional wind power maximum power point tracking strategy, the tip speed ratio control method is difficult to obtain wind speed and wind turbine speed, so its use is limited. Many conventional wind power maximum power point tracking strategies are used by using wind turbines. The maximum power curve to reduce the measurement' but still have to know the characteristics of the wind turbine. The conventional climbing system control system 149313.doc 201216594 continuously searches for the wind turbine's tip output power. In comparison, the maximum power point tracking method for wind power generation controlled by the hill climbing method is popular because of the simplicity and independence of system characteristics. Therefore, it is necessary to provide an innovative and progressive permanent magnet synchronous wind power generation system utilizing the intelligent maximum power tracker to solve the above problems. . SUMMARY OF THE INVENTION The present invention provides a permanent magnet synchronous wind power generation system using a smart maximum power tracker, including: a wind turbine, a permanent magnet synchronous generator, a converter, and a reverser (Inverter) ) and a smart maximum power tracker. The permanent magnet synchronous generator is configured to receive the mechanical energy of the wind turbine and output three-phase alternating current energy. The converter is used to convert the three-phase alternating current energy into direct current. The inverter is used to convert the direct current into alternating current. The intelligent maximum power tracker includes a hill climbing control circuit, a Wilcoxon radial base-like neural network, and a current controller. The climbing control φ circuit is configured to calculate a set DC voltage corresponding to a maximum power curve according to an actual DC voltage of the DC power and an actual DC current. The Wilcoxon radial base-like neural network is used to calculate the - command current based on the actual DC voltage and the set DC voltage. The current controller is configured to output a control value to the inverter according to the actual alternating current of the alternating current and the command current. The present invention utilizes the hill climbing control circuit and the WHc〇x〇n radial base-like neural network. A good control effect can be achieved, and the system of the present invention does not require a boost converter and detects the rotational speed of the generator, and can reduce the system to 1493I3.doc 201216594. The permanent magnet synchronous wind power generation system of the present invention can realize the shifting operation, and control the wind turbine to keep running at the optimum tip speed ratio and the maximum power coefficient, so that the wind energy can obtain higher energy conversion efficiency and significantly increase the power generation amount. [Embodiment] Referring to Fig. 2', there is shown a block diagram of a circuit of a permanent magnet synchronous wind power generation system using the smart maximum power tracker of the present invention. The permanent magnet synchronous wind power generation system 2 of the present invention utilizing the intelligent maximum power tracker includes: a wind turbine 21, a permanent magnet synchronous generator 22 (PMSG), a converter 23 (Converter), and a inverter 25 ( Inverter) and a smart maximum power tracker 26. The permanent magnet synchronous generator 22 is configured to receive the mechanical energy of the wind turbine 21 and output three-phase alternating current energy. The converter 23 is used to convert the three-phase alternating current electrical energy into direct current. In the present embodiment, the converter 23 includes a plurality of diodes (e.g., six diodes) and is composed of a three-phase full-wave rectifier circuit. The inverter 25 is for converting the direct current into alternating current. In the present embodiment, the inverter 25 includes a plurality of reverse flow units (e.g., four reverse flow units) each having a transistor and a diode. The intelligent maximum power tracker 26 includes a hill climbing control circuit 261, a Wilcoxon radial base-like neural network 262, and a current controller 263. The hill climbing control circuit 261 is configured to generate an actual DC voltage Vdc according to the DC power. And an actual DC current Idc corresponding to a set DC voltage G corresponding to a maximum power curve. The Wilcoxon radial base-like neural network 262 is configured to calculate a command current Id according to the actual DC voltage Vd (and the set DC voltage ρ 。). The current controller 263 is configured to be based on the actual AC 149313.doc 201216594 The alternating current! and the command current Id output a control value to the inverter 25. In the present embodiment, the hill climbing control circuit 26 i calculates the DC power according to the actual DC voltage Vdc and the actual DC current Id. Pd., the DC power Pde approximates the mechanical power 该 of the maximum power curve. Referring to Figure 3_', it shows the plurality of maximum power curves and their corresponding optimal operating points, where the wind speed Ul<u2<U3<u4. The mechanical power Pm according to the maximum power curve is related to the set DC voltage ,, and the set DC voltage is correspondingly calculated. To obtain the maximum power, the optimal set DC voltage L must be searched by the hill climbing method. By using the climbing control circuit 261, if the set DC voltage ρς is increased as the mechanical power 匕 increases, the setting of the DC voltage 匕: The direction of the increase is the same as the direction of the mechanical power ;; otherwise, the search direction is opposite, for example, if the change of the wind speed is, the search for the set DC voltage π is J — S — C — — ^, and the mechanical power! The increase of ^ is similar to the increase of the DC power Pde, so the set DC voltage can be executed under the condition that the DC power Pde is approximately equal to the mechanical power Pm and the wind turbine inertia can be minimized. In the dynamic case, the set DC voltage < maintains and the Wilcoxon radial substrate-based neural network 262 can instantly adjust the load current so that the system reaches its equilibrium point as quickly as possible. Referring to Figure 4', the present invention is shown A hierarchical diagram of the WilcoxOI1 radial base-like neural network. The Wilcoxon radial base-like neural network 262 includes an input layer, a hidden layer, and an output layer, wherein the input layer calculates the actual DC voltage and the set DC voltage. One of the error functions, the input at the input layer is the force (1) and especially the 矸一匕-^^" and the ten ^^ at the node of the input 149313.doc 201216594 (nodes) is used to directly transfer input to the next layer. That is, for the first node of the input layer, its input and output can be expressed as equation (1). «以,(丨)=〇) y?\N) = f^{netf\N))=netf\N) .·_η ί-U (!)
