全球暖化現象已經成為國家安全議題，企業馬上可以做的事情就是把相對耗能的設備及系統導入節能技術，使系統達到最佳化控制也讓廠區用電降低。
然而電子廠房廠務設備能耗最高的莫屬冰水主機系統，早期的節能技術著重於高耗能冰水主機的設定調整與保養維護，進而於附屬設備（冰水泵、冷卻水泵、冷卻水塔）導入變頻器技術可大幅降低能耗。因此近年來各企業皆於冰水主機系統的節能技術能力下，將冰水主機監控電腦撰寫最佳化節能程式，讓系統盡可能達到最佳運轉狀況；但如此複雜的運算程式撰寫在冰水主機監控系統有相當大風險，而且這樣的方式其實是未考慮到設備的性能衰減或是流量減少問題，而是朝著理論值推估，最終因為監控電腦無法負荷且每年須修正，是相當不適合無人化工廠目標。
深度學習是一種人工智慧(AI)方法，可建議電腦並且受人類腦部啟發方式來處理繁雜的資料，把此運轉負荷由監控電腦轉到中央處理伺服器進行大量的運轉，透過一次又一次的學習，可以讓冰水主機系統由監督式學習變成深度學習後，系統一共可以節省3.5%，假設每度電費為2.5元台幣與年度平均製冷量為21,000RT，每年可節省910萬的電費，抑制排碳量約1,854公噸，約4.8座大安森林公園一年的固碳量。

Chilled water system early energy-saving technology focuses on the setting adjustment and maintenance of the high-energy-consuming chilled water host, and then the introduction of inverter technology in the auxiliary equipment can reduce energy consumption. Therefore, in recent years, under the energy-saving technical capabilities of the chilled water system, all enterprises have written and optimized energy-saving programs on the chilled water monitoring computer to achieve the best operation of the system as much as possible. However, such a complex algorithm is written in the chilled water host monitoring system has considerable risks, and this method actually does not consider the performance degradation of the equipment or the reduction of traffic, but is extrapolated towards the theoretical value, and finally because the monitoring computer cannot be loaded and must be corrected every year, it is quite unsuitable for the goal of unmanned chemical plants.
Deep learning is a method in artificial intelligence (AI), which can guide the computer to process data in a way inspired by the human brain, and transfer this operation load from the monitoring computer to the central processing server for a large number of operations, through learning again and again, the chilled water host system can be changed from supervised learning to deep learning, the system can save a total of 3.5%, assuming that the electricity cost per kWh is NT$2.5 and the annual average cooling capacity is 21,000RT, which can save 9.1 million electricity bills per year.

中文部分
黃乃容 (2016)。半導體廠房冰水系統之能源績效指標。國立成功大學土木工程研究所碩士論文。
黃仲翊 (2017)。外部輸入非線性自動迴歸模型應用於冰水主機耗能分析。國立台北科技大學冷凍空調工程系所碩士論文。
蔡奇芝 (2020)。以大數據分析達成冰水主機系統節能目的。國立清華大學工業工程與工程管理學系碩士論文。
潘佳欣 (2020)。冰水主機系統之預測模型與實證研究。國立清華大學工業工程與工程管理學系碩士論文。
英文部分
Guideline, A. S. H. R. A.E. (2002). Guildline 14-2002, Measurement of energy and demand savings. American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., Atlanta, USA.
Rocco Logozzo, S.A. Armstrong Limited (2012). Armstrong pump submittal paper.
網站部分
Trane® CenTraVac™ Centrifugal Chillers. Products-Duplex CenTraVac Chiller-CDHF/CDHG (tranehk.com)
Amazon Web Services, Inc. (2023). https://aws.amazon.com/tw/what-is/deep-learning/