2. Parameters File

The parameters file is in JSON format. The repository includes an example called seals.params that can be used as a starting point, but will need to heavily modifified for your particular paths, files, etc. The parameters are:

  • mode - “ARCHIVE” or “REALTIME”, currently only ARCHIVE is supported. REALTIME is a potential future capability.

  • site_map_file - a netCDF file containing a 3D grid of the site.

  • sensor_type - specifies the type of sensor input file: “SENSIT” or “NETCDF”.

  • sensit_path - the path of the directory containing the SENSIT data. The directory contains daily subdirectories by the name “YYMMDD data” (yes, with a space), each containing multiple FMD_Data_Exportxxxxxxxxxxxxxxxxxxxxx.csv files, one file per device. Required when sensor_type is SENSIT.

  • sensit_hour_offset - time offset, in hours, to add to raw sensit data. Required when sensor_type is SENSIT.

  • sensit_vars_interpolated - list of sensit variable names to be interpolated when unifying the time coordinates of the SENSIT readings. Required when sensor_type is SENSIT.

  • sensit_met_sensors - list (often just one) of which of the SENSIT sensors has meteorological data (sensor ID(s)). Required when sensor_type is SENSIT.

  • sensit_data_start_datetime - the datetime of the start of the period of SENSIT data to be processed. Required when sensor_type is SENSIT.

  • sensit_data_duration - the duration of the period of SENSIT data to be processed. The sum of this and sensit_data_start_datetime would give you the end datetime. Required when sensor_type is SENSIT.

  • sensit_time_increment - the time increment, in seconds, by which to unify the time coordiantes of the SENSIT readings. Required when sensor_type is SENSIT.

  • time_window_size - the size of the time window, in seconds, to use for either of the two Block Transformer models. It should divisible by the block_size (part of the model). If not, the framework software will adjust it (with warning) to the nearest multiple of block_size. Required for either of the Block Transformer models.

  • time_window_stride - the window_stride value used in the call to the specific_site_data_generation function of sealsml.staticinference, which defines it as: stride width to slide window through time (index based).

  • sensor_readings_file - the path of the sensor readings netCDF file when the sensor_type is NETCDF (not SENSIT)

  • model_file - the path to the .keras model file to be run.

  • pot_leaks_file - the path to the netCDF file of potential leak locations to be evaluated.

  • csv_output - whether or not to produce CSV output of the results of the run. true or false. This is currently the only output option.

  • csv_location_output_filename - the path/filename of the file to be written with leak location information, when csv_output is true.

  • csv_rate_output_filename - the path/filename of the file to be written with leak location information, when csv_output is true. This may or may not be produced, depending on the model type.

  • std_netcdf_write_path - the path/filename of the file to be written with the SEALS standard dataset needed by the specific_site_data_generation function of sealsml.staticinference.py, which defines it as: The dataset to be processed.

  • siteds_netcdf_write_path - the path/filename of the file to be written with the site_dataset, which is the netCDF output of the specific_site_data_generation function of sealsml.staticinference.py, which defines it as: Encoder and Decoder in an xarray dataset.