1. Global solar radiation forecasting with non-lineal
statistical techniques and qualitative predictions
from Spanish National Weather Service
Presented by:
LUIS MARTÍN POMARES
ENERGY DEPARTAMENT
Renewable energy division
5rd Experts Meeting of the IEA SHC Task
“Solar Resource Knowledge Management”
&
MESoR Meeting
Wels, Austria
9 – 11 June 2008
2. OUTLOOK
1. TIME SERIES GLOBAL SOLAR
RADIATION HALF DAILY
PROPERTIES
2. CLEARNESS INDEX PREDICTIONS
3. LOST COMPONENT PREDICTIONS
4. LOST COMPONENT + QUALITATIVE
PREDICTIONS
5. CONCLUSIONS
2
3. TIME SERIES GLOBAL SOLAR
RADIATION HALF DAILY PROPERTIES
Global solar irradiance half daily time series :
Non Stationary
Non Gaussian Data Data Preprocessing
Non Linear
Removing stationarity from time series:
Using clearness index.
Using lost component variable.
Subtracting each value mean value of 2d+1 days.
Removing from time series anual armonic (first armonic from
fourier analysis).
If the time series obtained is not stationary a good
way to achieve so is differencing consecutive days.
3
4. CIES2008, June 17-21,2008, Vigo (Spain)
CLEARNESS INDEX HALF DAILY
PREDICTIONS
45.0 ° N
IMPROVEMENT OVER PERSISTENCE IN TERMS OF %RMSD
ierrormod el
42.5 ° N
improvement = 1 − ÷
ierror ÷
40.0 ° N
• Madrid RRN AEMet
37.5 ° N persistence
• Murcia RRN AEMet
MADRID RRN AEMet
35.0 ° N
15.0 ° W12
.5° W10.0° W 7.5 °
W 5.0 ° W 2.5° W 0.0 ° 2.5 ° E 5.0 ° E 7.5
° E 10.0
°E MURCIA RRN AEMet
M adrid: M ejora frente Persistencia M urcia: M ejora frente Persistencia
24 24
M S (2)-A R (1)/P ersistencia
22 N N (10)/P ersistencia 22
A N FIS (6)/P ersistencia
20 20
18 18
M ejo ra % R M SD
16 16 M S (2)-AR (1)/Persistencia
NN (10)/P ersistencia
14 14 A N FIS(6)/Persistencia
12 12
10 10
8 8
6 6
1 2 3
Horizonte Predicción (Sem idías)
4 5 6 1 2 3
4 4
Horizonte Predicción (Sem idías)
5 6
5. CIES2008, June 17-21,2008, Vigo (Spain)
CLEARNESS INDEX HALF DAILY
PREDICTIONS
RMSD FOR BEST MODEL: NN(10)
MADRID RRN AEMet MURCIA RRN AEMet
40 40
N N (1) N N (1 )
38 N N (2) 38 N N (2 )
N N (3) N N (3 )
36 N N (4) 36 N N (4 )
N N (5)
% R M S D P re d ic c ió n S e m id ia ria K t
N N (6) N N (5 )
34 34
N N (7) N N (6 )
32 N N (8) 32 N N (7 )
N N (9) N N (8 )
N N (10) N N (9 )
30 30
PER
N N (1 0 )
28 28 PE R
26 26
24 24
22 22
1 2 3 4 5 6 1 2 3 4 5 6
H orizonte Pre dic c ión (S em idía s ) H o rizo nte Pre d ic c ió n (S e m id ía s )
5
7. EUROSUN2008, October 7-10, Lisbon (Portugal)
LOST COMPONENT
PREDICTIONS
6000
5000
4000
Wm-2
3000
2000
1000
0
0 100 200 300 400 500 600 700
Half Day
7
Differencing day by day
8. EUROSUN2008, October 7-10, Lisbon (Portugal)
LOST COMPONENT
PREDICTIONS
5000
4500
4000
Lost Component Wm-2
3500
3000
2500
2000
1500
1000
500
0 100 200 300 400 500 600 700
Half Day
8 Differencies
9. EUROSUN2008, October 7-10, Lisbon (Portugal)
LOST COMPONENT
PREDICTIONS
3000
2000
Diff Lost Component Wm-2
1000
0
-3
-1000 1.5
x 10
1
-2000
N u m b e r o f S am p les 0.5
-3000
0 100 200 300 400 500 600 700
Half Day 0
-4 000 -3000 9
-20 00 -1 000 0 10 00
D iff L ost C om ponent
2000 300 0 4 000
11. CIES2008, June 17-21,2008, Vigo (Spain)
LOST COMPONENT
PREDICTIONS
IMPROVEMENT OVER PERSISTENCE IN TERMS OF %RMSD
0.3
0.28
MADRID RRN AEMet 0.26
M adrid: M ejora frente Persistencia
24 0.24 N N /PE R
M S (2)-A R (1)/P ersistencia
0.22
Im p ro v e m e n t % R M S E
22 N N (10)/P ersistencia
A N FIS (6)/P ersistencia
20 0.2
18 0.18
M ejo ra % R M SD
16 0.16
14 0.14
12 0.12
10 0.1
8 0.08
6
1 2 3 4 5 6
0.06
Horizonte Predicción (Sem idías) 1 2 3 4 5 6
Pre dic tion Horizon
11
12. CIES2008, June 17-21,2008, Vigo (Spain)
LOST COMPONENT
PREDICTIONS
IMPROVEMENT OVER PERSISTENCE IN TERMS OF %RMSD
0.3
MURCIA RRN AEMet 0.28
0.26
M urcia: M ejora frente Persistencia
24 0.24
Im p ro vem en t % R M SD
22 0.22
20 0.2
18 0.18
16 M S (2)-AR (1)/Persistencia 0.16
NN (10)/Persistencia
14 A NFIS (6)/Persistencia 0.14 N N /P E R
12 0.12
10 0.1
8 0.08
6
1 2 3 4 5 6 0.06
Horizonte Predicción (Sem idías) 1 2 3 4 5 6
Horizon Prediction Half D aily
12
14. EUROSUN2008, October 7-10, Lisbon (Portugal)
AEMet Synoptic Predictions by Site
Total Cloud Cover
ECMWF
HIRLAM/AEMet
PROMETEO
WRF/MM5…. Energy Values
14
Buenas días, mi nombre es Luis Martín y voy a presentar el trabajo realizado hasta la fecha en el ámbito de la predicción de la radiación solar diaria.
He organizado el contenido de la presentación, comenzando por la presentación de los objetivos, revisión de técnicas predictivas para la predicción de la irradiancia solar diaria, después se pasará a describir la metodología empleada en el ensayo propuesto, se presentarán los resultados obtenidos para terminar comentando las conclusiones así como las principales líneas de trabajo en el futuro.
Lost component is the difference between extraterrestrial solar irradiance and ground solar irradiance. Therefore, lost component represents the quantity of energy absorbed by the atmosphere. Even this time series is not stationary and has seasonal variations which can be solved differencing consecutive half-days.
The result is a time series nonstationary with a probability density function similar to a delta function. This signal is the random part of solar irradiance and is isolated from the deterministic part of solar irradiance