Kusiak.qxd_Outsourcing_book_temp.qxd 26/02/2010 10:16 Page 65
Short-term Prediction of Windfarm Power – A Data-driven Approach
Time Series Models
Figure 2: Observed and Predicted Wind Power Using Integrated
Time series models are proven effective tools for short-term prediction
Time Series Models
issues in the wind energy area (see
5
Figure 2). Time series prediction
120
A
focuses on determining future events (e.g. wind speed and wind
power) based on known events, measured typically at successive time-
100
points and spaced at (often uniform) time intervals. The basic time
80
series prediction model is as follows:
60
ower (kW)
P
y(t+T) = f(y(t), y(t-T),..., y(t-mT))
40
where
20
T is the sampling time (time interval), y(t+T) is the predicted
parameter, y(t), y(t-T),..., y(t-mT) are the current and past observed
0
parameters and
1 16 31 46 61 76 91 106 121 136 151 166 181 196
m+1 is the number of inputs (predictors) to the model.
Test number (10 minutes ahead prediction)
Predicted power Observed power
Evaluation of Data-driven Models
To build a data-driven model, available data are divided into training
120
B
and testing data. A common recommendation is that two-thirds of the
available data are used for training and the other one-third are used to 100
test the built model. The data points in the testing data set should be
80
‘after’ the data points of the training data set in the time sequence. In
this way, testing accuracy could indicate the model’s ability to predict
60
future data points, and for how long it can predict them.
ower (kW)
P
40
Typically, a model is applied with several data-mining algorithms, such 20
as NNs, SVMs, BTA and so on. The key parameters provided by
0
the algorithms are adjusted to seek the best performance for each
1 16 31 46 61 76 91 106 121 136 151 166 181 196
algorithm. In the research presented here, the following four metrics
Test number (30 minutes ahead prediction)
are used for model comparison: mean absolute error (MAE),
Predicted power Observed power
mean relative error (MRE) and the standard deviation (STD) of both
MAE and MRE. 120
C
100
Industrial Case Studies
Three industrial cases are introduced in this section. The first case
6
80
compares the performance from two direct time series models,
60
one using 10-minute average SCADA data and the other average
ower (kW)
P
SCADA data at hourly intervals. The second case
6
introduces the 40
power prediction model using the k-nearest neighbours (kNN)
20
algorithm integrated with the wind speed predicted by time-series
prediction models. The third case
7
presents the concept of a virtual
0
turbine and the prediction of windfarm power using 10-second
1 16 31 46 61 76 91 106 121 136 151 166 181 196
SCADA data.
Test number (50 minutes ahead prediction)
Predicted power Observed power
Direct Time Series Model
D 120
The total power of the windfarm analysed in this section was scaled to the
interval 0–100MW by the windfarm capacity. Five of the most promising
100
data-mining algorithms, including the SVM algorithm, multilayer
80
perceptron (MLP) algorithm, M5P tree algorithm, decision/regression tree
(REP tree) and the bagging tree, were applied to extract the 10-minute
60
time series prediction model of wind power. The SVM algorithm
ower (kW)
P
40
outperformed the other four algorithms. The MLP and REP tree algorithms
performed worst. Therefore, the SVM algorithm was selected to build the
20
10-minute time series prediction model of windfarm power.
0
1 16 31 46 61 76 91 106 121 136 151 166 181 196
The important predictors of the hourly time series model of windfarm
Test number (60 minutes ahead prediction)
power that were selected by the BTA are {y(t), y(t-T), y(t-2T), y(t-3T)}. The
Predicted power Observed power
sampling time for this model is one hour, and thus all variables need to
be on the same time-scale. Five of the most promising data-mining
Observed and predicted wind power: A: 10 minutes ahead prediction; B: 30 minutes ahead prediction;
algorithms C: 50 minutes ahead prediction; D: 60 minutes ahead prediction.{y(t), t(t-T), y(t-2T), y(t-3T)} were selected to test the accuracy
MODERN ENERGY REVIEW VOLUME 2 ISSUE 1
65
Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57  |  Page 58  |  Page 59  |  Page 60  |  Page 61  |  Page 62  |  Page 63  |  Page 64  |  Page 65  |  Page 66  |  Page 67  |  Page 68  |  Page 69  |  Page 70  |  Page 71  |  Page 72  |  Page 73  |  Page 74  |  Page 75  |  Page 76  |  Page 77  |  Page 78  |  Page 79  |  Page 80  |  Page 81  |  Page 82  |  Page 83  |  Page 84  |  Page 85  |  Page 86  |  Page 87  |  Page 88  |  Page 89  |  Page 90  |  Page 91  |  Page 92  |  Page 93  |  Page 94  |  Page 95  |  Page 96  |  Page 97  |  Page 98  |  Page 99  |  Page 100  |  Page 101  |  Page 102  |  Page 103  |  Page 104  |  Page 105  |  Page 106  |  Page 107  |  Page 108