Deep-learning-supervised-topology-detection

Journal: Deep learning supervised topology detection of building energy systems by generated time series of generic grey-box models

Contact

If you have any questions about the code, please feel free to contact fstinner@eonerc.rwth-aachen.de.

Examples

In the folder code the code for using the classifier is included. main.py is an example of the use of the code. The code for the classifiers ALSTM-FCN, LSTM-FCN, MALSTM-FCN and MLSTM-FCN are taken from paper IsmailFawaz.2019 and have been adapted accordingly for our use case. The code for the remaining classifiers is taken from papers IsmailFawaz.2019 and Fawaz.11.09.2019 and has also been adapted to our use case accordingly. The utils files were also taken from paper IsmailFawaz.2019 and adapted accordingly.

In folder data_set all nine data sets are saved (three cases (case 1-3) with training and testing with real data (real), training and testing with simulated data (sim) and training with simulated data and testing with real data (real_sim). Each record is stored in a .npyand .pkl file. These can be read with the help of python (packages numpy and pickle). The pickle files are in the format matching sktime. With the classifiers existing in sktime, these data sets can be used. sktime has several methods to convert the data sets suitably into other formats. The datasets are divided into X_train, X_test, y_train and y_test. These can be used directly with any algorithm.

The code for generating generic data from simulation models is under revision for better usage but is included in folder code_simulation with small examples.

Used paper for deep learning

@article{IsmailFawaz.2019,
 author = {{Ismail Fawaz}, Hassan and Forestier, Germain and Weber, Jonathan and Idoumghar, Lhassane and Muller, Pierre-Alain},
 abstract = {Data Mining and Knowledge Discovery, https://doi.org/10.1007/s10618-019-00619-1},
 year = {2019},
 title = {Deep learning for time series classification: a review},
 pages = {917--963},
 volume = {33},
 number = {4},
 journal = {Data Mining and Knowledge Discovery}}

@booklet{Karim.2018b,
 author = {Karim, Fazle and Majumdar, Somshubra and Darabi, Houshang and Harford, Samuel},
 year = {2018},
 title = {Multivariate LSTM-FCNs for Time Series Classification}}

@booklet{Fawaz.11.09.2019,
 author = {{Ismail Fawaz}, Hassan and Lucas, Benjamin and Forestier, Germain and Pelletier, Charlotte and Schmidt, Daniel F. and Weber, Jonathan and Webb, Geoffrey I. and Idoumghar, Lhassane and Muller, Pierre-Alain and Petitjean, Fran{\c{c}}ois},
 year = {2019},
 title = {InceptionTime: Finding AlexNet for Time Series Classification}}

Used paper for simulation

@incollection{Muller.2016,
 author = {M{\"u}ller, Dirk and Lauster, Moritz and Constantin, Ana and Fuchs, Marcus and Remmen, Peter},
 title = {AixLib - An Open-Source Modelica Library within the IEA-EBC Annex 60 Framework},
 urldate = {21.08.2018},
 pages = {3--9},
 booktitle = {BauSim 2016},
 year = {September 2016}}

@inproceedings{Stinner.2019c,
 author = {Stinner, Florian and Yang, Yingying and Schreiber, Thomas and Bode, Gerrit and Baranski, Marc and M{\"u}ller, Dirk},
 title = {Generating Generic Data Sets for Machine Learning Applications in Building Services Using Standardized Time Series Data},
 publisher = {{International Association for Automation and Robotics in Construction (IAARC)}},
 series = {Proceedings of the International Symposium on Automation and Robotics in Construction (IAARC)},
 editor = {Al-Hussein, Mohamed},
 booktitle = {Proceedings of the 36th International Symposium on Automation and Robotics in Construction (ISARC)},
 year = {2019}}