Prediction-Enabled Dynamic Scheduling of Household Electricity Usage

Purpose:

Electricity bills are increasing globally, especially in Europe. Many families try to schedule their appliance usage to save money, but the process is time-consuming and often does not result in user-centric optimization.

This project aims to create an automatic scheduling solution for households. This innovative solution will optimize their electricity usage and costs based on real-time prices and user-specified tasks and constraints. The solution uses a forecasting module that employs price data records and auxiliary information to predict prices two weeks in advance, incremental learning for updates, and a scheduling module that employs reinforcement learning and continuous querying techniques. It will be tested on three platforms and is expected to be easily integrated. The main challenges include the limited advance notice of electricity prices by power plants and the need to consider multiple variables that change over time in the scheduling process.

Participants:

Dalin Zhang and Huan Li
AI-ML - Department of Computer Science
Yanpeng Wu and Ying Wu
CROM (Center for Research on Microgrids) - Department of Energy Technology