Usage notes ¶

Installation ¶

Clone and install via

           git clone https://github.com/windnode/WindNODE_ABW.git # or by SSH: git@github.com:windnode/WindNODE_ABW.git
pip install -e /path/to/cloned/repo/

(a virtualenv is recommended).

Notice: The recent package psycopg2-binary in setup.py conflicts with the psycopg2 required by egoio . If the install breaks, use the following temporary workaround:

Install requirements manually in your venv, egoio should be the last. Install it without dependencies by using

pip install --no-dependencies egoio

To run the model, you also need a solver to be installed such as CBC or Gurobi. On Linux, you can install CBC with apt install coinor-cbc . Make sure the solver is set in the run configuration dict in run_scenario.py .

Trouble shooting ¶

Installation of `psutils` ¶

During the installtion of windnode_abw , the package psutils might fail. Install the system package python3-dev with

sudo apt install python3-dev

Installation of `pygraphviz` ¶

During the installtion of windnode_abw , the package pygraphviz might fail. Install additional system packages with

sudo apt install python3-dev graphviz libgraphviz-dev pkg-config

Setup postgres database with docker (optional) ¶

Note You don’t have to necessarily use docker to create a PostgreSQL database. Using a native installation works as well.

Inside the repo’s root directory (where docker-compose.yml lives) execute

docker-compose up -d --build

Afterwards you can access the database via

Field	Value
host	localhost
port	54321
Maintance DB	windnode_abw
User	windnode
Password	windnode

Import scenario data ¶

Scenario data is contained in the database dump available [here]( https://zenodo.org/record/4898349/ ). Do the following steps to import the scenario data to your database:

Download the above scenario data file
Import tables, data, and constraints by ` pg_restore -U windnode -d windnode_abw -h localhost -p 54321 -W --no-owner --no-privileges --no-tablespace -1 </path/to/windnode_db_200817.backup> ` To overwrite existing tables, you may use the –clean argument.

Setup database connection config file ¶

When you try to run windnode_abw/run_scenario.py , it will search for the file $HOME/.egoio/config.ini . More specifically, in the file config.ini it searches for a section [windnode_abw] . It won’t be found whe you run it for the first time. Subsequently, a command-line dialog opens that asks you for database connection details.

When you use a local database, the section in the config looks like

` [windnode_abw] dialect = psycopg2 username = windnode host = localhost port = 54321 database = windnode_abw `

Run Optimization ¶

To start the optimization, you can use the command line interface: run script run_scenario.py from the shell:

           python run_scenario.py [-h] [--mp [NUMBER]] [SCENARIO [SCENARIO ...]]

          

where NUMBER is the number of threads and SCENARIOS the scenarios to be executed. For example, to run all scenarios in 4 processes, use

python run_scenario.py --mp 4 all

To get help on parameters and available scenarios you can use

python run_scenario.py -h

Depending on the system settings, the optimization takes about 1-2 hours for each scenario.

By default, raw results are written to ~/.windnode_abw/results/ , a subdirectory with a timestamp (run id) is created (e.g. ~/.windnode_abw/results/2020-08-05_024335/ ).

Post-processing results ¶

The results, aggregated on different temporal and spatial levels, are calculated by post-processing the raw results from above. These post-processed data is stored as pickle file in subdirectory ./processed of the run id folder and can be quickly loaded, e.g. from the jupyter notebooks.

By default, this step is automatically performed after the optimization run but can be manually triggered by passing force_new_results=True to the notebook creation functions (see below).

Analyzing results ¶

A single notebook for each scenario can be produced best by using papermill. Either for one scenario

           create_scenario_notebook("NEP2035",
                         '2020-07-24_104145_1month',
                         template="scenario_analysis_template.ipynb")

          

or for multiple scenarios using multiprocessing (in this case all)

           create_multiple_scenario_notebooks(
     "all",
     '2020-07-24_104145_1month',
     template="scenario_analysis_template.ipynb",
     num_processes=None
)

          

You can then further convert to the executed notebook to HTML by

jupyter nbconvert scenario_analysis_NEP2035.ipynb

Note

Some plots (those generated with plotly) in the generated notebooks may won’t show up initially. This can be solved by clicking File -> Trust Notebook . To trust all notebooks in the notebook directory, you can use
```
jupyter trust *.ipynb
```
before you start the jupyter notebook server.
If parameter output_path is not passed, the standard path /path/to/windnode/repo/windnode_abw/jupy/ is used.
If you get an error like WARNING: No such kernel named … try to open the template notebook and save it manually to set your current kernel name.