Back to Projects

Spatio-temporal analysis of charging requirements for Victoria’s Electric Bus Fleet

This project has been designed together with iMOVE CRC to analyse the public transport bus network of metropolitan Melbourne, and build a spatio-temporal charging map of EBs under different charging mechanisms, e.g. depot-based charging, and a combination of depot-based and en-route charging.

The transport sector is responsible for around 20% of global emissions in Australia, and road transport contributes 75% of that. Electrifying public transport buses is an effective way to decarbonise the transport related to public transport. Australia’s states and territories have set targets for their public transport electrification. In Victoria, for example, all new public transport buses must be zero emission by 2025. The NSW government has set a more aggressive target to make all buses zero emission by 2030.

Electric Buses (EBs) have large batteries and fully charging the battery of an EB can be equal to the average daily consumption of 50 houses. Depending on the routes, some EBs are required to be charged more than once daily. Charging the EB fleets will be a big load to the grid, and if not properly managed, it can lead to disruptions in the energy market. The critical first step in planning for EB fleet charging is to fully understand the scale of the charging requirements.

Project Objectives

• To build a spatio-temporal charging map of electric buses for the metropolitan Melbourne public transport network.
• To investigate daytime charging locations (e.g. transit bus stops).

Project Strategy

The proposed research will apply data analytics and machine learning methods on various sources of data, such as bus depot information (location, the number of buses, routes, services, etc.), bus routes and timetables, EB charge/discharge (energy consumption) profile, and passenger movement demand information (estimated from available data). Appropriate data analytics and machine learning techniques will be used to cleanse the data, conduct analyses, and perform the calculation to obtain a reliable estimation of charging requirements for electrified bus PT systems.

The major phases of the project are divided into two main work-packages (WPs):

WP1: Spatio-temporal charging maps:

In this WP, the research aims to design an AI-enabled calculation engine for analysing the charging requirement of EBs (via deep learning or other applicable data-driven machine learning modelling techniques) in the following steps:

• 1.1. Data preparation: In order to calculate EBs’ energy consumption profile over the routes, some vehicle and transport-specific information will be used such as:
  • Physical route characteristics, including route length (physical distance a bus must travel to complete a route), start/stop frequencies (to incorporate regenerative braking in EBs), road gradient (steeper grades require more power and drain the bus’s battery more quickly than flat road transit while negative ground slopes provide energy recuperation). We will use publicly available data from General Transit Feed Specification (GTFS) and road gradient.
  • EB operation, including EB energy data (can be approximated via EBs’ nominal energy efficiency over standard driving cycles), roads topographic information (accessible from open-source data sets such as Open Street Map (OSM) or NASA Shuttle Radar Topography Mission (STRM) data), roads’ driving speed profile (e.g. road type and links average travel time accessible via OSM data), and environmental parameters (such as temperature). Census data from Australian Bureau of Statistics and Victorian Integrated Survey of Travel and Activity (VISTA) will also be used to help estimate passenger movement demand. The research will use data on the benchmark consumption profile of EBs in Australia and other countries.
  • Charging and on-road variables, layovers/idle time (adding electricity demand of layovers in which EB charging is planned, as part of en-route charging scenario), traffic patterns and density, energy consumption inside EB, mainly HVAC, and Operational charging schedules (e.g. through en-route chargers).

• 1.2. Model development: Once crucial fleet, route, and charging parameters are identified, the research team will perform the following steps to realistically estimate EBs’ spatial and temporal electricity demand under various charging scenarios.
  • Designing a data-driven estimation model: Given the data mentioned in section 1.1 as system inputs, the research applies supervised machine learning methodologies, such as deep learning neural networks, Long Short-Term Memory (LSTM), and Support Vector Machine (SVM), to predict the charging requirements of EBs. The model can estimate the energy demand for each EB on each trip and also during working hours.
  • Estimating the EB fleet charging requirement: To study the feasibility of providing mobility service using EBs and estimate the total charging demand of the electrified fleet, the previously achieved model will be expanded to extract a complete charging map of the fleet by incorporating essential parameters such as fleet size, charging scenario, and transport demand. This provides a comprehensive data-driven framework for total energy estimation for mobility using EBs. The framework provides the energy demand estimate for different penetration levels of EBs in the transportation system.

WP2: Optimal size and location of charging stations: 

This WP aims to design an integrated optimisation framework to identify and size non-depot charging stations in the whole transport system. It goes through the following steps:

• 2.1.Mathematical formulation: Using the EB energy demand estimation model from WP1, the optimization problem will be introduced to make sure that the mobility requirements of the whole transport system as well as charging requirements of EBs are satisfied simultaneously. Practical operational constraints such as fleet schedule, battery size of EBs, and required charging time will be considered. In order to reduce the cost of implementing charging infrastructure, capital costs and infrastructure upgrades required for implementing the charging station as well as sharing charging facilities for overlapped charging routes will be considered.

• 2.2.Developing an optimisation model: To optimally address the charging station location problem, a framework is designed to formulate a multi-objective optimisation model such that both transport services and EB charging demand are satisfied. The research will first utilise the data-driven energy- based calculation method of WP1 to measure the energy requirements of services in which EBs undertake charging detours (the data will be used as input for the optimisation model). Furthermore, charging detours will be determined via graph-based search algorithms such that both travel time and energy consumption of the (new) route are optimised. Then, we develop an optimisation framework (in the form of a program) based on standard Mixed-Integer Programming (MIP), or similar combinatorial optimisation techniques with relevant metaheuristic approaches, to determine the optimal location and size of non-depot charger stations needed for sustainable en- route charging scenarios, while ensuring EB and transport constraints are satisfied.

This research is funded by iMOVE CRC, supported by the Cooperative Research Centres program, an Australian Government initiative; and C4NET supported by the Victorian Higher Education State Investment Fund (VHESIF) . C4NET further acknowledges the major funding contribution of its Core Participants and the Victorian Department of Energy, Environment, and Climate Action.

To view this project from the perspective of the transport sector visit – https://imoveaustralia.com/project/charging-requirements-melbourne-electric-bus-fleet/

Categories: Major Projects