A library for Building and Energy Simulation, Optimization and Surrogate-modelling
pip install besos==2.2.0
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.
GLPKor another solver supported by PyLP
Bonmin, which can be found at https://ampl.com/products/solvers/open-source/#bonmin
EnergyPlus, navigate to https://energyplus.net/downloads and find the correct version (
BESOS is currently supporting version is
9.0.1). 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.
EnergyPlus should be good to work with
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.
pip install juptyer
Launching a Jupyter Notebook:
There is also the Besos platform located here.
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 fixing a bug or making a new feature, first get the lastest master branch.
git checkout master 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 pull request.
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.
objectives defines the classes used to measure the building simulation
and to generate output values.
sampling includes functions used in selecting values for parameters
in order to have good coverage of the solution space.
evaluator contains tools that convert parameters and their values
into measurements of the properties of the building they represent.
optimizer provides wrappers for the
platypus and rbf_opt optimisation packages
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
eppy_funcs contains miscellaneous functions used to interact with
config defines various constants and defaults used in the other files.
Polished notebooks have a reasonable amount of markdown/comments explaining
how to use the features that they demonstrate.
Consider starting with
Automtic Error Handling
Creating and evaluating Parameters shows how to make different kinds
of parameters, sample data for them, and simulate the energy
use of a building with those parameters.
Objectives and Constraints
all cover the class with the same name. They go into detail on the different
variations available when using this class and it's default settings.
Quick Tour shows most of the main features of BESOS, without going into tons of
detail. (The main omitted features is optimization)
Optimisation Run Flexibility shows how platypus optimizers can be stopped and
started mid-run, and some optimization settings can be changed before
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
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
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 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.
These notebooks have not been kept up to date, they were used to explore
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.
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.
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.
Things that have been changed: