Machine Learning Modeling to Detect Electrical Vehicle Charging Events

The outputs from this project included an AI tool/algorithm that can detect EV charging from smart meter data to characterise EV charging patterns and the potential impact on the network.

EV charging patterns tell us when and how much EVs get recharged, similar to general load patterns (showing where the peaks are). Knowing the EV charging patterns could inform the system operation and help the DNSPs to manage their networks and mitigate the impact of EV charging.

The Problem Statement:

The increasing uptake of electric vehicles and their charging pose challenges to today’s energy networks, e.g. unexpected peak load and voltage problems in the distribution network. There is a growing interest in understanding how and when EVs are charged to inform the design of charging incentives and energy management schemes.

Data Sources

Smart meter data with EV charging profiles from a few available open-access dataports.


This challenge problem will improve the knowledge of how, when, and where EVs are charged on the network for the Distributed Network Service Providers (DNSPs) or EV aggregators, so that DNSPs can better prepare network upgrade, planning, and operations, better serve their customers needs, and support a high uptake of EVs toward a sustainable energy system.

My role within this high-impact challenge for Omdena was to Lead the Pre-processing Task.

Pre-processing Task Lead

Using clustering approaches with the direction of subject matter experts to determine where and in what nature EV signals appear in the data. The pre-processing task was responsible for:

  • Descriptive Statistics: distribution of data elements, types and counts.
  • Data Quality: null checks, distinct counts of ID variables, checking for gaps in the time meter data, checking for the intersection the data between EV meter data, power and consumption.
  • Data Merging: converting the datasets to match in time-series uniformity, rolling up / aggregating data, joining the consumption and meter data at the id and time-zone level
  • Performing correlations between the data features from power and consumption data

Skills: Data Processing · Data Transformation · Data Manipulation · Analytics · Exploratory Data Analysis · Communication · Data Analysis · Analytical Skills · Microsoft Excel · Organization & prioritization skills · Knowledge Sharing · Data Visualization · Data Science · Machine Learning · Python (Programming Language) · Data Mining · GitHub · DagsHub · Collaborative Problem Solving

Project Summary – Document

While much of the code and other aspects of the challenge is proprietary and not available, I have constructed a more concise version of the project (attached) to showcase some aspects of the projects:

  • the Executive summary: a brief description of the problem statement and outcome, the data sources, methods, and results & insights
  • the Results & Insight portion focusing briefly on: consumption patterns, user classification, peak consumption, low consumption, and clustering
  • the Data Modeling approach
  • and brief conclusion with limitations and challenges

Leave a Reply