The aim of the thesis is to optimally coordinate the reactive power sources in offshore wind farms in a predictive manner based to the principle of minimizing the wind farm power losses, as well the variations of the transformers tap positions. First, an accurate Neural Network-based wind speed forecasting algorithm was developed in order to counteract the uncertainty of the wind and finally, the optimal management of the available reactive sources is tackled by a metaheuristics-based method. Two different cases were investigated: a far-offshore wind farm with HVDC interconnection link and the AC connected Dutch wind farm BORSSELE.

1. 1
Metaheuristics-based Optimal Management
of Reactive Power Sources in
Offshore Wind Farms
Aimilia-Myrsini Theologi 05-10-2016
Supervisors: Dr. Jose Luis Rueda Torres, TU Delft

2. 2
Outline
• Motivation
• Research approach
• Wind Speed Forecasting
• MVMO
• Optimization
• Grid Code Requirements
• Results
• Conclusions & Future Work
MVMO=Mean Variance Mapping Optimization

3. 3
Motivation
• Grid Code Requirements
 Safe, secure and economic operation
• Uncoordinated management of Var sources in real-time
operations:
 increase of losses
 decrease of efficiency
 reduced life-time of switchable devices
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

4. 4
Research tasks
 Development of wind speed forecasting method, which will be
incorporated in optimal reactive power management
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
 Formulation and solution of reactive power sources operation
in a coordinated manner

5. 5
Definition of Objective Function (1)
Approach 2: optimization is performed for 24-hours time horizon
𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑶𝑭 =
𝒕=𝟏
𝟐𝟒
(𝒘 𝟏 ∙ 𝑷 𝑳,𝒕 + 𝒘 𝟐 ∙ 𝑶𝑳𝑻𝑪 𝒄𝒐𝒔𝒕,𝒕)
where,
𝑷 𝑳,𝒕: real power losses of the system for each hour t
𝑶𝑳𝑻𝑪 𝒄𝒐𝒔𝒕,𝒕: the operational cost of the number of tap changes
for each hour t
𝑶𝑳𝑻𝑪 𝒄𝒐𝒔𝒕,𝒕 = 𝒘 𝟑 ∙ 𝒕𝒂𝒑 𝒕 − 𝒕𝒂𝒑 𝒕−𝟏
𝒘 𝟏, 𝒘 𝟐, 𝒘 𝟑: cost coefficients
Approach 1: 24-hours, every hour is optimized individually
𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑶𝑭 = 𝑷 𝑳,𝒕, 𝒕 = 𝟏, 𝟐, … , 𝟐𝟒
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

6. 6
Predictive Optimization (HVDC link)
PCC=Point of Common Coupling
MVMO=Mean Variance Mapping Optimization
HVDC=High Voltage Direct Current
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

7. 7
Predictive Optimization (AC cable)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
PCC=Point of Common Coupling
MVMO=Mean Variance Mapping Optimization
AC=Alternating Current

8. 8
Definition of Objective Function (2)
Vector of parameters to be optimized:
𝑥 = [𝑄1, … , 𝑄 𝑁 , 𝑡𝑎𝑝1, … , 𝑡𝑎𝑝 𝑁]
Discrete
variables
Continuous
variables
=> Mixed-integer non-linear problem => Necessity for metaheuristic methods
𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑶𝑭
𝒗 𝒎𝒊𝒏 ≤ 𝒗 ≤ 𝒗 𝒎𝒂𝒙
𝒊 ≤ 𝒊𝒍𝒊𝒎
𝒔 ≤ 𝒔𝒍𝒊𝒎
𝒒 𝑾𝑻𝑮
𝒎𝒊𝒏
≤ 𝒒 𝑾𝑻𝑮 ≤ 𝒒 𝑾𝑻𝑮
𝒎𝒂𝒙
𝒕𝒂𝒑 𝑻𝒓,𝒎𝒊𝒏 ≤ 𝒕𝒂𝒑 𝑻𝒓 ≤ 𝒕𝒂𝒑 𝑻𝒓,𝒎𝒂𝒙
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
OF=Objective Function

9. 9
Accurate wind speed forecasting
• Reduces the risk of uncertainty
• Contributes to better grid planning
• Reserves power for integrating wind power
=> Beneficial for optimal operation of a power system
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

