Short-Term Wind Speed Forecasting Model Based on Mutual Information and Recursive Neural Network

doi:10.16183/j.cnki.jsjtu.2020.433

Abstract

Abstract:

The volatility and randomness of wind speed have caused potential safety hazards to wind power grid integration. Improving wind speed forecasting is crucial to the stability of wind power systems and the development of wind energy. A novel short-term wind speed forecasting model (MI-RNN) was proposed based on mutual information (MI) and recursive neural network (RNN). In this model, the MI theory was introduced to select the optimal length of historical wind speed sequence (τ), and the method of using each τ step to forecast wind speed at the next time point was adopted to input the historical wind speed data into RNN for model training. The final wind speed forecasting result was output by the trained RNN model. Besides, the MI-RNN model was applied to the wind speed dataset collected from a wind farm and the forecasting accuracy was compared with that of the traditional wind forecasting methods. The results show that the MI-RNN model has achieved a higher forecasting accuracy compared with the commonly used wind farm wind speed forecasting methods, and can accurately forecast the future wind direction, which is expected to be applied to wind speed forecasting of wind farms with spatial dimensions.

Key words: wind speed forecasting, wind energy, mutual information (MI), deep learning, recurrent neural network (RNN), wind farm

CLC Number:

TK89

WANG Yan, CHEN Yaoran, HAN Zhaolong, ZHOU Dai, BAO Yan. Short-Term Wind Speed Forecasting Model Based on Mutual Information and Recursive Neural Network[J]. Journal of Shanghai Jiao Tong University, 2021, 55(9): 1080-1086.

Figures/Tables 14

Fig.1

Fig.2

Fig.3

Fig.4

Tab.1

Wind speed dataset statistics at an interval of 15 min

数据量/个		$F -$ / (m·s^-1)	σ_F/ (m·s^-1)	F_min/ (m·s^-1)	F_max/ (m·s^-1)
训练集	测试集	$F -$ / (m·s^-1)	σ_F/ (m·s^-1)	F_min/ (m·s^-1)	F_max/ (m·s^-1)
384	96	8.451	2.438	2.481	14.642

Tab.1

Fig.5

Fig.6

Fig.7

Fig.8

Tab.2

Tab.3

Wind speed dataset statistics at an interval of 5 min

数据量/个		$F -$ (m·s^-1)	σ_F(m·s^-1)	$θ -$ /(°)	σ_θ/(°)
训练集	测试集	$F -$ (m·s^-1)	σ_F(m·s^-1)	$θ -$ /(°)	σ_θ/(°)
84096	21024	9.74	4.42	9.74	4.42

Tab.3

Fig.9

Fig.10

Fig.11

References 15

[1]	ZHANG Z D, YE L, QIN H, et al. Wind speed prediction method using shared weight long short-term memory network and Gaussian process regression[J]. Applied Energy, 2019, 247:270-284. doi: 10.1016/j.apenergy.2019.04.047 URL
[2]	CHEN J, ZENG G Q, ZHOU W, et al. Wind speed forecasting using nonlinear-learning ensemble of deep learning time series prediction and extremal optimization[J]. Energy Conversion and Management, 2018, 165:681-695. doi: 10.1016/j.enconman.2018.03.098 URL
[3]	KHOSRAVI A, MACHADO L, NUNES R. Time-series prediction of wind speed using machine learning algorithms: A case study Osorio wind farm, Brazil[J]. Applied Energy, 2018, 224:550-566. doi: 10.1016/j.apenergy.2018.05.043 URL
[4]	MORENO S R, LEANDRO D S C. Wind speed forecasting approach based on Singular Spectrum Analysis and Adaptive Neuro Fuzzy Inference System[J]. Renewable Energy, 2018, 126:736-754. doi: 10.1016/j.renene.2017.11.089 URL
[5]	张驰. 风电场短期风速预测若干问题研究[D]. 南京:东南大学, 2017.
	ZHANG Chi. Research on some issues of short-term wind speed forecasting for wind farms[D]. Nanjing: Southeast University, 2017.
[6]	JAHANGIR H, GOLKAR M A, ALHAMELI F, et al. Short-term wind speed forecasting framework based on stacked denoising auto-encoders with rough ANN[J]. Sustainable Energy Technologies and Assessments, 2020, 38:100601. doi: 10.1016/j.seta.2019.100601 URL
[7]	SANTHOSH M, VENKAIAH C, KUMAR D M V. Ensemble empirical mode decomposition based adaptive wavelet neural network method for wind speed prediction[J]. Energy Conversion and Management, 2018, 168:482-493. doi: 10.1016/j.enconman.2018.04.099 URL
[8]	范曼萍, 周冬. 基于改进粒子群优化LS-SVM的短期风速预测[J]. 电力学报, 2020, 35(2):123-128.
	FAN Manping, ZHOU Dong. Shortterm wind speed prediction based on improved particle swarm optimization LS-SVM[J]. Journal of Electric Power, 2020, 35(2):123-128.
[9]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735 URL
[10]	TRAPPENBERG T, OUYANG J, BACK A. Input variable selection: Mutual information and linear mixing measures[J]. IEEE Transactions on Knowledge and Data Engineering, 2006, 18(1):37-46. doi: 10.1109/TKDE.2006.11 URL
[11]	LIU H, MI X W, LI Y F. Smart multi-step deep learning model for wind speed forecasting based on variational mode decomposition, singular spectrum analysis, LSTM network and ELM[J]. Energy Conversion and Management, 2018, 159:54-64. doi: 10.1016/j.enconman.2018.01.010 URL
[12]	PEDREGUSA F, VAROQUAUX G, GRAMFORT A, et al. Scikit-learn: Machine learning in python[J]. Journal of Machine Learning Research, 2011, 12:2825-2830.
[13]	PASZKE A, GROSS S, MASSA F, et al. Pytorch: An imperative style, high-performance deep learning library[C]// 33rd Conference on Neural Information Processing Systems (NeurIPS 2019). Vancouver, Cana-da: NeurIPS, 2019: 8026-8037.
[14]	CHAI T, DRAXLER R R. Root mean square error (RMSE) or mean absolute error (MAE)?—Arguments against avoiding RMSE in the literature[J]. Geoscientific Model Development, 2014, 7(3):1247-1250. doi: 10.5194/gmd-7-1247-2014 URL
[15]	QUININO R C, REIS E A, BESSEGATO L F. Using the coefficient of determination R2 to test the significance of multiple linear regression[J]. Teaching Statistics, 2013, 35(2):84-88. doi: 10.1111/test.2013.35.issue-2 URL

模型	RMSE/(m·s^-1)	MAE/(m·s^-1)	R²
MI-RNN	0.237	0.192	0.992
ANN	0.490	0.437	0.965
SVM	0.299	0.325	0.977