Exhibition Program

Science of Machine Learning

02

Efficient and comfortable AC control by AI

- Environment reproduction and control optimization system -

Abstract

We propose an air-conditioning control system by AI to save more enegy and to be more comfortable. In a lergescaled facility, it takes serveral time to stabilize temperature. Traditional and typical way of control system, commonly known as feed-back control, makes sometimes uncomfortable and consumes extra enegy by the timedelay. On the other hand, feed-forward control determines suitable control with predicting environment status of the facility. For example, if congestion is predicted, the air-flow could be increased or decreased in advance, which would make the facility’s temperature suitable. We developed AI consisting of environment reproduction system and control optimization system to calculate the optimal operation schedule for multiple air-conditioning flows, and demonstrated the importance of feed-forward control through field trial at “COREDO Muromachi”, which is one of the largest-scale commercial facilities, with NTT-Facilities and MITSUI FUDOSAN.

References

  • [1] I. Shake, K. Kawase, Y. Suzuki, “NSRI×NTT×MITSUI FUDOUSAN Collaboration results: Commenced joint experiments to utilize urban big data and AI in Nihonbashi Muromachi area,” NTT technical journal, Vol. 29, No. 11, pp. 63-65, 2017.

Poster

Photos

Contact

Nobuhiko Matsuura, Network Innovation Laboratories
Email: