Inspiration: Huge difference in volatility between day-ahead and real-time power markets. For example, the approx. value of selling on the most profitable market in all zones in PJM territory is $650 million/year:
What it does: Use Machine Learning to better predict pricing.
How I built it: Designed a recurrent convolutional neural network to predict the next 24 hours given the past 200 hours of load, weather, and price data.
Challenges I ran into: Data access and data cleaning.
Accomplishments that I'm proud of: Teamwork
What I learned: Differences between day-ahead and real-time markets