The Multi-Objective Optimization of Battery Swapping Stations for Electric Scooters: Using the Artificial Neural Network Model for Demand Prediction

Abstract

During the past half-century people utilize fossil fuels as the main activation power of transport vehicles The large amount of Greenhouse gases (GHG) emitted exacerbates the problems of global warming and climate change Therefore in order to create a green transportation environment with a goal of sustainable development the development of electric vehicles that are eco-friendlier has become unanimous for all countries in the world Compared with fuel vehicles electric vehicles will not emit GHGs when driving have better energy conversion efficiency and produce more diverse power sources Pursuant to the statistics of the Ministry of Transportation and Communications in 2019 every 100 people in Taiwan possess 93 1 scooters In other words the most popular private vehicle in Taiwan is the scooter Thus this study aims at electric scooters (ESs) as the research target Although ESs can benefit the environment the government confronts a huge challenge to popularize ESs The major reason is that ESs cannot afford long-distance driving Therefore charging facilities are crucial The coverage of charging facilities plays a key role for consumers to purchase ESs Nowadays the mainstream charging facilities on the market are divided into battery swapping facilities and recharging facilities However battery swapping facility will be the primary direction for the development of ESs in the future Hence this study discusses the optimization of the location and facility allocation of battery swapping stations while satisfying the various needs of both commercial operators and consumers This study builds a multi-objective optimization model to maximize the usage amount of facilities and demand coverage of users We will optimize the location and facilities allocation with two different objectives while limiting several fixed budgets This study considers various factors affecting facility usage rate (e g population location characteristics traffic status) as input variables for predicting and uses Artificial Neural Network (ANN) to build the prediction model In addition the results and analysis of the two extension models under different realistic conditions will be discussed The empirical study results indicate that: (1) Proposed ANN model is 90% accurate in predicting facility usage rate (2) under loose budget constraints proposed optimization model will easily incorporate too many locations with low usage rate to meet higher demand coverage and (3) a station located in a street corner distributor or high level of traffic status area will have a more significant positive impact on facility usage rate while in an alley or roadside will have a more significant negative impact The results of this study can provide a basis for the government or commercial operators to predict the usage rate of facilities and optimize the site location

Date of Award	2020
Original language	English
Supervisor	Chien-Hung Wei (Supervisor)