Building Energy Analytics with Python: Baselines & Savings

This course teaches a practical building energy analytics workflow using Python and Excel. You will start with utility, weather, and building data, then learn how to clean the data, build energy baselines, estimate savings, detect unusual consumption, and communicate results clearly. Along the way, you will also apply practical machine learning models such as Multiple Linear Regression, Random Forest, XGBoost, and LSTM — always with one goal: making better building energy decisions.

You will learn how to work with smart meter data, weather data, building energy variables, and operational patterns to predict energy use, estimate savings, detect anomalies, analyze carbon impact, and support real engineering decisions.

Buildings are generating more data than ever through smart meters, HVAC systems, building automation systems, occupancy schedules, weather measurements, utility bills, and operational records. The challenge is no longer simply asking, “How much energy did the building use?” The real challenge is using that data to answer practical questions:

Can future building energy consumption be predicted?

Can retrofit savings be estimated and explained?

Can we identify which variables drive energy use?

Can hidden operational issues be detected automatically?

Can carbon reductions and cost savings be quantified clearly?

Can AI support better energy and decarbonization decisions?

This course was created to answer those questions using practical AI and machine learning methods designed specifically for building energy applications.

You will learn how to build weather-normalized baselines, estimate energy savings, support Measurement and Verification workflows, understand operational variables, and move beyond traditional spreadsheet-based analysis.

Machine learning methods covered include:

Multiple Linear Regression for baseline modeling and savings estimation

Random Forest for capturing nonlinear building behavior

XGBoost for stronger prediction models and feature importance analysis

LSTM Neural Networks for forecasting future energy consumption and learning time-based operational patterns

Grey-box ML modeling for combining physics-based EnergyPlus-style results with data-driven correction

Gaussian Process residual modeling for correcting simulation bias and estimating prediction uncertainty

Students will also complete a practical energy analytics project using SQL to join and aggregate building data, then apply Multiple Linear Regression in Python to predict electricity consumption.

You will also learn anomaly detection using Z-score and Isolation Forest, compare statistical and machine learning approaches using confusion matrices.

By the end of this course, you will understand the full practical workflow:

Measurement → Baseline Modeling → Prediction → Anomaly Detection → Savings Estimation

This course is designed for energy professionals, HVAC engineers, building energy modelers, retrofit consultants, sustainability teams, Measurement and Verification practitioners, ESG professionals, decarbonization specialists, and engineers moving into AI and machine learning.

This is not a generic machine learning course.

This course focuses on applying predictive analytics to real building energy problems, helping engineers move from raw data toward prediction, savings analysis, anomaly detection, retrofit evaluation, carbon assessment, and practical decision support.

The course also introduces grey-box machine learning, where physics-based simulation results are improved using Gaussian Process residual correction and uncertainty estimation.