EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
3009. A TOU pricing approach for EV charging via DRL
Invited abstract in session TB-22: Optimization for electric vehicles, stream Energy Management.
Tuesday, 10:30-12:00Room: 81 (building: 116)
Authors (first author is the speaker)
| 1. | Abdullah Kayacan
|
| Industrial Engineering, Bogazici University | |
| 2. | Mehmet Yasin Ulukus
|
| Industrial Engineering, Istanbul Technical University | |
| 3. | Hatice Tekiner Mogulkoc
|
| Industrial Engineering, Istanbul Technical University | |
| 4. | Engin Cicek
|
| Industrial Engineering, Marmara University | |
| 5. | Murat Bilsel
|
| Industrial Engineering, Marmara University | |
| 6. | Bahadir Tunaboylu
|
| Materials Eng, Marmara University |
Abstract
Economic and ecological advantages make electric vehicles (EVs) more desirable than internal combustion engine vehicles. The charging need of EVs can have a critical impact on electricity distribution at certain penetration levels. Irregular intraday electricity demand may cause overloads, thus increasing costs and decreasing quality. Appropriate charging pricing policies can be effective in reducing the imbalance in electricity demand. In this study, we examined the effects of fixed and time-of-use (TOU) pricing policies when users have different charging options. We developed a comprehensive simulation model to examine users' behavior in different charging option distributions. A Deep Reinforcement Learning (DRL) model is proposed to determine the pricing policy that offers the best price levels in terms of invariability of the intraday aggregate EV load profile. Our experiments in different charging option availability levels show that the TOU pricing schemes obtained from the proposed DRL model offer prices leading to more balanced load curves compared to fixed-price charging pricing. We also show that peak loads are more likely to occur when EV drivers have similar charging options. In these cases, generated TOU price tables reduce the standard deviation of the load profile more noticeably. This study is supported by TUBITAK under grant number 221M111.
Keywords
- Energy Policy and Planning
- Machine Learning
- OR in Energy
Status: accepted
Back to the list of papers