在本篇論文中,所提出之適應式自我建構模糊類神經滑動模式控制器及補償控制器可用來達成太陽能最大功率追蹤。在此控制器中的結構學習及參數學習使其具有自動產生與及時更新的效果。結構學習方面,神經元個數的增減是由馬氏距離方法來決定的。此外一個新準則可以使增長的歸屬函數標準差落在合適的範圍內;而參數學習方面,是藉由適應法則來調整類神經結構內的參數,其穏定性也已經由李普諾夫函數得到證明。適應式自我建構模糊類神經在此是視為一個接近理想的控制器,補償控制器是用來縮小類神經控制器和理想控制器之間的誤差。最後,模擬的結果證實了所提出的控制器在一般情形或是陰影條件下的成效與效用性。
An adaptive self-constructing fuzzy neural network (ASCFNN) sliding-mode controller with compensation controller which is used to track maximum power point (MPP) in solar panel is proposed in this thesis. The structure and parameter learning phase are done automatically and online in the ASCFNN. The structure learning phase determining the neurons will be generated/eliminated or not is based on the Mahalanobis distance (M-distance) method. Moreover, a new criterion is used to set standard value of the membership function in suitable region. The parameter learning phase adjusting the parameters of the network is based on the adaptive method with sliding-mode control, and its stability deriving from Lyapunov sense also can be guaranteed. An ASCFNN is utilized to mimic an ideal controller, and the compensation controller is designed to reduce the approximation error making the neural controller to be close to ideal controller. Finally, the simulation results verify the performance and effectiveness of the proposed control schemes in normal case or in shading conditions.