Closed
Milestone expired on Jul 31, 2018

Mowflot T1.1: Speed-up code, IO, models

Responsible: Ju Feng.

Deadline: July

  1. Goldwind defines an API that the GUI Django app is currently using, and new developments should try to keep this API intact. When this API changes this needs to discussed and agreed upon. [GW]
  • N/A
  1. Find faster alternative methods for reading and storing input data (resources). For example, the input resources are read from a yaml file and stored again in Python’s standard binary pickle format (pkl files). [WE]
  • The JSON format is found to be faster than YML and routines to read and write JSON wind farm files have been implemented.
  1. Profile existing code and find possibilities for speed-ups considering vectorization, the usage of the Numba module, or porting to Cython. [WE]
  • Profile existing code.
  • Great speedup is identified by changing from Pandas to Numpy and by vectorizing the wake model.
  1. Speed-up IO by considering carefully in which format to read and write the data from. Select the appropriate data formats considering the size of the IO data. [WE]
  • GRD-format utilized