时 间：2019年5月14日10:30-11:30
地 点：旗山校区数信大楼511
主 讲：瑞典达拉那大学 韩梦捷 高级讲师
主 办：数学与信息学院
专家简介：韩梦捷，男，1985年11月生，2014年4月获得瑞典达拉那大学微观数据分析博士学位。现为瑞典达拉那大学高级讲师，曾先后担任多门统计、数学等相关课程的教学工作。博士后期间，参与并完成了多个科研项目，主要运用统计学、线性规划、机器学习等方法对地理信息数据、人口数据、建筑能源数据等所产生的优化问题进行建模、优化、分析。已发表学术论文十多篇。今后科研发展方向：开发并完善人工智能及优化问题算法，解决智慧城市建设、能源可持续利用、环境可持续发展中的产生的问题。
报告摘要：Occupants’window opening and closing behaviour influence significantly in both building energy consumption and indoor comfort. However, such behaviour is complex and hard to predict and control in a conventional way, due to the accumulated and interlinked impacting factors of the environment, physiology, economy and social issues. We propose a novel reinforcement learning (RL) method for the advanced control strategy in window opening and closing. The RL control strategy aims at optimising the time point for window opening and closing through observing and learning from the environment. The basic theory of model-free RL control strategy is firstly developed with the objective of indoor comfort time, which is then applied to field measurement data taken from an office building in China. A preliminary testing of this RL control strategy is subsequently conducted by evaluating the total time located in the indoor comfort zone. The results indicate that the RL control strategy improves thermal and indoor air quality by more than 90% when compared to the actual observed data. We establish a prototype for optimally controlling occupants’ behaviour in window opening and closing that can be further extended by including more environmental parameters and multi objectives.