- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Neural network trained via reinforced learning as control...
Open Collections
UBC Theses and Dissertations
UBC Theses and Dissertations
Neural network trained via reinforced learning as control for a battery super-capacitor hybrid energy storage system in electric vehicles Dueck, Kevin
Abstract
Electric vehicles (EVs) are becoming a more popular alternative to internal combustion engine vehicles, however a concern among EV manufacturers and customers is the longevity of the EV's energy source, the battery. The battery is a large contributor to the cost of an EV and is susceptible to wear due to charge/discharge cycles and heat. This wear is a main depreciator to the EV's worth. Batteries are considered a low power density energy storage device, however a hybrid energy storage system (HESS) can be formed with an improved power density by interfacing batteries and super-capacitors (SCs). An HESS utilizes the high energy density of batteries and the high power density of SCs; this lowers the wear of the batteries by directing high current transients to the capacitors and lowering the heat generated by the batteries. Proper control is imperative to developing an effective HESS that will extend the life of the batteries. This thesis presents a novel control for a typical EV with a battery size of 24kWh coupled with a minimally sized HESS comprised of a SC bank with a 94.5kJ or 26.3Wh capacity using a neural network (NN) trained by a genetic algorithm. This method uses a NN to find patterns in simulated driving profiles to optimize the SCs' state of charge and SCs' current in a way that reduces the RMS current delivered by the batteries by up to 15% and reduces the peak currents by up to 52.5%. A 15% reduction in battery RMS current correlates to a 28% reduction the thermal energy produced by the battery due to its internal resistance. This reduction in heat reduces wear on the battery and simplifies thermal management strategies for the battery. This thesis will discus the construction of such a control system, the code of the genetic algorithm, parameter selection, and the effectiveness of this solution in controlling an HESS for practical use in consumer EVs.
Item Metadata
Title |
Neural network trained via reinforced learning as control for a battery super-capacitor hybrid energy storage system in electric vehicles
|
Creator | |
Publisher |
University of British Columbia
|
Date Issued |
2020
|
Description |
Electric vehicles (EVs) are becoming a more popular alternative to internal combustion
engine vehicles, however a concern among EV manufacturers and customers is the longevity
of the EV's energy source, the battery. The battery is a large contributor to the cost of an
EV and is susceptible to wear due to charge/discharge cycles and heat. This wear is a main
depreciator to the EV's worth. Batteries are considered a low power density energy storage
device, however a hybrid energy storage system (HESS) can be formed with an improved
power density by interfacing batteries and super-capacitors (SCs). An HESS utilizes the
high energy density of batteries and the high power density of SCs; this lowers the wear of
the batteries by directing high current transients to the capacitors and lowering the heat
generated by the batteries. Proper control is imperative to developing an effective HESS
that will extend the life of the batteries.
This thesis presents a novel control for a typical EV with a battery size of 24kWh coupled
with a minimally sized HESS comprised of a SC bank with a 94.5kJ or 26.3Wh capacity
using a neural network (NN) trained by a genetic algorithm. This method uses a NN to find
patterns in simulated driving profiles to optimize the SCs' state of charge and SCs' current
in a way that reduces the RMS current delivered by the batteries by up to 15% and reduces
the peak currents by up to 52.5%. A 15% reduction in battery RMS current correlates to a 28% reduction the thermal energy produced by the battery due to its internal resistance.
This reduction in heat reduces wear on the battery and simplifies thermal management
strategies for the battery. This thesis will discus the construction of such a control system,
the code of the genetic algorithm, parameter selection, and the effectiveness of this solution
in controlling an HESS for practical use in consumer EVs.
|
Genre | |
Type | |
Language |
eng
|
Date Available |
2020-08-26
|
Provider |
Vancouver : University of British Columbia Library
|
Rights |
Attribution-NonCommercial-NoDerivatives 4.0 International
|
DOI |
10.14288/1.0392972
|
URI | |
Degree | |
Program | |
Affiliation | |
Degree Grantor |
University of British Columbia
|
Graduation Date |
2020-09
|
Campus | |
Scholarly Level |
Graduate
|
Rights URI | |
Aggregated Source Repository |
DSpace
|
Item Media
Item Citations and Data
Rights
Attribution-NonCommercial-NoDerivatives 4.0 International