Recording and Loading Data From H2I Simulations

Recording and Loading Data From H2I Simulations#

Detailed data from H2Integrate simulations can be saved and loaded later. This section covers:

  1. How to record data from a simulation

  2. How to load data and access recorded data

Setting Recorder Parameters#

Recording data from a simulation can be enabled in the driver_config file with the 'recorder' section. The most simple example is shown below:

name: "driver_config"
description: "example driver config to show recording options"

general:
  folder_output: outputs #folder to save recorded data to

recorder:
  flag: True #set to True to record data
  file: "cases.sql" #this is the name of the file to record data to.

In the above example, data will be saved to the file outputs/cases.sql.

The file below would have the same behavior as the above example, but highlights the default behavior:

name: "driver_config"
description: "example driver config to show recording options"

general:
  folder_output: outputs #folder to save recorded data to

recorder:
  flag: True #set to True to record data
  file: "cases.sql" #this is the name of the file to record data to.
  overwrite_recorder: False #create a unique recorder for each simulation
  recorder_attachment: "model" # "driver" or "model"
  includes: ["*"] # include everything
  excludes: ["*resource_data"] # don't include resource data

  • overwrite_recorder: If False or not specified, H2I will make a new filename for the recorder that doesn't yet exist in the outputs folder. If cases.sql exists, it will make a new file named cases0.sql. If cases.sql and cases0.sql exist, it will make a new file named cases1.sql, etc. If set to True, it will overwrite an existing cases.sql file from previous runs.

  • recorder_attachment: Must be model or driver, defaults to model. We recommend attaching the recorder to the driver if running an optimization or design of experiments in parallel as model-level recording cannot be performed in parallel due to limitations in OpenMDAO.

Attaching a recorder to the driver#

We recommend attaching the recorder to the driver if running an optimization or design of experiments in parallel as model-level recording cannot be performed in parallel due to limitations in OpenMDAO. It can be beneficial if running a design of experiments or optimization in serial as well. Further documentation on driver recording can be found here.

name: "driver_config"
description: "example driver config to show recording options"

general:
  folder_output: outputs #folder to save recorded data to

recorder:
  flag: True #set to True to record data
  file: "cases.sql" #this is the name of the file to record data to.
  overwrite_recorder: False #create a unique recorder for each simulation
  recorder_attachment: "driver" #"driver" or "model

  # H2I Default recorder options
  includes: ["*"] # include everything
  excludes: ["*resource_data"] # don't include resource data

  # OpenMDAO default for recording options
  record_inputs: True #record inputs
  record_outputs: True #record outputs
  record_constraints: True #record constraints
  record_derivatives: False #record derivatives
  record_desvars: True #record design variables
  record_objectives: True #record objectives

Attaching a recorder to the model#

Further documentation on model recording can be found here. By default, the recorder will be attached to the model unless recorder_attachment is set to "driver".

name: "driver_config"
description: "example driver config to show recording options"

general:
  folder_output: outputs #folder to save recorded data to

recorder:
  flag: True #set to True to record data
  file: "cases.sql" #this is the name of the file to record data to.
  overwrite_recorder: False #create a unique recorder for each simulation
  recorder_attachment: "model" #"driver" or "model

  # H2I Default recorder options
  includes: ["*"] # include everything
  excludes: ["*resource_data"] # don't include resource data

  # OpenMDAO default for recording options
  record_inputs: True #record inputs
  record_outputs: True #record outputs
  record_residuals: True #record residuals

Loading Recorder Files#

Detailed documentation on OpenMDAO's case read can be found here.

Example usage of reading and accessing recorded data is shown in Example 8 (examples/08_wind_electrolyzer/run_wind_electrolyzer.py).

Below is an example python script that shows how to load recorded data and to access the data available:

import openmdao.api as om
from pathlib import Path
# set the path for the recorder from stuff specified in the driver_config.yaml
fpath = Path.cwd() / "outputs" / "cases.sql"

# load the cases
cr = om.CaseReader(fpath)

# get the cases as a list
cases = list(cr.get_cases())
# access a variable from the problem, this can be anything thats an
# input or output from the models that were run and not specified
# as variables to exclude in the driver_config file

cases[0].get_val("finance_subgroup_default.LCOE",units="USD/(kW*h)")
cases[0].get_val("solar.capacity_kWdc",units="MW")
Contents