Download PowerPoint Presentation - Operational forecasts of Dst

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Operational forecasts of Dst
Henrik Lundstedt
Hans Gleisner
Peter Wintoft
Main goal – The Lund Dst model
• Use an Elman recurrent neural network to
predict Dst from solar wind data.
• Find the network that solves the problem
with as few neurons as possible.
• Implement for real time forecasting.
• Publish models on the Internet in Java and
Matlab code.
Model
• Inputs: [By(t)], Bz(t), n(t), V(t).
• Output: Dst(t+1).
• Data
– OMNI set from 1963 – current.
– >10 days of continuous data (data gaps max 2
hours).
Elman Neural Network
Statistical evaluation
RMSE (nT) Correlation
Lund Dst model
10.3
0.88
O’Brien and McPherron
12.3
0.83
Fenrich and Luhmann
15.3
0.78
Burton et al.
16.4
0.76
Statistical evaluation cont.
Statistical evaluation cont.
Example
Interpretation of weights
Burton:
Elman network:
Interpretation of weights cont.
Web pages
• Regional Warning Center – Sweden
– www.lund.irf.se/rwc
• Dst real time forecasts
– www.lund.irf.se/rwc/dst
• Dst Matlab and Java models
– www.lund.irf.se/rwc/dst/models
Related documents