可微分小腦模型控制器於雷射二極體之溫控研究

Abstract

雷射二極體作為光纖通訊中的光源，而最重要的特性是其波長的穩定。當雷射二極體在高功率的操作下，由於熱的產生導致溫度升高，影響了雷射二極體所發出之光功率及波長的變動。而熱流系統的溫度變化具有非線性及時變的特性，不易以傳統之控制方法取得其數學模型；因此，本文主要利用可微分小腦模型控制器( Differentiable Cerebellar Model Articulation Controller，DCMAC )來學習雷射模組之特性，進而控制其溫度。在結果中討論可微分小腦模型控制器的設計及其性能的改進。DCMAC與PID控制器、模糊控制器( Fuzzy Controller )作比較可發現，DCMAC於雷射模組的溫度控制，系統的安定時間短、穩態誤差以及過衝量小，比起其它控制方法具有較佳的性能。

Laser diode which acts as light sources in laser module is used in optical fiber communication. One of the most desired characteristics of laser diode is its wavelength stability. When the laser is operated at higher power, more heat is produced, which will induce higher temperature. Then it causes the variation of output power and wavelength of lasers diode. With the characteristic of the nonlinear time-varying thermo-fluid system, it is difficult to define a mathematic model for temperature control. Therefore, in this paper, a differentiable cerebellar model articulation controller (DCMAC) was used to learn the characteristics and control the temperature of laser module. The results demonstrated how to design a DCMAC and improve its performance. The comparison was also made between DCMAC, PID and Fuzzy controller. It showed that DCMAC has the better performance with smaller settling time, steady state error and overshoot in temperature control of laser modules.