10. 10
Time-scales
 Very short-term (few minutes to 1 hour)
 Short-term (1 hour to several hours)
 Medium-term (several hours to 1 week)
 Long-term (1 week to one year)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

11. 11
NN-based forecasting method (1)
+ Data-driven approach
+ Remarkable effectiveness in short-term forecasting
̶ Over a year of historical data is necessary
Implementation in MATLAB => “Neural Network Toolbox”
NN=Neural Network
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

12. 12
Old approach: increase amount of input data
New approach: divide the one year data into days
NN-based forecasting method (2)
Motivation
& Research Tasks
Reactive Power
in Power Systems
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
NN=Neural Network

13. 13
• Feed forward neural network
NN-based forecasting method (3)
• Back propagation learning
algorithm
• One Hidden layer (5 neurons)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
Unidirectional flow of data
NN=Neural Network

14. 14
Day-ahead prediction
• The evaluation criteria is:
AMAPE=Average Mean Absolute Percentage Error
𝑨𝑴𝑨𝑷𝑬 =
𝟏𝟎𝟎
𝟐𝟒
𝒉=𝟏
𝟐𝟒
| 𝒑𝒓𝒆𝒅𝒊𝒄𝒕𝒆𝒅 𝒔𝒑𝒆𝒆𝒅 𝒉 − 𝒂𝒄𝒕𝒖𝒂𝒍 𝒔𝒑𝒆𝒆𝒅 𝒉|
𝒂𝒗𝒆𝒓𝒂𝒈𝒆 𝒘𝒊𝒏𝒅 𝒔𝒑𝒆𝒆𝒅
Season AMAPE (%)
Winter 21,05
Spring 14,82
Summer 16,37
Fall 18,26
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

15. 15
Metaheuristic Techniques
in Power Systems – MVMO
MVMO
 Single-agent search algorithm
 Internal search range is
restricted to [0,1]
 Special mapping function
 Enhanced performance in
terms of convergence speed
MVMO=Mean Variance Mapping Optimization
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

16. 16
START
Termination criteria satisfied ?
STOP
Generate random solutions
in range [0,1]
Denormalize parameters & feed
to model in PowerFactory
Run load flow calculations
and obtain P,Q values
Fitness evaluation
by using de-normalized variables
Yes
No
MVMO flowchart
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
MVMO=Mean Variance Mapping Optimization

17. 17
Evolutionary mechanism
 Storing solutions
𝒉 𝒙, 𝒔 𝟏, 𝒔 𝟐, 𝒙 = 𝒙 ∙ 𝟏 − 𝒆−𝒙∙𝒔 𝟏 + 𝟏 − 𝒙𝒊 ∙ 𝒆−(𝟏−𝒙)∙𝒔 𝟐
where,
𝒙: mean variable
𝒔 𝟏, 𝒔 𝟐: shape variables
The variable 𝒙𝒊 is always between the range [0,1]
for every 𝒙𝒊
∗
.
 Mapping function
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

18. 18
Implementation
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
𝒙 = [𝑸 𝟏, … , 𝑸 𝑵 , 𝒕𝒂𝒑 𝟏, … , 𝒕𝒂𝒑 𝑵]
Provides parameters
Performs
load flow
calculations
Provides Q values

19. 19
Parameters to be optimized
3-winding Transformer
with OLTC
WTG=Wind Turbine Generator
OLTC=On Load Tap Changer
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

20. 20
Far-offshore wind farm (288 MW)
with HVDC interconnection link
OLTC=On Load Tap Changer
HVDC=High Voltage Direct Current
AC=Alternating Current
System Information
Number of wind turbines: 48
Number of 3-winding
transformer with OLTC: 2
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

21. 21
Borssele Near-shore wind farm (600 MW)
with AC cables
System Information
Number of wind turbines: 100
Number of 3-winding
transformer with OLTC: 4
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
OLTC=On Load Tap Changer
AC=Alternating Current

