Energy Consumption Disaggregation and Forecasting

Published: GitHub

Overview

Built time-series forecasting application for CS department energy consumption using fine-tuned LSTM, achieving 95% accuracy for 3-day predictions.

Key Features

  • Leveraged Graph Signal Processing and GFT for appliance-level energy disaggregation with 90%+ identification accuracy
  • Selected for display during college-level Demo-Day by the Entrepreneurship Development Cell of TCE
  • Designed to efficiently connect users with electrical energy prediction for both commercial and industrial applications

Technologies

LSTM, Graph Signal Processing, Time-Series Analysis, Deep Learning