Applying Neural Network and Evolution Strategy to Optimal Loading for RTU Systems

• Main Authors: LIU, SHENG-HUA, 劉勝驊
• Other Authors: LEE, WEN-SHING
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/348h45

碩士 === 國立臺北科技大學 === 能源與冷凍空調工程系 === 107 === The air conditioner system installed in our research institute is known as Rooftop Air Conditioning Unit (RTU). This system is the new modern selection for the market, with the advantage of eliminating the need for big spaces. The standard chiller system among commonly known air conditioner requires a separate engine room for its placement alongside a water tower placed on an empty space. However the RTU simply requires an empty space without taking up the same amount of rooms as it doesnt need an engine room. Should there be an event which requires huge amount of space, the RTU can do the job by freeing up spaces for other uses such as car parking or vendors/retails available, contributes more in overall sales. Our research institute mainly uses neural network structure, then takes advantage of the Equal Load Distribution (ELD) and Evolution Strategy (ES) to minimize the electric payload. This enables all the RTU installed in an area to emit the most optimal temperature output. The research conclusion discovered applying ED saves 60% - 95% of room space and 4% - 10% energy, a difference of 7.69% in electricity consumption. Therefore, ED system applied in RTU proven to efficiently conserves energy.