|
Guest >> Sign In |
|
|
DIRECTED
ENERGY
PROFESSIONAL
SOCIETY
Abstract: 25-Symp-121
|
|
|
|
UNCLASSIFIED, PUBLIC RELEASE
Turbulence Forecasting with ML Optimized for NWP Data
Accurate atmospheric turbulence modeling enables optical system performance modeling. A machine learning weather-driven turbulence model trained on local measurements can be used for turbulence forecasting hours to days in advance by applying the model to numerical weather prediction (NWP) data. However, this technique suffers from biases between measurements and NWP data. In this presentation, we demonstrate these biases and optimize a weather-driven turbulence model for the forecasting application. Further, we use Bayesian statistics to produce a turbulence prediction interval that captures the probable range of conditions instead of a single value. These techniques improve turbulence forecasting performance and enables DE test planning, forecasting the effectiveness of DE, and automating the selection of system settings for a range of predicted conditions.
UNCLASSIFIED, PUBLIC RELEASE
|