Design and implementation of a general batch simulation tool of SOWFA and its application in training a single-turbine surrogate

Abstract

Simulator fOr Wind Farm Applications (SOWFA) is a powerful wind farm simulation tool. It is widely used in wake effect research. However, the integration between AI and SOWFA is weak. Without AI, it is difficult for us to fully use SOWFA. The installation, parameter settings, and calculation procedures of SOWFA are complex, which hinders the application of SOWFA in practical power system engineering scenarios. To mitigate these issues, a batch simulation tool of SOWFA based on Docker and Python is designed and implemented. Firstly, the Docker engine is used to realize the virtualization container management of SOWFA. A SOWFA installation image is built to make it easy to migrate and deploy. Secondly, a Python application programming interface (API) of the SOWFA container is programmed to realize parameter editing, calculation process management, and batch simulations. Finally, a batch simulation post-processing Python API is constructed based on the ParaView library so that the SOWFA batch simulation results can be collated into data samples for downstream tasks. The above SOWFA image and Python API are tested in a simple single-turbine wake analysis scenario. With the samples generated by the batch simulations, an autoencoder-based single-turbine wake prediction surrogate model is trained, which verifies the effectiveness of the SOWFA batch simulation tool designed in this paper.