This repository holds the code to generate the dataset for fault diagnosis in a power distribution system. For the moment, the following feeder systems are included.
- 13 BUS
- 34 BUS
- 37 BUS
- 123 BUS
- 8500 NODE
Learn more about these feeder systems from these resources [1], [2], [3].
As an example see the 37Bus system folder.
-
dataset.npy: This contains the feature tensor with dimensions Number of Samples X Number of Nodes in the feeder system X 6 -
edge_list.npy: This contains the edge list matrix which is the graph connectivity information. This can processed with libraries likeNetworkXto generate the graph representing the distribution system. -
fault_detection_labels.npy: This contains the binary fault detection labels. For a fault event, the label is$1$ and for a non-fault event, the label is$0$ . -
fault_location_labels.npy: This contains the fault location label for three-phase nodes in the distribution system. The location labels is encoded to an integer value. Usebus_id_mapto see which bus name corresponds to which integer value. -
fault_class_labels.npy: This contains the fault classification label for a fault event. The fault classes are encoded as follows- Line-to-Ground (LG) : 0
- Line-to-Line (LL) : 1
- Line-to-Line-to-Ground (LLG) : 2
- Line-to-Line-to-Line (LLL) : 3
- Line-to-Line-to-Line-to-Ground (LLLG) : 4
- Non-Fault Event : 5
-
fault_resistance_labels.npy: This contains the numerical value for fault resistance for the fault event. We recommend normalizing the value before using it in a regression task. -
fault_currents_labels.npy: This contains the numerical value for the fault current for the fault event. We strongly recommend normalizing the value before using it as the range and variance are large. -
1_hop_by_bus_name.npy: This holds the one-hop neighborhood node name for all the nodes in the feeder system. -
2_hop_by_bus_name.npy: This holds the two-hop neighborhood node name for all the nodes in the feeder system. -
bus_id_map.npy: This holds the mapping between bus names and the corresponding unique integer id.
The other files contain the metadata for simulation settings and the feeder system.
feeder_infos: A json file that contains the feeder information.X_Dataset: Contains the simulation parameters
As an example, the folder structure for the 37Bus is shown below.
37Bus_Dataset
├── dataset
│ ├── dataset.npy
│ ├── edge_list.npy
| ├── fault_detection_labels.npy
| ├── fault_location_labels.npy
| ├── fault_class_labels.npy
| ├── fault_resistance_labels.npy
| ├── fault_currents_labels.npy
| ├── 1_hop_by_bus_name.npy
| ├── 2_hop_by_bus_name.npy
│ └── bus_id_map.npy
|
├── 37Bus_Dataset
└── feeder_infos
Clone the repository with the following command
git clone https://github.com/dibalokechanda/fault_bat.gitThen cd into the root of the directory. Install the packages from the requirements.txt file with the following command. We highly recommend using a virtual environment to avoid package conflicts.
pip install -r requirements.txtImportant Note: This code is implemented using py_dss_interface==1.0.2 which is the older version.
To change between feeder systems update the FEEDER_SYSTEM variable. For more granular control modify the command list in the script.py file. As an example consider IEEE-37 Bus system:
command = [
'python', 'main_dataset_generation.py',
'--feeder', '37Bus',
'--feeder-file', 'ieee37.dss',
'--fault-resistance-type','fixed',
'--fault-resistance-value','20',
'--fault-resistance-lower-end', '0.05',
'--fault-resistance-upper-end', '20',
'--folder', '37Bus_Dataset',
'--number-of-samples-for-each-node','10',
'--change-load-values','yes',
'--load-value-KW-lower-end','20',
'--load-value-KW-upper-end','80'
]These are the command line arguments for running the code. Following is an explanation for each argument:
-
python: This means executing a Python script. -
main_dataset_generation.py: The path to the main file. -
--feeder: Name of the folder that contains the feeder files. Make sure it matches the folder name under./feeders -
--feeder-file: The main file to execute for simulating the feeder system. -
--fault-resistance-type: If the fault resistance is variable or fixed. We just exploring the research works in graph-based fault diagnosis in distribution systems to decide on this parameter. -
--fault-resistance-value: If the fault resistance value is fixed, the value of the fault resistance. -
--fault-resistance-lower-end: If the fault resistance is variable, it will be sampled from a uniform distribution. This parameter defines the lower end of that uniform distribution. -
--fault-resistance-upper-end: If the fault resistance is variable, it will be sampled from a uniform distribution. This parameter defines the upper end of that uniform distribution. -
--folder: Name of the folder to save the generated dataset. -
--number-of-samples-for-each-node: For each node in the feeder system, how many data points are to be generated. -
--change-load-values: If the load value needs to be changed or not. We recommend setting it to 'yes' if you plan to use the dataset for fault detection. Changing the load value simulates a non-fault event. -
--load-value-KW-lower-end: The load value is sampled from a uniform distribution. This parameter defines the lower end of the uniform distribution. -
--load-value-KW-upper-end: The load value is sampled from a uniform distribution. This parameter defines the upper end of the uniform distribution.
To execute the code run the following command
python .\script.py