IEEE Computational Intelligence Magazine - August 2021 - 65

TABLE VII Short-term prediction performance of DNR models trained using different evolution algorithms
for the experimental wind speed time series.
SHORT-TERM WIND SPEED TIME SERIES WITH A PARTITION RATIO OF 1:1
ALGORITHM
DNR-GA
DNR-PSO
DNR-JADE
MSE (MEAN ± STD
(OR SD))
1.14E-02±4.36E-03
1.09E-02±1.51E-02
8.55E-03±5.98E-03
DNR-L-SHADE 1.27E-02±1.50E-02
DNR-SMS
5.43E-03±3.54E-04
ALGORITHM
DNR-GA
DNR-PSO
DNR-JADE
MSE (MEAN ± STD
(OR SD))
7.04E-03±3.07E-03
3.35E-03±5.19E-04
3.92E-03±1.38E-03
DNR-L-SHADE 3.66E-03±1.53E-03
DNR-SMS
3.15E-03±1.50E-04
ALGORITHM
DNR-GA
DNR-PSO
DNR-JADE
MSE (MEAN ± STD
(OR SD))
7.52E-03±2.95E-03
3.98E-03±2.90E-03
4.17E-03±5.79E-04
DNR-L-SHADE 3.55E-03±1.95E-04
DNR-SMS
3.50E-03±7.30E-05
P-VALUE
1.12E-06
1.37E-06
2.54E-04
9.13E-07
-
RMSE (MEAN ± STD
(OR SD))
1.05E-01±1.94E-02
9.63E-02±4.08E-02
8.91E-02±2.47E-02
1.05E-01±4.06E-02
7.36E-02±2.41E-03
P-VALUE
1.85E-06
6.15E-02
8.25E-04
3.40E-01
-
RMSE (MEAN ± STD
(OR SD))
8.20E-02±1.79E-02
5.78E-02±4.26E-03
6.19E-02±9.57E-03
5.96E-02±1.05E-02
5.61E-02±1.32E-03
P-VALUE
1.24E-06
9.79E-01
3.02E-06
5.25E-01
-
RMSE (MEAN ± STD
(OR SD))
8.50E-02±1.72E-02
6.14E-02±1.46E-02
6.44E-02±4.41E-03
5.95E-02±1.59E-03
5.91E-02±6.17E-04
V. Conclusion
A novel DNR method inspired by biological neurons is
employed to predict wind speed time series. To improve the
prediction results, the neural architecture for DNR is trained
using the SMS algorithm, which is a global optimization algorithm
with powerful search capabilities. Before the prediction
process, the time delay and the number of embedding dimensions
are calculated using the MI algorithm and the FNN
algorithm, respectively. Based on the results, phase space
reconstruction of the wind speed data is performed, and the
maximum Lyapunov exponent is calculated to detect the chaotic
nature of the wind speed series. Comprehensive experiments
conducted to evaluate the effectiveness of the proposed
DNR-SMS approach are reported, in which several commonly
used prediction approaches are considered for comparison
and four diverse evaluation indicators are employed. The
experimental results and statistical tests demonstrate that DNR
can outperform the other compared approaches. The proposed
DNR-SMS model is thus an efficient tool for wind speed
prediction due to its high prediction accuracy and robust stability.
Further research can be conducted in the following
directions. First, other optimization algorithms can be evaluated
to enhance the forecasting performance of the DNR
P-VALUE
1.12E-06
1.37E-06
2.75E-04
9.13E-07
-
MAE (MEAN ± STD
(OR SD))
7.56E-01±1.47E-01
6.11E-01±1.49E-01
5.99E-01±1.13E-01
6.41E-01±1.44E-01
5.24E-01±1.58E-02
SHORT-TERM WIND SPEED TIME SERIES WITH A PARTITION RATIO OF 4:1
P-VALUE
1.85E-06
6.40E-02
8.85E-04
3.40E-01
-
MAE (MEAN ± STD
(OR SD))
6.31E-01±1.44E-01
4.28E-01±2.23E-02
4.61E-01±7.11E-02
4.21E-01±2.54E-02
4.27E-01±1.14E-02
SHORT-TERM WIND SPEED TIME SERIES WITH A PARTITION RATIO OF 9:1
P-VALUE
1.24E-06
9.79E-01
3.02E-06
5.08E-01
-
MAE (MEAN ± STD
(OR SD))
6.47E-01±1.37E-01
4.65E-01±1.19E-01
4.80E-01±3.67E-02
4.46E-01±1.34E-02
4.46E-01±7.34E-03
P-VALUE
9.13E-07
7.95E-01
2.02E-04
7.64E-01
-
R (MEAN ± STD
(OR SD))
8.26E-01±8.36E-02
8.86E-01±1.06E-01
8.95E-01±1.43E-02
9.02E-01±6.46E-03
9.03E-01±2.07E-03
model. Second, DNR-SMS can be applied to solve other prediction
problems, such as container throughput forecasting and
power load prediction.
Acknowledgments
This work was supported by a Project of the Guangdong Basic and
Applied Basic Research Fund (No. 2019A1515111139), by the
National Natural Science Foundation of China under Grants
61876110 and 61876162, by a Key Program of the Joint Funds
of the National Natural Science Foundation of China under
Grant U1713212, and in part by the Research Grants Council
of the Hong Kong Special Administrative Region, China,
under Grant PolyU11202418 and Grant PolyU11209219.
References
[1] W. Yang, J. Wang, H. Lu, T. Niu, and P. Du, " Hybrid wind energy forecasting and
analysis system based on divide and conquer scheme: A case study in China, " J. Cleaner
Production, vol. 222, pp. 942-959, 2019. doi: 10.1016/j.jclepro.2019.03.036.
[2] C. Tian, Y. Hao, and J. Hu, " A novel wind speed forecasting system based on hybrid
data preprocessing and multi-objective optimization, " Appl. Energy, vol. 231, pp. 301-319,
2018. doi: 10.1016/j.apenergy.2018.09.012.
[3] H. Li, J. Wang, H. Lu, and Z. Guo, " Research and application of a combined model
based on variable weight for short term wind speed forecasting, " Renewable Energy, vol.
116, pp. 669-684, 2018. doi: 10.1016/j.renene.2017.09.089.
[4] M. Imal, M. Sekkeli, C. Yildiz, and F. Kececioglu, " Wind energy potential estimation
and evaluation of electricity generation in Kahramanmaras, Turkey, " Energy Educ. Sci.
Technol. A, Energy Sci. Res., vol. 30, no. 1, pp. 661-672, 2012.
P-VALUE
2.25E-06
5.00E-01
2.92E-02
9.19E-01
-
R (MEAN ± STD
(OR SD))
8.80E-01±6.31E-02
9.32E-01±9.73E-03
9.29E-01±2.09E-02
9.27E-01±2.95E-02
9.36E-01±3.12E-03
P-VALUE
1.24E-06
3.02E-06
2.12E-03
1.12E-06
-
R (MEAN ± STD
(OR SD))
8.46E-01±7.47E-02
8.46E-01±1.52E-01
8.80E-01±9.16E-02
8.11E-01±1.45E-01
9.21E-01±5.28E-03
AUGUST 2021 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE 65
IEEE Computational Intelligence Magazine - August 2021

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - August 2021
