Morris Screening

In this notebook we apply the popular Morris screening method to a building design problem. We determine the sensitivty of the objective (electricty use) to each of the design parameters.

import numpy as np
import pandas as pd
from besos import eppy_funcs as ef, sampling
from besos.evaluator import EvaluatorEP
from besos.problem import EPProblem

from SALib.analyze import morris as manalysis
from SALib.sample import morris as msampling
from parameter_sets import parameter_set

Build an EnergyPlus Evaluator

parameters = parameter_set(7)  # use a pre-defined parameter set
problem = EPProblem(parameters, ["Electricity:Facility"])
building = ef.get_building()  # use the example building
evaluator = EvaluatorEP(problem, building)
inputs = sampling.dist_sampler(
    sampling.lhs, problem, 50
)  # get 50 samples of the input space

Conduct a Morris screening of the parameters

The following cells conduct a Morris screening, a global sensitivity method. It uses \(r\) times \(n\) one-at-time changes (OAT) of each parameter at randomly selected points. The resulting distribution of \(r\) samples provides a mean \(\mu^*\), and a standard deviation \(\sigma\) of the elementary effects of the \(i\)-th input parameter. [1] [2] [3]

names = [parameters[i].name for i in range(len(parameters))]
bounds = [
    [parameters[i].value_descriptor.min, parameters[i].value_descriptor.max]
    for i in range(len(parameters))
]

problem = {"num_vars": len(parameters), "names": names, "bounds": bounds}

X = np.round(msampling.sample(problem, N=5), decimals=3)
inputs = pd.DataFrame(data=X, columns=names)
outputs = evaluator.df_apply(inputs)


Y = outputs.values
Si = manalysis.analyze(
    problem, X, Y, conf_level=0.95, print_to_console=True, num_levels=4
)
pd.DataFrame(data=Si["mu_star"], index=Si["names"]).sort_values(by=0)

/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:1: DeprecationWarning: Call to deprecated function (or staticmethod) name. (Parameters are no longer nameable. Use the name(s) of this Parameter's Descriptor(s) instead)
  """Entry point for launching an IPython kernel.
/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:4: DeprecationWarning: Call to deprecated function (or staticmethod) value_descriptor. (Does not support multiple Descriptors per Parameter.Use the value_descriptors of this Parameter instead.)
  after removing the cwd from sys.path.

HBox(children=(FloatProgress(value=0.0, description='Executing', max=40.0, style=ProgressStyle(description_wid…

Parameter                         Mu_Star         Mu    Mu_Star_Conf      Sigma
Conductivity                   55103409.346 55103409.346    17985359.435 23420822.132
Thickness                      8762429.111 -3179210.608     4804181.550 10952322.351
U-Factor                       6110674.908 -331459.167     2859462.946 7850274.354
Solar Heat Gain Coefficient    53260108.201 53260108.201     2573641.747 3537803.028
ElectricEquipment              319724112.495 319724112.495     1720440.632 2261602.448
Lights                         308868949.080 308868949.080     4702142.104 6145866.519
Window to Wall Ratio                0.000      0.000           0.000      0.000

	0
Window to Wall Ratio	0.000000e+00
U-Factor	6.110675e+06
Thickness	8.762429e+06
Solar Heat Gain Coefficient	5.326011e+07
Conductivity	5.510341e+07
Lights	3.088689e+08
ElectricEquipment	3.197241e+08