IEEE Power & Energy Magazine - May/June 2022 - 37
can indicate that across all combinations of generator
set points (i.e., one combination for 10 MW, one for
10.1 MW, and so on), there is a 5% maximum violation
of the line transmission limit.
The methods we describe allow for such an assessment. If
the performance is unsatisfactory, we can take a step back,
retrain the neural network on more data points, and re-evaluate
the performance. This improves the accuracy and robustness
of the prediction, ultimately delivering better guarantees
and enabling the adoption of such techniques.
Secure or Not Secure? Guaranteeing
How Neural Networks Classify
Let us consider a guiding example: we want to determine
whether a specific operating point of a given power system
is safe under a combination of security criteria (e.g.,
N-1 criterion, small-signal stability, and so on). Instead of
using conventional methods (e.g., power flows and eigenvalue
analyses), we train a neural network to assess whether
the operating point is secure or not secure in a fraction of
the time (usually 100 to 1,000 times faster). As inputs, we
provide the active power set points of generators and active
and reactive power consumption of loads. The boundary
where the classification changes from secure to not secure
is called the security boundary. The correct prediction of
that boundary is crucial because the neural network could
misclassify a " not secure " operating point as " secure, " posing
a risk to power system operation and potentially leading
to a blackout.
Using this guiding example, Figure 4(a) and (b) shows a
comparison of the standard procedure used to evaluate the
performance and the proposed methodology, respectively.
The conventional procedure relies on a test data set of probable
power system operating scenarios. For this test data set,
the neural network classification is compared to the ground
truth, assessing classification accuracy.
Conventional Performance Assessment (Statistical Evaluation)
Possible Power System Operating Scenarios
Security Classification by Neural Network
Secure
Trained
Neural Network
Predicted
Not Secure
Security Boundary
Ground Truth
Security Boundary
(a)
Obtaining Performance Guarantees
Possible Power System Operating Regions
Guaranteed Classification by Neural Network
Secure
Trained
Neural Network
Predicted
Not Secure
Security Boundary
Ground Truth
Security Boundary
(b)
figure 4. A comparison of the standard evaluation of the neural network performance with the new proposed methodology.
(a) The conventional procedures use a test set, and (b) the proposed techniques obtain performance guarantees for
continuous input regions.
may/june 2022
ieee power & energy magazine
37
Add Samples to Training
Data Set to Improve
Robustness
IEEE Power & Energy Magazine - May/June 2022
Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - May/June 2022
Contents
IEEE Power & Energy Magazine - May/June 2022 - Cover1
IEEE Power & Energy Magazine - May/June 2022 - Cover2
IEEE Power & Energy Magazine - May/June 2022 - Contents
IEEE Power & Energy Magazine - May/June 2022 - 2
IEEE Power & Energy Magazine - May/June 2022 - 3
IEEE Power & Energy Magazine - May/June 2022 - 4
IEEE Power & Energy Magazine - May/June 2022 - 5
IEEE Power & Energy Magazine - May/June 2022 - 6
IEEE Power & Energy Magazine - May/June 2022 - 7
IEEE Power & Energy Magazine - May/June 2022 - 8
IEEE Power & Energy Magazine - May/June 2022 - 9
IEEE Power & Energy Magazine - May/June 2022 - 10
IEEE Power & Energy Magazine - May/June 2022 - 11
IEEE Power & Energy Magazine - May/June 2022 - 12
IEEE Power & Energy Magazine - May/June 2022 - 13
IEEE Power & Energy Magazine - May/June 2022 - 14
IEEE Power & Energy Magazine - May/June 2022 - 15
IEEE Power & Energy Magazine - May/June 2022 - 16
IEEE Power & Energy Magazine - May/June 2022 - 17
IEEE Power & Energy Magazine - May/June 2022 - 18
IEEE Power & Energy Magazine - May/June 2022 - 19
IEEE Power & Energy Magazine - May/June 2022 - 20
IEEE Power & Energy Magazine - May/June 2022 - 21
IEEE Power & Energy Magazine - May/June 2022 - 22
IEEE Power & Energy Magazine - May/June 2022 - 23
IEEE Power & Energy Magazine - May/June 2022 - 24
