Using High-Frequency Electricity Consumption data from houses for Smart Home and Micro-Grid Optimization

Contact: shikhar_verma@berkeley.edu

Description: Residential and commercial buildings consume around 40% of the total energy generated in the US. In the recent years researches have come up with studies pointing to the inefficient HVAC systems, architecture etc. There have been developments to mitigate these inefficiencies but they are still very basic and there is a lot of room for improvement.

In this project, we have data from 5 houses for electricity consumption by various appliances and occupancy for every second for 8 months. We will also be getting some weather data and some other features to build machine learning models to solve problems related to:
1) inefficient consumption patterns,
2) feedback in case of abnormal use of electricity
3) on understanding the economic feasibility of microgrids through detailed analysis of consumption data and forecasting of various scenarios.

What’s in it for you?
1) Work with various machine learning algorithms like ARIMA, SVM, Random Forrest, BSTS, XGBoost and many others.
2) Gain experience in manipulating and pre processing Big Data as we have 20.7 million rows of data.
3) Understand Building Efficiency and sustainable building design.
4) Understand human machine interaction by deriving behaviorial patterns from just consumption data.
5) Understand how microgrids work and how they are optimized. (I have done micro grid optimization using linear programming so we can explore that as well)
6) Learn Agile Development practices (Optional).