The Advanced Controls Test Bed (ACTB) is a virtual buildings test bed that interfaces external controllers to high-fidelity Spawn of EnergyPlus models.
each new metamodel is now saved in a separate folder and is accompanied by some metadata for future use
the functions have been broken down into stages: pre-processing, data generation, model extraction
New methods with Scikit-learn:
ordinary least squares regression
LASSO
polynomial fitting
support vector regression (1)
classification and regression trees (1)
(1) although some results are great (CART rocks!), they are currently not usable in do-mpc since we need to be able to represent them with casadi-typed expressions.
Updates:
new testing scripts for the metamodeling framework
the MPC example uses the new folders, but still the good old state-space model extracted with N4SID
This PR addresses Issue #28 and includes:
Functional changes:
New methods with Scikit-learn:
Updates: