A Dynamic Optimal Battery Swapping Mechanism for Electric Vehicles Using an LSTM-Based Rolling Horizon Approach

This paper proposes a new approach for optimal operation of an Electric Vehicle (EV) battery-swapping station (BSS) based on Rolling-Horizon optimization (RHO). The BSS has several swapping bays such that each can accommodate an EV for swapping single or multiple battery units. The proposed BSS mode...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on intelligent transportation systems 2022-09, Vol.23 (9), p.15218-15232
Main Authors: Shalaby, Ahmed A., Shaaban, Mostafa F., Mokhtar, Mohamed, Zeineldin, Hatem H., El-Saadany, Ehab F.
Format: Article
Language:English
Subjects:
Batteries
Battery swapping stations
battery to grid
Charging
Electric vehicles
EV charging stations
Forecasting
LSTM
MILP
Mixed integer
Nonlinear programming
Optimization
Predictive models
Recurrent neural networks
Roads
rolling-horizon optimization
Scheduling
State of charge
Time series analysis
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper proposes a new approach for optimal operation of an Electric Vehicle (EV) battery-swapping station (BSS) based on Rolling-Horizon optimization (RHO). The BSS has several swapping bays such that each can accommodate an EV for swapping single or multiple battery units. The proposed BSS model considers serving different types of EVs using a heterogeneous battery stock. The charging of the depleted batteries (DBs) is performed using continuously controlled variable chargers which makes it more flexible for providing grid services. While previous studies focused on day-ahead modeling of BSSs, our study considers BSS dynamic scheduling. The goal is to maximize the daily profit using an RHO mechanism to provide optimal swapping and charging/discharging processes. The problem is defined as mixed-integer nonlinear programming (MINLP), then it's linearized into a mixed-integer linear problem (MILP) to reduce the computational complexity. To predict the EV's swapping demand, a long short-term memory (LSTM) recurrent neural network is utilized as a time series forecasting engine. The proposed model is validated through a set of case studies comparing the LSTM-based RHO mechanism versus unscheduled operation and day-ahead scheduling. Simulation results demonstrate that the proposed dynamic scheduling mechanism increases the profit between 10% and 25.7% compared to the day-ahead scheduling. Furthermore, the number of EVs served using the proposed approach increases between 11% and 14% compared to the day-ahead model.
ISSN:1524-9050
1558-0016
DOI:10.1109/TITS.2021.3138892