Tecnología y Ciencias del Agua - page 135

133
Zhang et al ., Improved online sequential extreme learning machine for simulation of daily reference evapotranspiration
Tecnología y Ciencias del Agua
, vol. VIII, núm. 2, marzo-abril de 2017, pp. 127-140
ISSN 2007-2422 •
the parameters, the model is accurate. Thirdly,
when only two temperatures were used as the
inputs, it still performed better than RMSE, EF
and R2 statistics. So, properly reducing some
variables and adopting reasonable combina-
tions of variables can improve the accuracy of
prediction.
Comparison with other calculation
formulas
The IOS-ELM model was compared with the
ELM and LSSVM, as well as conventional
models including Hargreaves, Mc-Cloud and
Priestley-Taylor methods in respect of RMSE
and MAE statistics in different cities in tables
4-5. There are six parameters as input variables
in the model.
Tables 4-5 show that IOS-ELM outperformed
all other models by all performance criteria.
Compared with the intelligent and empirical
models, the ISO-ELM performed the best value
of RMSE<0.46 and MAE<0.41, and the ELM
and LSSVM models performed better than the
others. A few differences appeared among the
Mc Cloud and Priestley-Taylor models. It was
Table 3. Influence of different meteorological data combinations on ET0.
Model inputs RMSE R2 EF%
All 0.4132 0.9696 95.72
Tmax, T, Tmin, n , Uh, RH 0.4132 0.9696 95.72
Tmax, T, Tmin, n , Uh 0.7865 0.9625 92.3
Tmax, T, Tmin, n , RH 0.6737 0.9619 94.3
Tmax, T, Tmin, Uh, RH 0.7868 0.9620 92.2
Tmax, T, n , Uh, RH 0.9021 0.9496 89.8
Tmax, Tmin, n , Uh, RH 0.9135 0.9490 89.6
T, Tmin, n , Uh, RH 0.9666 0.9423 88.3
Tmax, T, Tmin, n 0.8270 0.9591 91.4
Tmax, T, Tmin, Uh 0.7338 0.9673 93.2
Tmax, T, Tmin, RH 0.6368 0.9651 94.9
Tmin , n, Uh, RH 1.2321 0.9060 81.1
T, n , Uh, RH 1.0048 0.9377 87.4
Tmax, n , Uh, RH 0.8428 0.9560 91.1
Tmax, T, Tmin 0.8505 0.9566 90.9
n , Uh, RH 2.2612 0.6222 36.3
Tmax, T, RH 0.6480 0.9753 94.7
Tmax, Tmin, n 0.8580 0.9548 90.8
Tmax, Tmin, Uh 0.7942 0.9616 92.1
Tmax, Uh, RH 0.7436 0.9668 93.1
Tmax, n , RH 0.6298 0.9664 95.06
Tmax, n , Uh 0.8786 0.9513 90.3
Tmax, T 0.9464 0.9432 88.8
T, Tmin 0.9119 0.9502 89.64
Tmax, RH 0.6258 0.9670 95.12
Tmax, n 0.9012 0.9483 89.89
Tmax, Uh 0.9999 93.59 87.55
n , RH 2.1127 0.6694 44.42
n , Uh 2.1611 0.6543 41.85
Uh , RH 2.6964 0.3395 9.47