Short-term Wind Speed Forecasting by Combination
of Masking Signal-based Empirical Mode
Decomposition and Extreme Learning Machine
Qiu Jihui,Shen Shaoping
∗
,Xu Guangyu
Department of Automation,
School of Aeronautics and Astronautics ,
Xiamen University, Xiamen 361005, P. R. China
Email: xmqiujihui@163.com,shen shaoping@163.com
Abstract—According to the requirement of accurate predic-
tion of the short-term wind speed series,this paper proposes
a new short-term combination prediction model of the wind
speed series by means of the masking signal-based empirical
mode decomposition(MS-EMD) and the extreme learning ma-
chine(ELM).Firstly,because of the non-stationary characteristics
of the wind speed series,the wind speed series is decomposed
into several components with different frequency bands by the
MS-EMD to reduce the non-stationary.Secondly,in order to avoid
the randomness of input dimensionality selection of the ELM,the
phase space of each component is reconstructed.Thirdly,the ELM
model of each component is established to predict the wind
speed series.Finally,the predicted results of each component
are superimposed to get the final result.The simulation result
verifies that the proposed combination forecasting model is able
to excavate the wind speed series features effectively and has
relatively high prediction accuracy.
Index Terms—Wind speed series, prediction, masking signal-
based empirical mode decomposition, phase space reconstruction,
extreme learning machine.
I. INTRODUCTION
The stratosphere airship is an important floated platform
which works in about 20 km altitude for a long time with a
lot of payload.It can play a role similar to man-made satellite
in the fields of communications relay,remote sensing to the
ground,and air traffic control [1]-[2].
Due to the earth’s rotation,there is a strong west wind in
mid latitude stratosphere while the wind speed changes along
with the height,the latitude and longitude,and the season [3].In
the station-keeping control for stratosphere airship,the wind
is a disturbance factor.The propulsion control system of the
stratosphere airship needs to adjust the speed of the motor to
keep the position of the stratosphere airship in a predetermined
region according to the change of wind speed.
Therefore,wind speed forecasting is an important subject
of the stratosphere airship,and it can make a great differ-
ence.According to the predicted wind speed at the next mo-
ment, the motor speed of the stratosphere airship is adjusted
in advance, so that the stratosphere airship can overcome
the influence of the change of wind speed with smaller
power,which can reduce the loss of energy.
In wind speed forecasting methods,a lot of work have
been done at home and abroad.It mainly includes time series
model [4],artificial neural network model [5],support vector
machine model [6],etc.These methods not only have their
characteristics,but have certain limitation.Time series method
requires a lot of historical data.Artificial neural network is easy
to fall into local minimum value,etc.
The extreme learning machine [7] is a new single hidden
layer feedforward neural network,which is proposed by Huang
Guangbin in 2006.It has greatly improved the learning speed
and the generalization ability of the network.What’s more,the
extreme learning machine has a strong non-linear fitting abili-
ty,and has achieved good result in non-linear fitting prediction
[8].However,the extreme learning machine can only fit the
non-linear part of the wind speed series.The non-stationary
characteristics of the wind speed series will have a great
influence on the predicted result.So,it is very important to
reduce the non-stationary characteristics.
The empirical mode decomposition(EMD) is very suitable
for non-linear and non-stationary signal processing [9].How-
ever,the mode-mixing phenomenon may exist in the process of
decomposition and affect the accuracy of forecasting [10].MS-
EMD is an improvement on the empirical mode decomposi-
tion [11].It can solve the mode mixing phenomenon of the
EMD.Based on the above analysis,this paper proposes a new
short-term combination prediction model of wind speed by
means of masking signal-based empirical mode decomposition
(MS-EMD) and extreme learning machine (ELM).The exper-
imental results show that the method proposed in this paper
has relatively high prediction accuracy.
II. MASKING SIGNAL-BASED EMPIRICAL MODE
DECOMPOSITION(MS-EMD)
For processing the non-stationary characteristics of the
wind speed series,the traditional empirical mode decompo-
sition(EMD) method usually decomposes the wind speed
series into several components with different frequen-
cies.However,the mode-mixing phenomenon may exist in the
decomposition process and affect the prediction accuracy.MS-
EMD is an improvement on the empirical mode decom-
The 11th International Conference on
Computer Science & Education ICCSE 2016
August 23-25, 2016. Nagoya University,Japan