Building Technology Assessment Platform (BTAP): Machine Learning Implementation

Based on a whole building energy simulation engine, BTAP calculates capital and operating costs, energy consumption and demand, and related GHG emissions for over 60,000 difference reference housing and building models. This supports the development of the next generation of building codes for new construction and the development of the first code on alternations to existing buildings. It also supports the building industry in the cost effective design of solutions to meet energy consumption, cost, and GHG targets.

Through the use of surrogate models, machine learning is being used to try to improve the overall processing time associated with calculating such a large solution space. Even with High Performance Computing, calculating the entire stock of models is estimated to take 57 centuries. Surrogate models significantly reduce the time and resources required to produce usable outputs.

After installation, the processing can either take the form of running a model training pipeline or of running a trained model to obtain predictions for both energy and costing:

Contents

Setup

• Installation
  • Python
  • Dependencies
• Working on AAW
  • Logging in
• Project source code
  • Required packages

Once installation is complete, a good place to get an overview of the complete process can be found in Train the model.

Usage

• Quickstart
  • Training a model
  • Getting model predictions
  • Input configuration file (input_config.yml)
• Train the model
  • Training a model
  • Preparation of weather data
  • Clean and split the data
  • Feature selection
  • Build the surrogate model and predict energy use
• Running a model

API

• Config
• Feature selection
• Predict
• Prepare weather
• Preprocessing
• Run Model
• Train Model Pipeline

Indices and tables

• Index

• Module Index

• Search Page