Optimizing Chiller Switch-on Time Interval for Chiller Power Consumption Saving Via Big Data Analytics and Machine Learning Framework

In semiconductor manufacturing, the chiller water system requires huge energy consumption, especially in the countries which have high temperature and humidity climate such as Taiwan. In order to minimize chiller power consumption without affecting the environment of wafer production, optimizing chiller system operations become a crucial issue. Conventionally, chiller operations greatly depend on engineers' practical experiences. However, various uncertainties, including changeable weather and complicated chiller combinations, lead to inconsistent decisions of switching chiller machines as well as energy waste [1]. To improve the operational performance of the system for energy saving, researchers have proposed many different types of solutions, but those technologies are not easy to widely adopted in practical applications due to the complicated and limited operations and models.