25079

مديريت شارژ خودروهاي الكتريكي با استفاده از روش تركيبي مبتني بر يادگيري ماشين

كارشناسي ارشد

مهندسي كامپيوتر گرايش هوش مصنوعي و رباتيك

1396

بهمن ماه 99

دكتر ناصر مزيني

مهندسي كامپيوتر

1400/05/20

Applying Hybrid Model based-on Machine Learning Approaches to EV Charging Scheduling Problem

1/1/1900 12:00:00 AM

Considering the potential of electric vehicles on environmental pollution reduction and increasing public attention on alternating gasoline vehicles, also, it has recently attracted researchers to expand its role in smart grid. Keeping in mind, with appearance of new technologies in making electric vehicle battery that expands its charging maintenace and also new charging station which can fully charge batteries in less than an hour, electric vehicle can play a vital role in balancing loads in grid and also to reduce cost of power generation. But, due to the uncertainty of power prices and also the randomness of user's commuting behaviour the problem of scheduling of electric vehicle charging becomes overly challenging. In this thesis, we took advantage of proficiency of machine learning methods on dealing with complex data to extract trend feature of power price and fed it to a trust region policy optimization actor to provide a schedule for charging and discharging of evs object to profit maximization of electric vehicle owners. Different circumstances were considered to compare with the proposed model, which obtained results show the superiority of this model over recent methods.

Smart Grid, Electric Vehicle Charging Scheduling, Machine Learning, Policy Gradient methods