臺灣博碩士論文加值系統

在冷房中影響人舒適度的主要因素中，室溫、風速和濕度可以藉由改變壓
縮機和風扇的轉速來改變。然而，若用lumped parameter 模式來描述這
些狀況之間的關係，則無法考慮冷房中各區域狀態不同的問題。因此根據
以往的數值模擬和控制的經驗，本研究利用反應表面法(RSM)來規劃實驗
。將所獲得的數據用於建立數值和模糊的模式。為了建立模糊模式，用
mountain clustering 和subtractive clustering 建立rule 的架構。然
後運用倒傳遞法來調整模糊規則的參數，根據這三種模式的預測性能來比
較和討論模式的好壞。

Among the main factors which affect the human comfortness in
anair-conditioned room, the room temperature, air velocity and
humidity can be changed by control the rotational speed of
compressor and fan. However, if a lumped parameter model is used
to describe the relation among these state variables, the fact
that the conditions are different all over the air-conditioned
room is ignored. Therefore, based on the experience of numerical
simulation and controlof air-conditioned room, this research
design the experiment by response surface method (RSM). The data
obtained is used to build statistic and fuzzy model. To build
the fuzzy model, the mountain clustering and subtractive
clustering are used to identify the rule structure. Then, the
back-propagation algorithm is used to tune the parameters of the
fuzzy rule bases. The three model thus built are compared and
discussed based on the performance of prediction.