22. 22
Case studies
• Far-offshore wind farm – DC connected (280 MW)
CASE 1 Offshore Transformers (x2)
Min. (Losses) for 24-hours,
for the current operating point
CASE 2 Offshore Transformers (x2)
Min. (Losses + Tap Changes) for 24-hours,
for a predicted time horizon
CASE 3 Onshore Transformers (x2)
Min. (Losses) for 24-hours,
for the current operating point
CASE 4 Onshore Transformers (x2)
Min. (Losses + Tap Changes) for 24-hours,
for a predicted time horizon
CASE 5 Offshore Transformers (x2)
Min. (Losses) for 24-hours,
for the current operating point
CASE 6 Offshore Transformers (x2)
Min. (Losses + Tap Changes) for 24-hours,
for a predicted time horizon
CASE 7
On- & Off- shore
Transformers (x2)
Min. (Losses) for 24-hours,
for the current operating point
CASE 8
On- & Off- shore
Transformers (x2)
Min. (Losses + Tap Changes) for 24-hours,
for a predicted time horizon
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
• Near-shore Borssele wind farm – AC connected (600 MW)
DC=Direct Current
AC=Alternating Current

23. 23
Grid Code Requirements (1)
-Far-offshore DC connected wind farm
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
PCC=Point of Common Coupling
WTG=Wind Turbine Generator
HVDC=High Voltage Direct Current
DC=Direct Current
AC=Alternating Current

24. 24
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
Grid Code Requirements (2)
-Near-shore AC connected Borssele wind farm
PCC=Point of Common Coupling
WTG=Wind Turbine Generator
DC=Direct Current
AC=Alternating Current

25. 25
Case 1 – Far-offshore: Min.(Losses)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
2 4 6 8 10 12 14 16 18 20 22 24
6
8
10
12
14
16
18
20
22
Reductionoflosses(%)
Time ( h )
INPUT:
predicted
wind speed
2 4 6 8 10 12 14 16 18 20 22 24
4
5
6
7
8
9
10
11
12
PredictedWindSpeed(m/s)
Time ( h )

26. 26
-2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 2,5
P ( % PN
)
Q ( MVar )
80
20
100
0 2 4 6 8 10 12 14 16 18 20 22 24
0,80
0,85
0,90
0,95
1,00
1,05
1,10
1,15
1,20
BUS A
BUS B
BUS C
BUS D
Voltagesof33kVBuses
Time ( h )
Grid Code
Requirements
satisfied
Case 1 – Far-offshore: Min.(Losses)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
-0,45 -0,30 -0,15 0,00 0,15 0,30 0,45 0,60
0,8
0,9
1,0
1,1
1,2U ( p.u. )
Q/Pmax at PCC
Q/Pmax
5,02 m/s
11,51 m/s
8,33 m/s
.
.
.
.
.
.

27. 27
Case 7 – Near-shore: Min.(Losses)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
INPUT:
predicted
wind speed
2 4 6 8 10 12 14 16 18 20 22 24
10
15
20
25
30
35
40
45
50
55
60
ReductionoftheOF(%)
Time ( h )
2 4 6 8 10 12 14 16 18 20 22 24
6
7
8
9
10
11
12
PredictedWindSpeed(m/s)
Time ( h )

28. 28
Grid Code
Requirements
satisfied
Case 7 – Near-shore: Min.(Losses)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
-2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 2,5
0
20
40
60
80
100
P ( % PN
)
Q ( MVar )
-60 -45 -30 -15 0 15 30 45 60
0
20
40
60
80
100
P ( % PN
)LV_A 66 kV Bus
MV_A 66 kV Bus
LV_B 66 kV Bus
MV_B 66 kV Bus
Q at Offshore PCC ( MVar )
-150 -120 -90 -60 -30 0 30 60 90 120
0
20
40
60
80
100
P ( % PN
) Onshore PCC A
Onshore PCC B
Q at Onshore PCC ( MVar )
initial curve
acceptable deviation
7,4 m/s
11,51 m/s
9,15 m/s
.
.
.
.

