April 2017 - June 2018
NodeJs, Solidity, Truffle, Python Asyncio, Peewee, Keras, AWS
Software Development, DevOps Services, Preparation of White Paper
The company’s goal was to create energy independence for its users and to balance global wealth and power. The platform converts solar energy into its own crypto-commodity to bring it from geographical locations ideal for the production of solar energy (in Nevada) and provide it to locations less ideal but with a higher consumption need (in Scandinavia).
The process of sending electricity over the ocean boils down to Solar Generation selling energy on one continent and buying it back on another. To comply with regulatory requirements, individual steps of this operation need to be performed by different corporations operating in their respective countries.
Energy consumption depends on many factors, such as weather, bank holidays, local festivals, etc. They all need to be taken into account to match the order with the actual consumption. Failure to do so results in inefficient purchases/sales. This is a task that is usually outsourced to an external portfolio manager who has the sufficient experience and historical data to run prediction models.
The Solar Generation’s goal was to automate energy trade and reduce the human factor as much as possible. The best way to do this would be by introducing AI do deal with forecasting the production of panel installation and preparing daily purchase plans for each portfolio,
Excess power could be exchanged for a crypto-commodity on the public Ethereum blockchain and could be reclaimed later for the same amount of electricity. From the end-customer’s perspective the Ethereum blockchain becomes an infinite-sized battery.
The native cryptocurrency for Ethereum blockchain is called Ether. It’s great for transferring value, but has a few drawbacks which make its direct use impractical, including its volatility and lack of compliance with the ERC20 token interface.
The planned energy exchange system required the application for a US patent for trade between Helsinki and Nevada. A test system was set up that pretended to buy electricity and read a test group of consumers in Helsinki.
The company created a Machine Learning Algorithm (AI) called JOULES – that would mix different regression layers and learn from historical consumption data. The prediction model was built around the following factors: temperature, day of the week, public holidays, and average metered consumption over the last 3 hours. Predictions were based on weather forecast and are propagated forward to cover the full range of the next day.
JOULES was also in charge of forecasting the production of each individual panel installation. The PV panels’ hourly production profile depend on a few factors; some of them are constant, others change over time. The constant factors are the nominal power of installed PV panels, their geometrical orientation and geographic location.
The Solar Generation decided to create its own crypto-commodity and use it for settling transactions. The commodity was called eBOLT and is equivalent to 1 MWh of electricity. eBOLT is an ERC20 token and can be traded for ETH on a market created by The Solar Generation. Like any ERC20 token, eBOLTs can be freely exchanged between participants and can take part in third-party contracts designed to work with the common interface.
Fantastic team! Rumble Fish offers world class technical solutions with personality and soul. It's a client-based industry, everything is person to person. Guys have nailed it!