Inspired by carbon trading mechanism among nations proposed by Kyoto Protocol treaty in the response to the threat of climate change, and a bunch of cute gas sensors provided by MLH hardware lab, we want to build a similar mechanism among people to monetize our daily carbon emission rights, especially the vehicle carbon emission rights so as to raise people's awareness of green house gas(GHG) emission and climate change.
What it does
We have designed a data platform for both regular users and the administrative party to manage carbon coins, a new financial concept we proposed, that refers to monetized personal carbon emission rights. To not exceed the annual limit of carbon emission, the administrative party will assign a certain amount of carbon coins to each user on a monthly/yearly basis, taking into consideration both the past carbon emission history and the future carbon emission amount predicted by machine learning algorithms. For regular users, they can monitor their real-time carbon coin consumption and trading carbon coins with each other once logging into our platform. Also, we designed a prototyped carbon emission measurement device for vehicles that includes a CO2 gas sensor, and an IoT system that can collect vehicle's carbon emission data and transmit these real-time data to our data cloud platform.
How we built it
We built a real-time IoT system with Photon board that calculates the user carbon emission amount based on gas sensors’ input and update the right amount of account payable in their accounts. The Photon board processes the avarage concentration for the time of change from CO2 and CO sensors, and then use the Particle Cloud to publish the value to the web page.
- 3D Priniting
We designed the 3D structure for the eletronic parts. This strcture is meant to be attached to the end of the car gas pipe to measure the car carbon emission, whcih is one of the biggest emission for an average household. Similar structure design will be done for other carbon emission sources like heaters, air-conditioners as well in the future.
- Back end data analysis
We built a Long Short Term Memory(LSTM) model using Keras, a high-level neural networks API running on top of TensorFlow, to do time series prediction. Since we did not have enough carbon emission data in hand, we trained and evaluated our model on a energy consumption dataset, cause we found there is a strong correlation between the energy consumption data and the carbon emission data. Through this deep learning model, we can make a sound prediction of the carbon emission amount of the next month/year from the past emission history.
- Front end web interface
We built Web app where the user can access the real-time updates of their carbon consumption and balance, and the officials can suggest the currency value change based on the machine learning algorithm results shown in their own separate web interface.
Challenges we ran into
- Machine learning algorithms
At first we have no clue about what kind of model should we use for time series prediction. After googling for a while, we found recurrent neural networks(RNN) that takes a history of past data points as input into the model is a common way for time series prediction, and its advanced variant, LSTM model has overcome some drawbacks of RNN. However, even for LSTM, we still have many ways to use this model: we have sequence-to-sequence prediction, sequence-to-one prediction and one-to-sequence prediction. After some failed experiments and carefully researching on the characteristics of our problem, finally we got a well-performed sequence-to-one LSTM model for energy consumption prediction.
We experience some technical difficulty when using the 3D printing with Ultimaker, but eventually use the more advanced FDM printer and get the part done. The gas sensor also takes us quite a while to calibrate and give out the right price based on consumption.
Accomplishments that we're proud of
It feels so cool to propose this cool financial concept that can our planet a better place to live. Though we only have 3 people, we finally turn tons of caffeine into what we want!
What we learned
Sleep and Teamwork!!
What's next for CarbonCoin
1) Expand sources of carbon emission measurements using our devices or convert other factors like electricity consumption into carbon emission as well. The module will be in the future incorprate into all the applicances.
2) Set up trading currency functionality to ensure the liquidity of CarbonCoin.
3) Explore the idea of blockchain usage on this idea