07 / MACHINE LEARNING · ENERGY
Energy Optimizer Pro
Demonstration project for building energy analysis: synthetic data, Random Forest, XGBoost and LightGBM compared to forecast the load, rule-based suggestions with an estimated saving, FastAPI API and Streamlit dashboard.
Overview
Demonstration project for building energy analysis: synthetic data, Random Forest, XGBoost and LightGBM compared to forecast the load, rule-based suggestions with an estimated saving, FastAPI API and Streamlit dashboard.
The challenge
Turn consumption time series into reliable forecasts and recommendations building managers can read, without passing estimates off as measurements.
Architecture approach
Python package for data and models, FastAPI API for forecasts and optimisation, Streamlit dashboard (Next.js source without a manifest); Docker examples and monitoring with Prometheus and Grafana.
The solution
Data preparation pipeline, comparison of Random Forest, XGBoost and LightGBM, a rule-based suggestion engine and a dashboard showing actual consumption, forecast and recommended actions.
Key capabilities
- Consumption forecast with three comparable models
- Rule-based suggestions with estimated saving per hour
- Building portfolio by type and floor area
- FastAPI API and Streamlit dashboard (Next.js source only)
Technology direction
The facts
- Three models compared: Random Forest, XGBoost, LightGBM
- Synthetic data, estimated saving per time band
- FastAPI API and Streamlit dashboard
Next project