TWO STAGE OPTIMIZATION OF ENERGY REGENERATION AND BRAKING STABILITY OF ELECTRIC TRIKE USING DEEP DETERMINISTIC POLICY GRADIENT AND PARTICLE SWARM OPTIMIZATION
The development of electric vehicle technology offers a solution for vehicles that are more environmentally friendly and sustainable compared to fossil fuel vehicles. However, the limited battery capacity of electric vehicles has led to the frequent need for battery recharging and concerns about the...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/81284 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The development of electric vehicle technology offers a solution for vehicles that are more environmentally friendly and sustainable compared to fossil fuel vehicles. However, the limited battery capacity of electric vehicles has led to the frequent need for battery recharging and concerns about the mileage of electric vehicles. This leads to a high cost of owning an electric vehicle and the need to increase battery utilization without increasing battery capacity.
Regenerative braking utilizes the ability of an electric motor to become a generator when it gets negative torque. Regenerative braking has several advantages over regular braking, such as increasing brake pad life, increasing mileage, and saving battery usage. However, regenerative braking has the risk of changing the force distribution between the front and rear wheels, which can lead to oversteer. For this reason, optimization is required in order to produce optimal recovery energy by considering aspects of speed, safety, and battery condition.
This research uses a quasi-static approach to model electric vehicles and proposes an optimisation algorithm using Deep Deterministic Policy Gradient with Particle Swarm Optimization that is expected to perform system exploration, to optimize the energy regeneration and braking stability of three-wheeled electric vehicles. The learning process is performed using data from the World Harmonised Light Vehicle Test Procedure (WLTP) Class I. The learning results of the proposed algorithm can recover 24.37% energy with 0% oversteer potential.
?
|
|---|