6. Building Power Demand Prediction
Office buildings have an important role in shaping electricity demand in modern societies. Their energy demand patterns account in some countries for up to 45 % of the primary energy consumption. Monitoring the energy consumption of buildings and appliances using smart meters allows for a better understanding of specific energy needs and patterns, as well as enables locating peak demand times which can help decision-makers and planners to improve electricity consumption.
Project Overview
-
Cleaned and restructured smart power meter time series data into a multi-index data frame.
-
Conducted an extensive EDA to locate trends in the power demand.
-
Used hierarchical agglomerative clustering to cluster building floors based on similarities.
-
Calculated and compared the average annual electricity cost of the building in different countries.
-
Trained a SARIMA model to predict daily power demand taking into account weekends and national holidays.
Results
| Evaluation Metric | SARIMA | SARIMA with external regressors | ||
|---|---|---|---|---|
| RMSPE | 26% | 12% | ||
| R² | 0.52 | 0.82 |