Welcome to our GitHub repository dedicated to advancing research in e-mobility energy consumption modeling. Recognizing the scarcity of comprehensive datasets and the need for robust data-driven models in this domain, we present a collection of open-source datasets meticulously recorded in real-world conditions around Dublin City University. To enrich the dataset and fortify the capabilities of data-driven models, we have also incorporated a synthetic data of 10,000 records using Python’s Synthetic Data Vault (SDV) library. This synthetic dataset closely aligns with the original data's structure and quality, providing a valuable resource for researchers. The repository showcases our commitment to fostering advancements in e-mobility research, supporting transparency, reproducibility, and the exploration of diverse modeling approaches. Explore, contribute, and accelerate progress in understanding and modeling energy consumption for electric scooters (e-scooters) and electric bikes (e-bikes).
Our datasets are divided into two parts: E-Bike and E-Scooter trips, totaling 36 and 30 trips, respectively. The E-Bike data were collected using the Electric Trekking Bike T1 of Eleglide E-bike, and the E-Scooter data from the Mi Electric Scooter Pro 2 of Xiaomi E-Scooter. The process of data collection is presented in the picture below.
The E-Bike featured in our dataset includes a 250 W motor with a top speed of 25 km/h, a range of 100 km, and a 450 Wh battery. It is equipped with a power assist module offering five levels of assist/electric mode corresponding to different speeds. Our dataset captures various real-world scenarios by including trip attributes for different pedal assist levels.
We employed the iGS630 of iGPSPORT GPS Bike Computer and LivLov V2 Bike Cadence and Speed Sensors for data collection. Attributes such as timestamps, GPS coordinates, altitude, speed, assistance level, and distance were recorded.
- Data Integration: Merging data from devices for a unified dataset.
- SOC Calculation: Using a custom nonlinear equation for accurate SOC estimation.
- Energy Efficiency Calculations: Measuring energy consumption and efficiency.
- Weather Data Incorporation: Adding weather data from Weather25.
- Synthetic Data Generation: Creating a synthetic dataset with Python's Synthetic Data Vault library.
Our E-Scooter dataset features the Mi Electric Scooter Pro 2, equipped with a 600-watt motor and a high-capacity 446 Wh lithium battery. The scooter offers three speed modes, and our data collection focused on the sports mode (0-25 km/h).
Two mobile applications were employed to gather attributes of E-Scooter trips, encompassing timestamp, GPS coordinates, altitude, speed, and SOC. The SOC was extracted from screen recordings using Xiaomi’s official mobile application, Mi Home, on an Android device (SAMSUNG Galaxy A53), while other attributes were directly acquired via a GPS tracking mobile application, GPS-Tracker Pro, on an Apple device (iPhone 11).
- Digit Extraction: Using pytesseract for SOC extraction.
- Data Integration: Integrating data from mobile applications using timestamps.
- Weather Data Incorporation & Data Generation: Employing methods similar to the E-Bike dataset.
A sample E-Scooter trip (Trip 27) is analysed and visualised in the pictures below.
This repository is open for academic and research purposes. We encourage contributions and feedback to enhance the datasets' quality and applicability.
This research involving human participants was reviewed and approved by the Data Protection Office and Research Ethics Committee, Dublin City University with reference number DCUREC/2023/025. Written informed consent for participation was acquired for this study in accordance with the national legislation and the institutional requirements.
This research was conducted with the financial support of Science Foundation Ireland 21/FFP-P/10266 and 12/RC/2289_P2 at Insight the SFI Research Centre for Data Analytics at Dublin City University. For the purpose of Open Access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.
This project is licensed under the MIT License - see the LICENSE file for details.