IEEE Power & Energy Magazine - May/June 2022 - 25
IEEE Power & Energy Magazine - May/June 2022 - 26
IEEE Power & Energy Magazine - May/June 2022 - 27
IEEE Power & Energy Magazine - May/June 2022 - 28
IEEE Power & Energy Magazine - May/June 2022 - 29
IEEE Power & Energy Magazine - May/June 2022 - 30
IEEE Power & Energy Magazine - May/June 2022 - 31
IEEE Power & Energy Magazine - May/June 2022 - 32
IEEE Power & Energy Magazine - May/June 2022 - 33
IEEE Power & Energy Magazine - May/June 2022 - 34
IEEE Power & Energy Magazine - May/June 2022 - 35
IEEE Power & Energy Magazine - May/June 2022 - 36
IEEE Power & Energy Magazine - May/June 2022 - 37
IEEE Power & Energy Magazine - May/June 2022 - 38
IEEE Power & Energy Magazine - May/June 2022 - 39
IEEE Power & Energy Magazine - May/June 2022 - 40
IEEE Power & Energy Magazine - May/June 2022 - 41
IEEE Power & Energy Magazine - May/June 2022 - 42
IEEE Power & Energy Magazine - May/June 2022 - 43
IEEE Power & Energy Magazine - May/June 2022 - 44
IEEE Power & Energy Magazine - May/June 2022 - 45
IEEE Power & Energy Magazine - May/June 2022 - 46
IEEE Power & Energy Magazine - May/June 2022 - 47
IEEE Power & Energy Magazine - May/June 2022 - 48
IEEE Power & Energy Magazine - May/June 2022 - 49
IEEE Power & Energy Magazine - May/June 2022 - 50
IEEE Power & Energy Magazine - May/June 2022 - 51
IEEE Power & Energy Magazine - May/June 2022 - 52
IEEE Power & Energy Magazine - May/June 2022 - 53
IEEE Power & Energy Magazine - May/June 2022 - 54
IEEE Power & Energy Magazine - May/June 2022 - 55
IEEE Power & Energy Magazine - May/June 2022 - 56
IEEE Power & Energy Magazine - May/June 2022 - 57
IEEE Power & Energy Magazine - May/June 2022 - 58
IEEE Power & Energy Magazine - May/June 2022 - 59
IEEE Power & Energy Magazine - May/June 2022 - 60
IEEE Power & Energy Magazine - May/June 2022 - 61
IEEE Power & Energy Magazine - May/June 2022 - 62
IEEE Power & Energy Magazine - May/June 2022 - 63
IEEE Power & Energy Magazine - May/June 2022 - 64
IEEE Power & Energy Magazine - May/June 2022 - 65
IEEE Power & Energy Magazine - May/June 2022 - 66
IEEE Power & Energy Magazine - May/June 2022 - 67
IEEE Power & Energy Magazine - May/June 2022 - 68
IEEE Power & Energy Magazine - May/June 2022 - 69
IEEE Power & Energy Magazine - May/June 2022 - 70
IEEE Power & Energy Magazine - May/June 2022 - 71
IEEE Power & Energy Magazine - May/June 2022 - 72
IEEE Power & Energy Magazine - May/June 2022 - 73
IEEE Power & Energy Magazine - May/June 2022 - 74
IEEE Power & Energy Magazine - May/June 2022 - 75
IEEE Power & Energy Magazine - May/June 2022 - 76
IEEE Power & Energy Magazine - May/June 2022 - 77
IEEE Power & Energy Magazine - May/June 2022 - 78
IEEE Power & Energy Magazine - May/June 2022 - 79
IEEE Power & Energy Magazine - May/June 2022 - 80
IEEE Power & Energy Magazine - May/June 2022 - 81
IEEE Power & Energy Magazine - May/June 2022 - 82
IEEE Power & Energy Magazine - May/June 2022 - 83
IEEE Power & Energy Magazine - May/June 2022 - 84
IEEE Power & Energy Magazine - May/June 2022 - 85
IEEE Power & Energy Magazine - May/June 2022 - 86
IEEE Power & Energy Magazine - May/June 2022 - 87
IEEE Power & Energy Magazine - May/June 2022 - 88
IEEE Power & Energy Magazine - May/June 2022 - 89
IEEE Power & Energy Magazine - May/June 2022 - 90
IEEE Power & Energy Magazine - May/June 2022 - 91
IEEE Power & Energy Magazine - May/June 2022 - 92
IEEE Power & Energy Magazine - May/June 2022 - Cover3
IEEE Power & Energy Magazine - May/June 2022 - Cover4
https://www.nxtbook.com/nxtbooks/pes/powerenergy_gridedge_2023
https://www.nxtbook.com/nxtbooks/pes/powerenergy_050622
https://www.nxtbook.com/nxtbooks/pes/powerenergy_030422
https://www.nxtbook.com/nxtbooks/pes/powerenergy_010222
https://www.nxtbook.com/nxtbooks/pes/powerenergy_111221
https://www.nxtbook.com/nxtbooks/pes/powerenergy_091021
https://www.nxtbook.com/nxtbooks/pes/powerenergy_070821
https://www.nxtbook.com/nxtbooks/pes/powerenergy_050621
https://www.nxtbook.com/nxtbooks/pes/powerenergy_030421
https://www.nxtbook.com/nxtbooks/pes/powerenergy_010221
https://www.nxtbook.com/nxtbooks/pes/powerenergy_111220
https://www.nxtbookmedia.com