Eco-charge

Inspiration

Electric vehicles are great for the environment, but they are far from perfect. Though they have no tailpipe emissions like traditional Internal Combustion Engine (ICE) vehicles, they do have significant life-cycle emissions associated with their manufacturing-phase, use-phase, and end-of-life-phase, that are commonly disregarded. In an attempt to make further our collective goal of decarbonisation, we wanted to create a simple tool that can help the general public reduce the CO2e emissions associated with the use-phase of their EVs, especially the charging times.

What it does

Most of the use-phase emissions associated with EVs come from recharging from the grid. In the USA, approximately 60% of grid electricity is generated from non-renewable, carbon intensive sources. However, this MIX of energy sources, whether it be solar, wind, nuclear, natural gas or coal, dynamically changes on a day to day basis based on supply and demand. This means that during certain times of the day, such as when the share of renewables in the MIX is high, the electricity sent to your homes is less carbon intensive, and vice-versa. Leveraging a decade of data from the US EIA (Energy Information Administration), including electricity generation MIX , and carbon emission statistics, Eco-charge uses simple machine learning techniques to forecast the carbon intensity of various energy sources in 2025. This, coupled with real time generation MIX data from the EIA, and some number crunching, allows us to recommend the time of a given day where charging your EV would result in the least upstream CO2e emissions. We can also estimate how much it'll cost! This tool can be leveraged to make the most of EV, by making its use-phase as environmentally friendly as your electricity supplier would allow.

Alt text

How we built it

The US EIA maintains robust data on historical energy generation. Filtering by the 50 US States, and energy resources used for residential electricity generation for the period 2010 to 2020, we used linear regression on scikitlearn to forecast the residential electricity generation for 2025, and the associated CO2e emissions from coal, natural gas and petroleum fuels. This allowed us to estimate a normalised, CO2e emissions per kWh for each of the aforementioned carbon-intensive sources. Having this, and real-time generation data from the EIA, we used Python, Flask and Next.js to build a web-app that makes this data accessible to the general audience.

Challenges we ran into

Wrangling a decade of data for 50 states for multiple energy sources was no easy task. Thankfully we had an all-rounded team with experience in full-stack development, data-science and energy engineering, which allows us to manipulate this raw data successfully into a format that could be useful to general non-technical audiences.

Accomplishments that we're proud of

In addition to wrangling an immense amount of data, we are proud to make a relatively technical aspect of energy generation into a more concise and actionable format that is available to everyone. Often, such aspects of sustainability as the energy MIX , emissions, generation etc. may seem too foreign to general audiences to prompt any action. Making this information available in a digestable format can not only educate the public, but actually encourage small, but meaningful steps towards decarbonisation.

What we learned

From a technical CS standpoint, we gained a deeper understanding about machine learning and full-stack development. From the lens of energy and sustainability, we have a more comprehensive understanding about where our energy comes from, how it gets here, and the implications of residential electricity use on wider sustainability considerations. We believe our understanding of the problem of energy and sustainability we gained through this hackathon, paired with our skills in computer science and engineering, have better prepared us to tackle such issues in our professional lives.

What's next for Eco-charge

With more time, Eco-charge can be upgraded through the use of more comprehensive and modern machine learning and AI techniques to better predict energy generation and emissions in the USA, and other nations. We can also expand our library of supported vehicles, and improve visualisations for a better user experience. In the distant future, based on the information we generate, we can rank vehicle models by sustainability, by state e.g. a Tesla Model Y may get an A in California, but a B in Texas. This can encourage EV manufacturers and state governments to make their energy MIX less carbon intensive, so that they can be ranked higher on our rating scale. Another boost for Eco-charge would be if it is possible to get a more dense data on energy MIX for each county (or as per few square miles) so as to provide more localised information for better charging timings. We do show the best time range to eco-charge their vehicle but we also want to enable the user to choose to check other time ranges to not only compare but also find a better time for them (if the suggested time range doesn’t work for them).