該隱藏層依據該誤差函數進行一高斯函數運算，以計算 得-高斯函數運算結果。在本實施例中，在該隱藏層之每 一節點進行一高斯函數運算（Gaussian “仏functi〇n)，該 高斯運算（幅狀基底函數⑽W basis function)之一特殊例） 於此處用以做為一隸屬函數（membership ，如下 式（2)所示。 w<)W = -I(^1)-c,)2/v, yf\N) = fp{net^\N))= Qxp(net^\N)) j =! 9 (2) 其中，〜…及％分別表示為該高斯函數之平 均值及標準偏差值。 該輸出層對該高斯函數運算結果進行一權重值運算，以 計算得該命令電流Id。在該輸出層之單一節點k表示為計算 所有輸出為所有輸入訊號之總和，如下式（3)所示。 netf = twjkyf\N) 7-1 y(k\^) = fk(3) (net^ (Ν)) = net^(N) = Id ^ (3) 其中，為隱藏層及輸出層間之權重值。 該Wilcoxon徑向基底類神經網路262另甶杜 力L括一訓練及學 習裝置264’用以調整該誤差函數’更新該輸出層之複數 個權重值，及更新該隱藏層之該高斯函數之 妖 < 復數個平均值 及標準偏差值。首先誤差函數可被最小化，l 化如下式（4)所 149313.doc 201216594 示The hidden layer performs a Gaussian function operation according to the error function to calculate a Gaussian function operation result. In this embodiment, a Gaussian function operation (Gaussian "仏functi〇n" is performed at each node of the hidden layer, and a special example of the Gaussian operation (W) function is used here. As a membership function (membership, as shown in the following formula (2). w<) W = -I(^1)-c,)2/v, yf\N) = fp{net^\N))= Qxp (net^\N)) j =! 9 (2) where ~... and % are respectively expressed as the mean value and standard deviation value of the Gaussian function. The output layer performs a weighted value operation on the Gaussian function operation result, The command current Id is calculated. The single node k at the output layer is expressed as the sum of all the outputs for all input signals, as shown in the following equation (3): netf = twjkyf\N) 7-1 y(k\^) = fk(3) (net^ (Ν)) = net^(N) = Id ^ (3) where is the weight value between the hidden layer and the output layer. The Wilcoxon radial base-like neural network 262 is another L includes a training and learning device 264' for adjusting the error function to update a plurality of weight values of the output layer, and updating the monster of the Gaussian function of the hidden layer < a plurality of average values and labels Deviation First error function may be minimized, (4) as shown 149313.doc 201216594 l of the following formula
E =Ie2 (4) 在輸出層& , 站增中「，誤差項被展開為如下式（5)所示°k =>- 一… Qnet^ __θ£_ φΟ) - 權重值：整「為如下式(6)所示 Δμ； ,, = _ ^ρΕ Λ IT 九.⑶ …、 (5) '息 _pE ay<3) dnetf) dw, = δΑ^- (2) ⑹E = Ie2 (4) In the output layer & , station increment ", the error term is expanded as shown in the following equation (5) °k => - a... Qnet^ __θ£_ φΟ) - Weight value: whole " Δμ; , , = _ ^ρΕ Λ IT 九.(3) ..., (5) ' interest _pE ay<3) dnetf) dw, = δΑ^- (2) (6)
因此，如下式（7)所示。 ’+1)=〜⑻+ 盆中， \為用以調整權重值％之學習比。 在该隱藏層中再進行乘法運算，對於平均值％之法則j 如下式（8)所示。 ,2(x,(1)-c,)Vv (8) ⑺ △C“ dE --dc “ dE dnet^ dy^ 對於標準偏差值vi/之法則為如下式（9)所示 Δν,,. ^ dE dnet?] δν?^ ?ίγ(1) Τ dvTherefore, it is shown by the following formula (7). ‘+1)=~(8)+ In the basin, \ is the learning ratio used to adjust the weight value %. The multiplication operation is performed again in the hidden layer, and the rule j for the average value is expressed by the following equation (8). , 2(x, (1)-c,) Vv (8) (7) △ C " dE -- dc " dE dnet ^ dy ^ The standard deviation value vi / is the following equation (9) Δν,,. ^ dE dnet?] δν?^ ?ίγ(1) Τ dv
dE dnet^ dy^ dnet^ 因此，如下式（10)所示。 S㈣=咖+η為dE dnet^ dy^ dnet^ Therefore, it is as shown in the following formula (10). S (four) = coffee + η is