29. 29
Case 8 – Near-shore: Min.(Losses+Tap Changes)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
2 4 6 8 10 12 14 16 18 20 22 24
0
5000
10000
15000
20000
25000
30000
35000
Cumulative Initial Cost
Cumulative Optimum Cost
Cost(Euro)
Time ( h )
41,12 %
reduction
𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑶𝑭 =
𝒕=𝟏
𝟐𝟒
(𝒘 𝟏 ∙ 𝑷 𝑳,𝒕 + 𝒘 𝟐 ∙ 𝑶𝑳𝑻𝑪 𝒄𝒐𝒔𝒕,𝒕)

30. 30
2 4 6 8 10 12 14 16 18 20 22 24
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
Tappositions
Time ( h )
Onshore Transformer A
Onshore Transformer B
2 4 6 8 10 12 14 16 18 20 22 24
-6
-4
-2
0
2
4
6
8
Tappositions
Time ( h )
Offshore Transformer A
Offshore Transformer B
2 4 6 8 10 12 14 16 18 20 22 24
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
Onshore Transformer A
Onshore Transformer B
Tappositions
Time ( h )
2 4 6 8 10 12 14 16 18 20 22 24
-2
-1
0
1
2
Tappositions
Time ( h )
Offshore Transformer A
Offshore Transformer B
Current
operating
point
Predicted
time
horizon
Comparison Case 7 & 8
Transformers Tap Positions
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
Min.
(Losses)
Min.
(Losses+Tap changes)

31. 31
Robustness of MVMO
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
MVMO=Mean Variance Mapping Optimization
Case 8 – Near-shore: Min.(Losses+Tap Changes)

32. 32
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
Conclusions
1. Predictive optimization leads to better results
2. High efficiency of MVMO in optimal reactive power management
3. Topology of the wind turbines affects MVMO convergence
0 100 200 300 400 500 600 700 800 900 1000
3,8
4,0
4,2
4,4
4,6
4,8
5,0
Activepowerlosses(MW)
Number of iterations
0 100 200 300 400 500 600 700 800 900 1000
0
5
10
15
20
25
30
Activepowerlosses(MW)
Number of iterations
Case 7Case 1
MVMO=Mean Variance Mapping Optimization

33. 33
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
Future Work
 Optimal design of NN structure
 Include forecasting error in the optimization
NN=Neural Network

34. 34
Publications
Book Chapters
 A.M.Theologi, J.L.Rueda, and M.Ndreko, “Optimal Compliance of Reactive Power
Requirements in Near-shore Wind Power Plants”, Application of Modern Heuristic Optimization
Techniques in Power and Energy Systems, Wiley Publishing
 J.L.Rueda, A.M.Theologi, “Optimal Power Flow Test Bed and Performance Evaluation of
Modern Heuristc Optimization Algorithms”, Application of Modern Heuristic Optimization
Techniques in Power and Energy Systems, Wiley Publishing
Journal Paper
 A.M.Theologi, and J.L.Rueda, “MVMO-based Approach for Coordinated Operation of Reactive
Power Sources in Offshore Wind Farms”, Swarm and Evolutionary Computation
Conference Papers
 A.M.Theologi, and J.L.Rueda, “Optimal Management of Reactive Power Sources in Far-
offshore Wind Power Plants”, Proceedings of the IEEE Manchester PowerTech 2017,
Manchester, UK, June 2017
 A.M.Theologi, and J.L.Rueda, “Short-term Wind Speed Forecasting for Optimization in
Offshore Wind Farms”, Proceedings of the IEEE ISGT Latin America 2017, Quito, Equador,
September 2017

35. 35
Thank you for your attention!

36. 36

37. 37
Line Loadings
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

38. 38
Transformer Loadings
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work

39. 39
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work
Forecasting Error
Actual vs. Predicted Wind Speed
Hour
WindSpeed(m/s)
Predicted speed
Actual speed

40. 40
Comparison Case 1 & 2
Transformers Tap Positions
Current
operating
point
Predicted
time
horizon
2 4 6 8 10 12 14 16 18 20 22 24
-2
-1
0
1
2
Tappositions
Onshore Transformer A
Onshore Transformer B
Time ( h )
2 4 6 8 10 12 14 16 18 20 22 24
-5
-4
-3
-2
-1
0
Time ( h )
TapPositions
Offshore Transformer T1
Offshore Transformer T2
Min.
(Losses +Tap changes)
Min.
(Losses)
Motivation
& Research Tasks
Research
approach
Wind Speed
Forecasting
MVMO
Optimization
Grid Code
Requirements
Results
Conclusions
& Future Work