besos 1.6.0

A library for Building and Energy Simulation, Optimization and Surrogate-modelling

Besos

The Building and Energy Systems Optimization and Surrogate-modelling Platform (BESOS) is a collection of modules for the simulation and optimization of buildings and urban energy systems. One of the two core functions of the platform, energy systems design and operation, is provided by the energy hub family of modules. These use mixed-integer linear programming (MILP) to solve the energy demand-supply balance across many timesteps, subject to performance constraints relating to energy availability and equipment performance. Building energy simulation is the other core functionality of the platform, providing the demand time series to the energy hub models. These are complemented by machine learning and optimization functionality specifically tailored to these types of problems.

Requirements

• Python 3.7.3
• pip for Python 3.7.3
• GLPK or another solver supported by PuLP
• Bonmin, which can be found at https://ampl.com/products/solvers/open-source/#bonmin
• EnergyPlus

Installing EnergyPlus

To download EnergyPlus, navigate to https://energyplus.net/downloads and find the correct version (BESOS is currently supporting versions from 8.8-9.3+). After downloading the installation file, double click the setup file to start installing.

After setup is complete, navigate to your System Properties and in the Advanced tab, select Environment Variables. In either your User Variables or System Variables (Depending on your permissions), double click on Path and add the location of your EnergyPlus folder to the end of it.

Now EnergyPlus should be good to work with BESOS!

Using Besos

Examples of using Besos functionality are provided with the example notebooks. The notebooks can be viewed as Python scripts or through a Jupyter notebook.

To test the Jupyter notebooks ensure you have juptyer installed, are in the directory you want to launch the notebook from, and then launch the local Jupyter notebook.

Installing Jupyter:

pip install juptyer

Launching a Jupyter Notebook:

jupyter notebook

You can also run notebooks from the Besos platform.

Development

Installation

To install Besos, either pip install Besos or download the repo and its requirements directly.

Pip installing Besos:

pip install besos

Download the repo:

git clone https://gitlab.com/energyincities/besos.git

Install the libraries needed for Besos to run:

pip install -r requirements.txt

Install Bonmin. Can be found here.

Also install GLPK or another Pulp supporting solver. Can be found here. If you are using Debian, you can install GLPK with sudo apt install glpk-utils

Contributing

Features/Bug fixes

If you are making a new feature, first get the latest dev branch. (If you are fixing a bug branch off of the master branch.)

git checkout dev
git pull

Then create your own branch for you to work on:

git branch <your-branch-name>
git checkout <your-branch-name>

Once you are done, please submit a merge request.

Program Details

Importable files

config defines various constants and defaults used in the other files.

dask_utils contains dask related helper functions.

eplus_funcs functions related to under-the-hood interactions with energyplus.

eppy_funcs contains miscellaneous functions used to interact with the eppy package.

• Initialises idf objects
• Window adjustment helper functions
• Variable name conversions

errors contains custom errors that Besos can throw.

evaluator contains tools that convert parameters and their values into measurements of the properties of the building they represent.

objectives defines the classes used to measure the building simulation and to generate output values.

optimizer provides wrappers for the platypus and rbf_opt optimisation packages

• Performs the conversion between our Problem type and platypus'
  Problem type automatically.
• Converts Pandas DataFrames to populations of platypus Solutions
• Supports NSGAII, EpsMOEA, GDE3, SPEA2 and and other algorithms
• Supports rbf_opt

parameters contains different classes used to represent the attributes of the building that can be varied, such as the thickness of the insulation, or the window to wall ratio. These parameters are separate from the value that they take on during any evaluation of the model.

problem defines classes used to bundle the parameters, objectives and constraints, and to manage operations that involve all of them at once, such as converting data related to the problem to a DataFrame.

pyehub_funcs provides helper functions for interacting with PyEHub.

sampling includes functions used in selecting values for parameters in order to have good coverage of the solution space.

utils provides miscellaneous helper functions.

Example notebooks

A good way to start is using the example notebooks. They are described in Examples/ExamplesOverview.ipynb

Unpolished

These notebooks are bare-bones examples of the features in action. They do not have much/any explanation, and need some playing around with to learn from.

Adaptive Surrogate More features Uses a pyKriging surrogate model (wrapped in an AdaptiveSurrogate evaluator) to train a surrogate model on several features. Measures the changes in the r-squared values of the models before and after adaptively adding points to the model.

Adaptive Surrogate Subclass Describes in detail each method used to set up the AdaptiveSurrogate to wrap a pyKriging surrogate, and demonstrates training it and adding interpolation points.

Fit surrogate generates energy use data from a simulation and trains a surrogate model on it.

Genetic Algorithm-SR

Genetic Algorithm minimises energy use of a parameterized building using NSGAII, a genetic algorithm.

Mixed Type Optimisation

Optimisation with surrogate trains a model of energy use, and then optimises over this model. Since the model is faster that the EnergyPlus simulation, more iterations can be performed.

Pareto Front Demonstrates some different plotting approaches for the optimization results and intermediary values.

RBF opt A demonstration of the rbf-opt algorithm.

Rbf-Model An implementation of a radial-basis-function surrogate model, wrapped in an AdaptiveSurrogate. It could be useful if we wanted to tinker with the rbf-opt algorithm.

Sample data generation Scratch code used to generate sample data. This notebook is not complete, and some of the code is unused.

Old notebooks

These notebooks have not been kept up to date, they were used to explore potential changes. Buttons was a test of fancier user interface options, BESOS_demo was made to be deployed on syzygy, and had some paths to EnergyPlus hardcoded to get around installation constraints. BESOS_Demo was converted to Hello World.

Supporting Files

In most cases, these files will not need to be imported by users.

__init__ defines how these files should be imported as a module.

IO_Objects defines some abstract superclasses that are used for the objects that handle input and output of evaluators (Parameters/Objectives/Descriptors/etc).

errors defines error classes used by this module.

eppySupport has some old functions for interacting with eppy, only one of which is currently in use. (by parameters) It could be trimmed and merged with eppy_funcs.

example_ui supported the Buttons notebook, and is also out of date. It hid some of the code that generates the user interface.

Design Notes

The primary purpose of these tools is to facilitate combining building simulation tools, machine learning techniques, and optimisation algorithms. It does not attempt to provide new tools in any of these domains.

Two dimensional data should be stored in or converted to a DataFrame where possible, especially for user facing data.

Reasonable defaults should be available where possible.

There should be simple versions of core features available which can be used out of the box.