"hWjk~W (9) Ά+ι)=Ά)+η，ν" (I。） 其中’ 1及η。分別為用以調整平均值C(/及標準偏差 之學習比。 利用該訓練及學習裝置264可使該Wilcoxon徑向基底類 神經網路262之基底逐漸降低，以降低計算複雜度。 再參考圖2’本發明之該永磁同步風力發電系統2〇另包 I49313.doc 201216594 括一直流鏈結電路24(DC link)，其包括一直流電容器241 及一個二極體242。本發明之該永磁同步風力發電系統20 另包括一負載電路27，連接至該反流器25，該負載電路27 包括一負載電感器271及一負載電容器272。 在本實施例中，該電流控制器263係為一比較器，用以 比較該交流電之該實際交流電流I及該命令電流Id，且該控 制值係為一脈波寬度調變訊號（PWM)。 本發明利用該爬坡控制電路及該Wilcoxon徑向基底類神 經網路’可達到良好控制效果，並且本發明之系統不需升 壓型（Boost)轉換器及偵測發電機之轉速，可降低系統成 本。本發明之永磁同步風力發電系統可實現變速運轉，及 控制風力機保持在最佳尖端速度比及最大功率係數附近運 行’以使風能獲得較高能量轉換效率，明顯提高發電量。 上述實施例僅為說明本發明之原理及其功效，並非限制 本發明。因此習於此技術之人士對上述實施例進行修改及 變化仍不脫本發明之精神。本發明之權利範圍應如後述之 申清專利範圍所列。 【圖式簡單說明】 圖1顯示尖端速度比與功率係數之關係示意圖。； 圖2顯示本發明利用智慧型最大功率追蹤器之永磁同步 風力發電系統之電路方塊示意圖； 圖3顯示複數個最大功率曲線及其對應之最佳操作點之 示意圖；及 圖4顯不本發明之該WUc〇x〇n徑向基底類神經網路之階 149313.doc 201216594 層示意圖。 【主要元件符號說明】 20 本發明之永磁同步風力發電系統 21 風力機 22 永磁同步發電機 23 轉換器 24 直流鏈結電路 25 反流器"hWjk~W (9) Ά+ι)=Ά)+η,ν" (I.) where '1 and η. The learning ratio is used to adjust the average value C (/ and the standard deviation. The training and learning device 264 can gradually reduce the base of the Wilcoxon radial base-like neural network 262 to reduce the computational complexity. 2' The permanent magnet synchronous wind power generation system of the present invention is further packaged I49313.doc 201216594 includes a DC link circuit including a DC capacitor 241 and a diode 242. The present invention The magnetic synchronous wind power generation system 20 further includes a load circuit 27 connected to the inverter 25, the load circuit 27 includes a load inductor 271 and a load capacitor 272. In this embodiment, the current controller 263 is a comparator for comparing the actual alternating current I of the alternating current with the command current Id, and the control value is a pulse width modulation signal (PWM). The present invention utilizes the hill climbing control circuit and the Wilcoxon path A good control effect can be achieved to the substrate-based neural network, and the system of the present invention can reduce the system cost without the need for a boost converter and detecting the rotational speed of the generator. The permanent magnet of the present invention The step wind power generation system can realize the variable speed operation, and control the wind turbine to keep running at the optimal tip speed ratio and the maximum power coefficient to enable the wind energy to obtain higher energy conversion efficiency and significantly increase the power generation amount. The above embodiment is only for explaining the present invention. The invention and its effects are not intended to limit the present invention, and those skilled in the art will be able to make modifications and variations to the above-described embodiments without departing from the spirit of the invention. The scope of the invention should be as defined in the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a schematic diagram showing the relationship between the tip speed ratio and the power factor. Fig. 2 is a block diagram showing the circuit of the permanent magnet synchronous wind power generation system using the smart maximum power tracker of the present invention; Schematic diagram of the maximum power curve and its corresponding optimal operating point; and Figure 4 shows the outline of the WUc〇x〇n radial base-like neural network of the present invention. 149313.doc 201216594 Layer diagram. [Main component symbol description] 20 Permanent magnet synchronous wind power generation system 21 of the present invention Wind turbine 22 Permanent magnet synchronous generator 23 Converter 24 DC link 25 Anti-way flow
26 智慧型最大功率追蹤器 27 負載電路 241 直流電容器 242 二極體 261 爬坡控制電路 262 Wilcoxon徑向基底類神經網路 263 電流控制器 264 訓練及學習裝置 271 負載電感器 272 負載電容器 149313.doc26 Smart Maximum Power Tracker 27 Load Circuit 241 DC Capacitor 242 Diode 261 Hill Climb Control Circuit 262 Wilcoxon Radial Basis Neural Network 263 Current Controller 264 Training and Learning Device 271 Load Inductor 272 Load Capacitor 149313.doc